The present topography of Tian Shan is related to the India-Asia collision, whereas the mechanisms for the topographic growth of Tian Shan remain at the center of debate partly due to the poorly constrained onset timing of the Cenozoic tectonic deformation. Our new apatite (U-Th)/He (AHe) data in the northern Chinese Tian Shan suggest that rapid cooling commenced at Ma along the northern margin, which is consistent with published fission-track data from the same area and confirms that the youngest component of fission-track has been totally annealed. Moreover, AHe data from the interior mountain suggest that enhanced cooling began at Ma, which is slightly older than that from the northern edge of the mountain. Although thermochronological data suggest that both the interior and northern margin of the northern Chinese Tian Shan have undergone rapid cooling since the late Oligocene-early Miocene, well preservation of relict low-relief surfaces along the northern rim of the northern Chinese Tian Shan and published thermochronological data indicate that the northern Chinese Tian Shan may have experienced differential exhumation in the Cenozoic. Combining the thermochronological and geomorphological evidence, we propose a progressive northward growth model for the northern Chinese Tian Shan. During the late Oligocene and early Miocene, compressive deformation derived from the India-Asia collision have arrived at the northern Tian Shan to reactivate both the interior mountain and its northern margin, while intense exhumation was concentrated in the interior mountain range. Then, deformation extended northward into the foreland basin that may be initiated in the middle-late Miocene.
The Tian Shan extends east-west for over 2500 km through central Asia (Figure 1(a)) with peaks exceeding 7000 m. The present-day tectonics and geomorphology of the range are driven by crustal shortening related to the India-Asia collision [1, 2]. Various tectonic models have been proposed to be responsible for the uplift of the mountain range, including the subduction of basins on both sides , northward movement and clockwise rotation of the Tarim Basin , the northward indentation of the Pamir [5, 6], and mantle convection beneath the mountain range . Although the present-day structures of active faults and folds, surface ruptures of historical or prehistoric large earthquakes, and modern seismicity are widely distributed along the northern and southern flanks of the mountain ranges, and even in the interior mountains, it remains incompletely understood on when the deformation commenced and how it propagated in the Cenozoic. The timing of initiation of deformation in the Chinese Tian Shan is estimated to be ranging from the late Oligocene to Pleistocene (e.g., [2, 8–24]). Such a large difference could be either derived from different methods used to estimate the initial deformation timing or from the nonsynchronous activities due to the large area of the Tian Shan or outward tectonic propagation. For the northern Chinese Tian Shan, despite minor differences in the magnetostratigraphic results, the Cenozoic deposits were consistently defined to be from the early Miocene to the Quaternary, and the change in sediment accumulation rate or facies transition provides evidence for mountain building during the middle to late Miocene (e.g., [11, 13, 22, 23, 25–27]). However, due to the lack of continuous sections of the deposits older than Miocene in the Junggar and Tarim foreland basins, the pre-Miocene mountain building history of the Chinese Tian Shan has been poorly studied. Meanwhile, sediment accumulation rate and facies transition in basins may be controlled by climate change, such that nondeposition evidence, such as low-temperature thermochronological data, would provide valuable information on the development history of the adjacent mountains and can be used to check the reliability of the evidence derived from foreland basins.
Although published thermochronological data have been reported widely across the western part of the Tian Shan (Kyrgyz Tian Shan), which reconstruct the Cenozoic denudation history and the Oligocene-Miocene reactivation (e.g., [28–35]), as to the Chinese Tian Shan, thermochronological data are limited along the northern and southern margins of the orogen and along the Dushanzi-Kuche (Du-Ku) and Urumqi-Korla (Wu-Ku) highways (e.g., [9, 10, 14, 15, 17–20, 29, 36–48]). In general, most of the published thermochronological samples have Mesozoic ages, which suggests that much of the Chinese Tian Shan has experienced modest unroofing (<3-5 km) through the Cenozoic time . Only limited areas along the southwestern Chinese Tian Shan and the interior mountain range along the Du-Ku highway have undergone significant amounts of exhumation to denudate the samples with late Cenozoic low-temperature thermochronological ages to the surface [9, 10, 14, 19, 20, 41]. These young Cenozoic thermochronological ages provide the minimum boundary on the initial deformation timing.
One of the robust thermochronological evidence of the initiation of intense Cenozoic deformation in the Chinese Tian Shan came from a transect along the Manas River on the northern flank of the range . They reported a young component AFT age of Ma to mark the initiation of Cenozoic deformation in the northern Chinese Tian Shan [15, 19]. However, as the broad fission-track length distributions with numerous short tracks suggest that these samples were not totally annealed, then even the young component age may predate the initiation of deformation. These samples may experience later deformation that has not been recorded by the young component ages for the fission-track system yet. Thus, a thermochronological system with lower closure temperature, like the apatite (U-Th)/He dating method, would potentially provide constraints on the precise timing of the initial deformation.
In order to better constrain the onset timing of mountain building in the northern Chinese Tian Shan, we conducted an apatite (U-Th)/He thermochronological study along two sections (Figure 1(b)). One (Manas section) was collected along the Manas River and the other (Guertu section) was along an elevation profile in the interior mountain range. Our thermochronological results reveal an episode of rapid exhumation initiated in the late Oligocene-early Miocene in the interior mountain range and at the northern margin of the Chinese Tian Shan.
2. Geological Setting
The ancestral Tian Shan formed during multiple accretions of island arcs and Precambrian continental terranes during the Paleozoic, which was then reactivated during the Mesozoic [8, 18, 24]. The pre-Cenozoic evolution of the Tian Shan was dominated by mainly two accretionary events, including a late Devonian to early Carboniferous collision between the Tarim and central Tian Shan terranes and a late Carboniferous to early Permian collision between the amalgamated Tarim-central Tian Shan block and island arc complex lay to the north [8, 24]. These two major accretionary events produced two sutures that divided the Chinese Tian Shan into three parts, the south, central, and north Tian Shan. The southern suture, which is also named as the Nalati fault, extends along a northward-dipping metamorphic zone with several ophiolitic slices . It is an active fault with both thrusting and sinistral strike-slip components in the late Quaternary [49, 50], whereas its Cenozoic deformational history has been poorly constrained. To the north of the southern suture, the central Tian Shan block narrows eastward in a triangular shape and consists of widespread Carboniferous granite and early Permian rhyolitic and basaltic dikes .
The central and north Tian Shan is divided by the west-northwest-striking northern Tian Shan fault (F1) (Figures 1 and 2). This fault is also known as the Bo’a fault for the part in China  and the Dzhungarian (or Junggar) fault in Kazakhstan . It extends for more than 1000 km from the east end of the Balkhash Lake to the south of the Turfan Basin. Detailed field investigations and satellite imagery interpretations suggest that the segment west of Jinghe at 83°E is dominated by right-lateral strike-slip motion [51, 52]. The fault extends southeastward into the interior mountains that have been poorly investigated. Based on our preliminary satellite imagery interpretations, we suggest that the fault extends continuously in right-lateral strike-slip sense until ~84.3°E. To the further southeast, there is no evidence suggesting that the fault is active in the interior mountain during the late Quaternary. Active faults emerge again south of the Turfan Basin. Miocene apatite fission-track ages were reported near Haxilegen Pass along the Du-Ku highway, suggesting significant Cenozoic unroofing that may be related to the activity of the northern Chinese Tian Shan fault [14, 36, 37]. However, the role of the structure in building the north Tian Shan in the Cenozoic remains poorly understood.
The north Tian Shan is principally composed of the Devonian-Carboniferous arcs and accretionary complexes and the Jurassic-Cretaceous sedimentation [24, 53] (Figure 2). Most of the thermochronological samples collected in the north Tian Shan have Mesozoic apatite fission-track ages [36, 37], suggestive of modest exhumation during the Cenozoic. A linear valley and topographic difference define a boundary, referred to as the Yilian fault (F2), between the rugged topography to the south and relatively low-relief surfaces to the north (Figure 1(b)). It has never been reported previously and may represent a reactivated structure that has played an important role in defining the late Cenozoic deformation in the north Tian Shan. The north Tian Shan is bounded by the northern Tian Shan range-front fault (F3) to the north, which marks the most significant topographic difference between the northern Tian Shan and Junggar Basin (Figure 1(b)).
To the further north into the southern Junggar Basin, the most distinct tectonic features are three rows of east-west-striking anticlines. The Jurassic-Cretaceous and late Cenozoic strata are well exposed along the rivers that cut through these anticlines. The strata would provide stratigraphic records on the Cenozoic uplift history of the northern Chinese Tian Shan. Several magnetostratigraphic studies have been conducted to constrain the chronology of these Cenozoic sediments [11, 13, 22, 23, 25–27]. The sediments comprise five formations, including the Anjihaihe, Taxihe, Dushanzi, and Xiyu formations from bottom to top. The exposed oldest Anjihaihe formation was dated to be ca. 30.5-28 Ma [23, 25]. Multiple episodes of accelerated deformation since the early Miocene were proposed along the southern Junggar Basin based on magnetostraphically dated change in lithology or sedimentary accumulation rate and occurrence of growth strata [11, 13, 22, 23, 25, 27].
3. Methods and Sampling
As most apatite fission-track samples in the north Chinese Tian Shan have Mesozoic ages, it indicates that the region has undergone less than 3-5 km unroofing during the Cenozoic assuming a geothermal gradient of 20.0-30.0°C/km. It means that the samples have not been totally annealed, such that even the young component of fission-track ages may not record or predate the onset of Cenozoic exhumation. Alternatively, apatite (U-Th)/He (AHe) thermochronology has even lower closure temperature of °C [54, 55], which records the cooling and exhumation process in the upper crust. Thus, the apatite (U-Th)/He method may contribute to precisely dating the Cenozoic exhumation history in the northern Chinese Tian Shan. Typically, low-temperature thermochronological samples collected from a nearly vertical profile within a horizontal distance less than 3 km, or along a sedimentary section, may provide records on the exhumation history of the area . In the case of a region that has undergone a significant acceleration in exhumation rate, a break-in-slope on age-elevation or age-depth plots indicates the onset of rapid exhumation .
In order to quantify the exhumation history in the northern Chinese Tian Shan, we collected 11 samples along a horizontal profile (Manas section) and a vertical profile (Guertu section) (Figures 1(b) and 2 and Table 1). The Manas section consists of six samples (N-1 to N-4, N-6, and N-7) that were collected from the same localities of the samples from Hendrix et al. , lying immediately adjacent to the northern margin of the Tian Shan. These samples are from the Jurassic-Neogene sandstone or conglomerate. It is important to note that although our sampling locations are exactly the same as those of Hendrix et al. , the depositional timing of the sampled sediments is different from each other based on each modified Chinese geological map, as our geological map in Figure 3(a) is modified from the original Chinese geological map  and the sampling locations have been checked in the field. The Guertu section has five samples (KT-01 to KT-05) that cover an elevation interval of ~500 m from 2100 to 2600 m with an elevation difference of 150-200 m between two adjacent samples. These five samples were collected from the late Paleozoic granite just south of the north Tian Shan fault (F1).
All the apatite (U-Th)/He analyses were conducted at the Institute of Geology, China Earthquake Administration. Often, euhedral crystals were hand-picked from separated apatite grains, with average radii around 50 μm and mostly between 40 and 60 μm (see Table 2). The picked grains were individually wrapped in Pt tubes for later 4He and U-Th measurements. Helium content was analyzed using an Alphachron of Australian Scientific Instruments. 4He was degassed by Nd-YAG laser heating under 8 A current for 5 min. Each grain was heated at least twice to confirm that >99% 4He has been released. The released gas then was purified and spiked with a small pipette of 3He isotope dilution, and the 4He/3He ratio was measured using a quadrupole spectrometer. Following 4He extraction and measurement, the wrapped apatite grains were dissolved by nitric acid and U-Th contents were measured on an Agilent 7900 inductively coupled plasma-mass spectrometer (ICP-MS) using the isotope dilution method. Durango apatite fragments were analyzed as standards to check the analytical accuracy. More detailed analytical procedures can be found in .
As the form of the age-elevation profile and the time of the break-in-slope could be affected by various factors, such as rate and amount of cooling, thermal history prior to the onset of final rapid cooling, and geothermal gradient, a break-in-slope may provide a minimum constraint on the onset of rapid cooling . Thus, to further extract the cooling and exhumation history of the measured samples, inverse thermal historic modeling was conducted using the published code (QTQt, v5.4.0), which is based on the Bayesian transdimensional Markov Chain Monte Carlo (MCMC) inversion scheme . The present temperature was set to be , and the present-day temperature offsets were set to meet an atmospheric lapse rate of ~6°C/km. A geothermal gradient of °C was applied during modeling based on organic maturation data . For each sample, the AHe age of the oldest measured grain and 70°C with a ±range of the same values were applied as the center of the modeling time and temperature, respectively. We tested different parameters, such as the number of iterations and number of T-t points, to reach a thermal history with acceptable rates of time and temperature parameters and consistent low birth and death acceptance rates in QTQt. To derive a stable inverse model for each sample or profile, we run 400,000 iterations with the first 200,000 used as burn-in which will be discarded and subsequent 200,000 used to infer the thermal history. For the Guertu section, we modeled all five granite samples together. No other temperature-time constraints were added to the model. For the Manas section, as the samples were collected from sediments along a subhorizontal profile that is nearly perpendicular to the north-dipping monocline (Figure 3(a)), it means that these samples have experienced differential degrees of exhumation and interior deformation. Thus, we prefer to model the samples from the Manas section separately. Because Hendrix et al.  did not provide fission-track density and length information, fission-track data imposed in the inverse modeling are resampled in QTQt based on track-density data provided by Dumitru et al. . Additionally, the depositional ages with a relatively large uncertainty of about ±30 Ma based on Hendrix et al.  and our interpretations on the Chinese geological mapping results (Figure 3(a)) and a paleosurface temperature of °C were imposed to the models. Sample N-2 has only two grains with scatter AHe ages and no AFT dating such that we do not conduct inverse thermal history modeling for this sample.
Totally, 41 AHe ages are obtained for the 11 samples (Tables 1 and 2). Radiation damage may affect dating results; then, we should observe a positive or negative correlation between the ages and the effective U concentration (eU) . Most of our samples have relatively young late Cenozoic ages, and all the apatite grains have low eU values of less than 1 nmol (or <60 ppm) (Table 2). Thus, our samples may be not affected by radiation damage significantly. There is no apparent age-eU (nmol/g) and age-radius correlations observed for all the samples (except for KT-04) (Figure 4). It indicates that our AHe ages are not significantly correlated with the radiation damage and grain volume.
4.1. Manas Section
Six samples were collected from sandstone or conglomerate deposited during the Jurassic and Neogene (Figure 3(a)). Except for N-2, other samples were collected at the localities in  (Figure 3(b)). The geologic units and structures have been well investigated by Xinjiang Bureau of Geology and Mineral Resources (XBGMR)  (Figure 3(a)). The sampled Jurassic-Neogene layers are located immediately north to the northern Tian Shan range-front fault, forming a north-dipping monocline that has been slightly folded and displaced by a south-dipping blind fault. It suggests that these sediments were exposed to the surface by foreland flexure related to the northward thrusting of the northern Tian Shan range-front fault. Hendrix et al.  estimated the ages of the sedimentary layers ranging from ca. 230 Ma to ca. 80 Ma and reported apatite fission-track ages from ca. 186.9 Ma and ca. 29.7 Ma. The data display an overall pattern of younger AFT ages with increasing paleodepth (Figure 3(c)). The shallowest sample (M1) has an AFT age of ca. 186 Ma that is much older than its depositional age of ca. 80 Ma. In our updated geological map, sample M1 is deposited in the early Neogene, which means that its depositional age is even much younger (Figure 3(a)). The sample has not been annealed since the deposition, such that the AFT age only records the exhumation history of the provenance region. The AFT ages of the remaining four samples are younger than the depositional ages, whereas the broadly distributed single-grain ages and track lengths suggest that these samples have not been totally annealed. Based on the age of the youngest component of the lower three samples (M3, M4, and M5), they indicate that significant cooling and exhumation began at Ma [14, 15, 19].
The northernmost sample (N-1) has a relatively old mean AHe age of Ma derived from five grains with ages ranging from Ma to Ma (Tables 1 and 2). All the AHe ages are older than the depositional age, supporting that AHe ages have not been reset during the Cenozoic. Sample N-2 has only two grains that have AHe ages of Ma and Ma, respectively. To the further south, the remaining four samples have relatively consistent mean AHe ages ranging from ca. 13.8 Ma to ca. 22.0 Ma (Figure 3(c)). The AHe ages of samples N-4 and N-7 are slightly younger than the youngest component AFT ages of the corresponding samples M3 and M5 that were collected at the same localities from Hendrix et al. , respectively, while sample N-6 (M4) has much younger AHe age of Ma and the age difference between the AHe and AFT ages is larger, which may be affected by the chemical component difference. We have no definite interpretations for this. The AHe ages of all the grains (except for N-2(b)) from the five southern samples yield a mean age of Ma. If we exclude the relatively younger sample N-6, the other four samples would yield a mean age of Ma. In general, the northernmost sample with a much older AHe age and the five southern samples with consistent younger AHe ages suggest a break-in-slope in the age-depth plot, which indicates the onset of accelerated exhumation.
4.2. Guertu Section
The Guertu section has five samples collected along a vertical profile immediately south of the northern Tian Shan fault (Figure 5(a)). The surface trace of the fault appears fresh on the satellite imagery (Figures 5(b) and 5(c)). The fault is characterized as a continuous straight line that displaced the bedrock and remnant alluvial/fluvial terraces. Displaced stream channels, no significant vertical offset, and uphill-facing fault scarps suggest that the northern Tian Shan fault is dominated by right-lateral strike-slip movement in the late Quaternary. Another fault (F2) is inferred to terminate here based on the linearity of the geomorphic expression, whereas no geomorphic expression of the late Quaternary activity was observed on the satellite imagery and in the field investigations.
The highest sample (KT-01) has a mean AHe age of Ma derived from four apatite grains ranging from Ma to Ma (Figure 5(d) and Tables 1 and 2). In contrast, the lower four samples have much younger mean AHe ages ranging from Ma to Ma. Especially for samples KT-02, KT-03, and KT-05, the AHe ages of all the grains are concentrated between 25 and 31 Ma, with a mean AHe age of Ma. Sample KT-04 has a mean AHe age of Ma derived from four grains, which is relatively younger than that of the other three samples. The age distribution of KT-04 shows an apparent correlation with the grain radius (Figure 4(d)), and all four grains have relatively small grain sizes with an equivalent spherical radius of less than 40 μm. Thus, we interpret that the relatively younger AHe ages of KT-04 mainly resulted from the effect of grain size. Consistent AHe ages for the lower four samples and older age for the highest sample KT-01 indicate a distinct break on the age-elevation relationship, which indicates the onset of accelerated exhumation (Figure 5(d)).
4.3. Thermal History Modeling
Samples in the Manas section are modeled separately (Figure 3(d) and Figures S2–S11). In general, except for the shallowest sample (N-1), all the other samples have experienced similar two periods of distinct cooling history since the deposition. Inverse modeling results suggest that sample N-1 experienced continuous cooling prior to the deposition in the early Neogene, followed by a period of isothermal history at the shallow subsurface. Both AHe and AFT ages are apparently older than their depositional age, suggesting that the thickness of lateral sediments upon sample N-1 is not enough to anneal the sample. Thus, sample N-1 only records the exhumation history of its provenance region. Samples N-3, N-4, and N-6 show similar cooling/exhumation history. Inverse modeling results indicate that these samples experienced reheating since the deposition until about 30-20 Ma, which was followed by a period of rapid cooling with a consistent rate of ~2-4°C/Ma to the present. Assuming a geothermal gradient of 22.0°C/km based on organic maturation data and used in previous studies [14, 15], it would yield an exhumation rate of 0.10-0.18 km/Ma. The modeling maximum reheating temperature is around 100°C, most likely between 80 and 100°C. Modeling results of sample N-7 show a period of reheating between the depositional time and ~50-40 Ma that was followed by a period of rapid cooling until the present. The rapid cooling rate is constrained to be ~2-3°C/Ma. The maximum reheating temperature is over ~110-120°C, suggesting that this sample has been totally annealed. We will discuss the cause of the difference in modeling results between N-7 and the other three samples (N-3, N-4, and N-6) in the next section. The predictions of both AHe and AFT ages are generally consistent with the observations, suggesting the validity of the inverse modeling results (Figures S2 to S11).
The inverse modeling results for the Guertu section show two periods of distinctly different cooling history since the late Cretaceous (Figure 5(e)). Prior to ~28 Ma, the samples experienced relatively isothermal history or slight reheating since the late Cretaceous, which was followed by a period of rapid cooling and exhumation initiated at ~28 Ma and continues until the present with a consistent cooling rate of °C/Ma. Assuming a reasonable geothermal gradient of 22.0°C/km, the rapid cooling rate is equivalent to an exhumation rate of ~0.1 km/Ma. To validate the inverse modeling results, the predicted and observed mean and single-grain AHe ages are plotted together (Figures S12 to S14). For mean AHe ages, all the five predictions are well consistent with observations. While for single-grain ages, the predictions of the highest sample (KT-01) are younger than the observations. For the lower younger samples (KT-02 to 05), AHe age predictions of a few grains are slightly off the observations.
5.1. Onset of Rapid Cooling/Exhumation
Our AHe data from the Manas section display an overall pattern of older age for the shallowest sample (N-1) and relatively consistent younger ages for the deeper five samples (Figure 3(c)). The shallowest sample has a mean AHe age of Ma that is much older than the sample’s depositional age (early Neogene). It suggests that the AHe ages have not been reset during the Cenozoic. Sample N-2 has only two replicate AHe ages. One is Ma, and the other is Ma. The younger one is consistent with the ages from the deeper samples. It indicates that N-2 may be located immediately above the base of the fossil AHe partial retention zone (PRZ). The deeper four samples (N-3, N-4, N-6, and N-7) have relatively consistent AHe ages from a depth of ~0.5 km to ~5 km. It suggests that all the grains have been completely reset and went through the fossil AHe PRZ rapidly. Thus, the base of the fossil AHe PRZ is located between samples N-2 and N-3. Nevertheless, the consistent AHe ages for the deeper four samples and a break-in-slope on the age-depth plot indicate a phase of rapid exhumation initiated at or slightly prior to Ma. Inverse thermal history modeling results of the lower three samples (N-3, N-4, and N-6) support this explanation, which reveals a reheating episode since the deposition time and a distinct rapid cooling episode since ca. 30-20 Ma. The reheating process may be related to the deposition of the Cretaceous and early Cenozoic units. The rapid cooling rate since ca. 30-20 Ma is estimated to be ~2-4°C/Ma, which is corresponding to an exhumation rate of 0.10-0.18 km/Ma assuming a 22.0°C/km geothermal gradient [14, 15]. The thermal history modeling results of these three samples are different from that of sample N-7, which has experienced rapid cooling since the early Cenozoic (~50-40 Ma) (Figure S10). The depth of N-7 is estimated to be >5 km, which would derive a maximum reheating temperature of ~110°C. It means that both the AHe and AFT ages of sample N-7 may have been totally reset, such that the early thermal history of the sample has been erased. Thus, the thermal history below the base of PRZ is not reliable, and we infer that the onset timing of the latest rapid cooling derived from the inverse modeling of sample N-7 is artificial. Combining the age-depth plot and inverse thermal history modeling results, we suggest that rapid cooling/exhumation in the northern margin of the Chinese Tian Shan began at or slightly prior to Ma, which is well consistent with the AFT results from Hendrix et al. . Relatively consistent AHe ages and the youngest component AFT ages of the deeper three samples support that the fission-tracks formed in the young component grains were fully annealed, such that the AFT ages of the young component grains can be explained as the time when the significant cooling began . The tilting and exhumation of the sampling horizontal strata are apparently related to the fault activity along the northern Tian Shan range-front fault (F3) (Figures 3(a) and 3(b)).
Additionally, there is yet little direct evidence on initial Cenozoic deformation history within the interior northern Chinese Tian Shan. Although the young apatite fission-track ages along the Du-Ku highway postdate the initiation of exhumation in the area, most available thermochronological data are distributed on the northern and southern margins of the range. Few data have been reported about the deformation history in the interior mountain range because of the inaccessibility of most areas of the Chinese Tian Shan. However, several large-scale structures are developed in the northern Chinese Tian Shan (Figure 1(b)) that may have undergone different deformation histories. This is supported by field observations and satellite image mapping, which suggest that low-relief surfaces are preserved along the northern rim of the northern Chinese Tian Shan [63–65]. The roles of these faults in the deformation history of the northern Chinese Tian Shan have been poorly addressed. The samples from the Guertu section were collected from the interior mountains and would provide critical information on the tectonic evolution of the interior northern Chinese Tian Shan.
As the samples from the Guertu section were from a vertical profile within 2.5 km in horizontal, a break-in-slope on the age-elevation plot would indicate a clear change in the exhumation and cooling rate . Except for sample KT-04 with a relatively younger mean AHe age of Ma that is affected by the grain size as we mentioned before, the other three lower samples (KT-02, KT-03, and KT-05) have a consistent mean AHe age of Ma (Figure 5(d)). It suggests that these samples below the break rapidly passed through the PRZ. The sample above the break (KT-01) has a much older mean AHe age of Ma, indicative of a relatively slower exhumation rate during the late Cretaceous and the Oligocene. We interpret that the base of a fossil PRZ was located between samples KT-01 and KT-02 and a phase of rapid cooling and exhumation initiated at or slightly prior to Ma. Inverse thermal history modeling results suggest a stable or slightly reheating episode between the late Cretaceous and ~28 Ma, followed by an episode of rapid monotonic cooling with a rate of °C/Ma (~0.1 km/Ma assuming a 22.0°C/km geothermal gradient) to the present (Figure 5(e)). Thus, both age-elevation plot and inverse thermal history modeling indicate that the Guertu section has experienced rapid cooling/exhumation since ~28 Ma. As faults F1 and F2 converge near the Guertu section and all the samples were collected from the hanging wall of the faults, we attributed the rapid cooling/exhumation to the fault activity of fault F1 or/and F2.
Conclusively, our new AHe data indicate that rapid cooling and exhumation initiated in the late Oligocene and early Miocene in the interior and the northern margin of the northern Chinese Tian Shan, respectively. The interior mountain range may initiate slightly earlier than the northern margin of the northern Chinese Tian Shan. Considering the AHe dating error, it might be hard to distinguish these two cooling events from each other.
5.2. Late Oligocene-Early Miocene Exhumation Episode
In the northern Chinese Tian Shan, our new thermochronological data from both the Manas and Guertu sections indicate that the initiation of building of the Tian Shan commenced in the late Oligocene-early Miocene. The break-in-slope on the age-depth plot of the Guertu section shows robust evidence for an accelerated exhumation at Ma. It is consistent with the dating results of the youngest fission-track component from the same section, which place the initiation of mountain building at about Ma . Corresponding to the rapid exhumation in the northern Chinese Tian Shan, increase in sedimentation rate, changing in sedimentary facies, and anisotropy of magnetic susceptibility indicate that the northern Chinese Tian Shan experienced a phase of uplift in the early Miocene [25, 68, 71]. This phase of rapid exhumation is also recorded by AFT data and thermal history modeling from the Bogda mountain east of Urumqi (e.g., [67, 70]). Along the Du-Ku highway, numerous thermochronological data have been achieved with most dating ages ranging from the late Mesozoic to the late Miocene [14, 18, 43, 48]. The late Oligocene-early Miocene is revealed by thermal history modeling of a single bedrock sample south of the Haxilegen Pass [14, 43, 48].
In the southern Chinese Tian Shan, the late Oligocene-early Miocene is recorded by both sedimentological and thermochronological data (Figure 6). Based on sedimentary facies changing constrained by biostratigraphy and magnetostratigraphy in the Kuche basin, Yin et al.  firstly suggest that initial crustal shortening may have occurred at 21-24 Ma. Then, based on detailed magnetostratigraphic survey and provenance analyses, Huang et al.  and Tang et al.  affirm the early Miocene uplift of the southern Chinese Tian Shan. Similarly, a marked increase in the accumulation rate at 20-18 Ma in the northwestern Tarim basin is recorded by magnetostratigraphic dating, which is attributed to the southward propagation of the Tian Shan related to the strike-slip motion of the Talas-Fergana fault . Meanwhile, growing thermochronological data from the southern Chinese Tian Shan support the late Oligocene-early Miocene crustal shortening. The late Oligocene-early Miocene estimates for the onset of deformation have been proposed for the Kokshaal Range in the southwestern part of the southern Chinese Tian Shan, which is suggested to be connected to the Pamir indentation and strike-slip movement along the Talas-Fergana fault [20, 38]. Detrital apatite fission-track samples from the western Kuche basin indicate that the uplift of southern Tian Shan initiated in the late Oligocene-early Miocene [17, 74], which is consistent with the sedimentological records [2, 72]. Recently, with lower closure temperature (°C), the apatite (U-Th)/He method has been successfully applied to reveal the exhumation history of the sediments in the northwestern Tarim basin, which suggests that exhumation started at around 25-20 Ma [75–77].
Moving west to the Kyrgyz Tian Shan, the late Oligocene-early Miocene is recorded in both the interior mountain and piedmont (Figure 6). The Kyrgyz Tian Shan experienced a period of tectonic quiescence between the late Mesozoic and early Cenozoic based on the sedimentary record and thermochronological data [29, 31, 32, 34, 35]. Then, the crustal shortening related to the India-Eurasia collision has propagated into the region in the late Oligocene and early Miocene to reactivate the inherited weak structures, which have been recorded in several basins and mountain ranges. Based on low-temperature thermochronological analyses in the Atbashi Range, Glorie et al.  constrained the initial building of the modern Tian Shan to the late Oligocene (~30-25 Ma). The onset of increasing depositional rates and deposition of coarse clastic sediments has been dated to be 25-20 Ma by thermochronological and magnetostratigraphical data in the central Terskey Range south of the Issyk Kul Lake [33, 69]. North of the Issyk Kul Lake, published apatite fission-track and (U-Th)/He data from the Trans-Ili (or Zaili) Range indicate a late Oligocene-early Miocene increase in cooling/exhumation, which has been interpreted as incipient building of the modern Tian Shan as a result of stress propagation from the India-Eurasia collision . Furthermore, along the southern side of the Kokshaal Range in the southwestern Chinese Tian Shan, rapid exhumation was estimated to commence at ~24 Ma . The Talas-Fergana fault is a large-scale right-lateral strike-slip fault in the region and is of great significance in terms of understanding the history and kinematics of the Pamir indentation. Based on apatite fission-track and (U-Th)/He data collected from the fault zone and adjacent regions, strike-slip motion along the Talas-Fergana fault was estimated to be commenced at ~25 Ma [28, 78, 79].
Although the construction of the modern Tian Shan commenced in the late Oligocene-early Miocene period, a growing consensus is that the Tian Shan underwent a widely distributed exhumation around 10 Ma (e.g., [14, 18, 20, 32–34, 38, 68, 71, 78]). The accelerated phase of exhumation around 10 Ma is widely distributed throughout the Tian Shan, particularly along its northern and southern piedmonts (Figure 6) (e.g., [9, 11, 12, 21, 22, 29, 33, 43, 44, 64, 71, 75, 77, 80]). Synthesis of the widely distributed late Oligocene-early Miocene and middle-late Miocene deformation throughout the Tian Shan in Central Asia allows us to suggest that the entire Tian Shan may have experienced a similar two-pulse deformation pattern during the Cenozoic: the building of the modern Tian Shan, as a result of stress propagation related to the India-Eurasia collision, commenced between the late Oligocene and the early Miocene, then intensified during the middle-late Miocene.
Although the India-Asia collision began in the Paleocene ( and reference therein), the resulting deformation in Asia occurred in several discrete periods (e.g., ). The late Oligocene-early Miocene deformation is reported across the Tibetan Plateau. Rapid exhumation began at 30-25 Ma along the Longmen Shan at the eastern margin of the plateau  and at ~24-18 Ma along the Anninghe-Xiaojiang fault to the south . Central Tibet may also uplift in the late Oligocene based on the discoveries of low-elevation tropical fossils and updated magnetostratigraphic dating results [84, 85]. Upper-crustal shortening and rapid exhumation within the Hoh Xil Basin occurred in the Oligocene and early Miocene . Major transpressional lateral-slip movement on the Eastern Kunlun fault began at ~25 Ma . On the other hand, the late Miocene deformation is also widely distributed in the Tibetan Plateau. Rapid exhumation along the mountain ranges and corresponding river incision in the southeastern Tibetan Plateau occurred at ~15-10 Ma (e.g., [82, 88]). A transition from lateral-slip movement on the Red River fault system is recently dated to be as early as the middle Miocene . The clockwise rotation of crustal around the Eastern Himalayan Syntaxis began at ~15-10 Ma . The middle-late Miocene deformation is widely recorded by thermochronology and deposition in the northern Tibetan Plateau [59, 91]. Though the details on the mechanisms of these two phases of tectonism remain uncertain, it seems possible in terms of a fundamental shift in the dynamics of mountain building from the late Oligocene-early Miocene to the late Miocene. It is also possible that the two phases of deformation resulted from continuous outward propagation of crustal thickening [86, 92]. Regardless, the approximate synchroneity of the two phases of accelerated deformation in the Tibetan Plateau and the Tian Shan suggests that the deformation propagated rapidly across the stable Tarim block. A strong cratonic lithosphere would introduce rapid propagation of strain, and the deformation concentrates along the edges of the strong region .
5.3. Implications for Growth of the Northern Chinese Tian Shan
Our new thermochronological data from two sections indicate that both the northern range front and the interior mountain of the northern Chinese Tian Shan underwent a phase of accelerated exhumation in the late Oligocene-early Miocene period. However, geomorphological observations indicate that the interior mountain south of F2 is dominated by rugged topography, while large fragments of low-relief surfaces are preserved along the northern rim of the northern Chinese Tian Shan [63–65]. Parts of the low-relief surface are covered by Neogene grey lacustrine shale and siltstone layers. Based on the occurrence of lacustrine fauna, Jolivet et al.  prefer to regard them as the Taxihe Formation with an older depositional age of ~20-15 Ma. Although these Neogene deposits are now distributed at an elevation ranging from about 1850 to 2500 m, they must be at a similar elevation as the lacustrine deposits in the Junggar Basin at the time when they were deposited. It means that these lacustrine layers were perched to the current elevation sometime after ~20-15 Ma . Based on the well preservation of the low-relief surface along the northern rim and rugged topography in the interior mountain, we suspect that the northern Chinese Tian Shan may experience differential deformation or exhumation along the north-south direction.
There are two possible explanations for the forming of the low-relief surfaces along the northern rim of the northern Chinese Tian Shan. On the one hand, the low-relief surfaces may represent local planation surfaces formed in the Jurassic or even earlier. As we suggest above, at least parts of the low-relief surfaces are covered by Mesozoic and Cenozoic sediments. These sediments are easily eroded such that the planation surfaces are preserved. In this case, the different geomorphological expressions in the northern Tian Shan may be mainly controlled by lithology. On the other hand, a unique widespread surface may exist prior to the Oligocene, and only the northern part along the northern Tian Shan has been preserved due to differential erosion. This is supported by low-temperature thermochronological data. Although the Cenozoic AFT ages were reported along the Manas River and its adjacent regions [15, 36, 45, 94], most thermochronological samples collected from the base bedrock along the northern margin of the mountain have much older AFT ages (>100 Ma), indicative of modest Cenozoic exhumation . In contrast, the late Cenozoic (<20 Ma) AFT and AHe ages are reported south of the Haxilegen Pass, which suggest that the interior mountain range of the northern Chinese Tian Shan has undergone significant Cenozoic unroofing [14, 18, 43]. Considering distinctly younger thermochronological ages in the interior mountain than that from the areas to the north and relict low-relief surfaces distributed along the northern rim, we infer that the northern Chinese Tian Shan may undergo differential unroofing during the Cenozoic. In general, the areas south of the northern Tian Shan fault (F1) have experienced at least 3-5 km of late Cenozoic unroofing, whereas the areas to the north have undergone modest total unroofing as the apatite fission-track ages have not been reset.
Combining the onset timing of the deformation in the northern Chinese Tian Shan and geomorphological observation, we propose a roughly two-pulse deformation model (Figure 7). During the late Oligocene and early Miocene, compressive stresses derived from the India-Asia collision have been transmitted to the northern Chinese Tian Shan, reactivating the core and northern margin of the northern Chinese Tian Shan. This episode of deformation was also recorded in the foreland basin in the southern Junggar Basin. Based on magnetostratigraphic data in the Jingou River, a distinct increasing sedimentation rate and a quick shift of depositional environment from the lacustrine facies in the Anjihaihe Formation to the deltaic facies in the Shawan Formation occurred during the early Miocene [23, 25, 68]. The Anjihaihe Formation is mainly composed of greyish green mudstone, whereas the overlying Shawan Formation is dominated by dark brown mudstone and conglomerate. It is also recorded by a regional angular unconformity marked by the deposition of Oligocene conglomerates in the Urumqi and Turfan Basins [8, 24]. During this episode, the deformation was concentrated in the interior mountain, which underwent significant exhumation then, whereas the regions along the northern rim of the northern Chinese Tian Shan experienced less exhumation and likely acted as a foreland basin that is similar to the present southern margin of the Junggar Basin. Otherwise, if the northern Tian Shan range-front fault (F3) acted as a major structure that controlled the uplift of the northern Tian Shan during the early Miocene, with a distance of only ~30 km to the northern Tian Shan range front, the Jingouhe River magnetostratigraphic section would be dominated by alluvial conglomerates like the present Xiyu Formation. Furthermore, as we mentioned above, the Neogene lacustrine layers with depositional ages of about 20-15 Ma are preserved upon the relict low-relief surfaces, which means that these low-relief surfaces were at a similar low elevation as the Junggar Basin then. Thus, the significant topographic growth is limited to the interior mountain during this stage. Then, during the middle-late Miocene (~10 Ma), tectonic deformation propagated basin-ward to the north. Mountain building was mainly along the northern Tian Shan range-front fault (F3) then and deformation continuously extended into the foreland basin to form several E-W-trending anticlines in the southern Junggar Basin. This episode of deformation occurred during the middle or late Miocene that is widely recorded by sedimentary records in the foreland basin based on accelerated depositional rate, growth strata, and onset of molasse deposition [11, 13, 22, 23, 25–27]. Thus, the present tectonics and geomorphology of the northern Chinese Tian Shan resulted from the reactivation of a pre-Cenozoic accretionary orogen and progressive northward growth of the Cenozoic deformation.
New apatite (U-Th)/He data from two sections reveal that enhanced cooling and exhumation commenced at or slightly prior to Ma and Ma for the northern margin and interior mountain of the northern Chinese Tian Shan, respectively. It suggests that tectonic deformation related to the India-Asia collision have arrived at the northern Tian Shan as early as the late Oligocene to the early Miocene. Moreover, the interior mountain may begin to uplift slightly earlier than the northern edge. Combining thermochronological and geomorphological evidence, we propose a two-pulse tectonic deformation model for the northern Tian Shan. Firstly, crustal shortening derived from the India-Asia collision arrived at the northern Tian Shan in the late Oligocene-early Miocene, while deformation was concentrated on the interior mountain range then. Then, tectonic deformation extended northward into the foreland basin, and the relict low-relief surfaces were uplifted during the middle-late Miocene to the present.
All the data supporting the results of this study have been presented in the paper.
Conflicts of Interest
The authors declare that they have no conflicts of interest.
This work was supported by the National Natural Science Foundation of China (42030301 and 41902199) and National Nonprofit Fundamental Research Grant of Institute of Geology, China Earthquake Administration (IGCEA2117). We gratefully thank Edward Sobel and Marc Jolivet for their helpful comments on the earlier version of the manuscript.