Cooling and exhumation of the Late Paleozoic Tulasu epithermal gold system, Western Tianshan, NW China: implications for preservation of Pre-Mesozoic epithermal deposits

Epithermal gold deposits are rarely well preserved in pre-Mesozoic terranes because of their low-temperature mineralization at shallow crust levels, and they are easily destroyed by subsequent erosion or depleted by tectonic events. However, several significant Late Paleozoic epithermal gold deposits have been found in the Tulasu volcanic basin in NW China, forming one of the largest gold districts in the western Tianshan Orogen. Here, we report a new 40Ar/39Ar age from a monzonite porphyry enclave hosted in andesite and apatite fission track data for 10 volcanic rocks from the Tulasu basin. These data, combined with the previous dataset, are used to perform inverse thermal modelling to quantify the district's cooling and exhumation history. Our modelling indicates a phase of burial reheating during Late Paleozoic sedimentation following mineralization, a subsequent rapid exhumation in the Jurassic to Early Cretaceous (c. 196–128 Ma), and a slow exhumation to the present. The Mesozoic exhumation is likely related to the far-field effects of the Cimmerian orogeny along the southern Eurasian margin. Therefore, we suggest that the rapid burial by thick sediments and the slow protracted exhumation after mineralization were crucial for the preservation of the Paleozoic epithermal gold system at Tulasu.

Epithermal gold deposits form at shallow crustal levels of <1.5 km, low temperatures of <200°C and low pressures of 10-50 MPa in volcanic arc or back-arc extensional settings (Hedenquist et al. 2000;Sillitoe and Hedenquist 2003). However, such conditions are generally unfavorable for their preservation because of the longterm and intricate geological processes after mineralization (Simmons et al. 2005). As a result, most of the world's epithermal gold deposits are geologically young, such as the many Mesozoic and Cenozoic epithermal gold deposits in the circum-Pacific and Tethyan belts (Corbett and Leach 1998;Cooke and Simmons 2000;Kerrich et al. 2000), and the preservation of Pre-Mesozoic deposits is relatively rare. Nonetheless, exceptions exist in the Paleozoic western Tianshan Orogen, which lies in the southwestern Central Asian Orogenic Belt ( Fig. 1a; Han et al. 2011). The orogen is well endowed with numerous pre-Mesozoic epithermal gold deposits that formed during Late Paleozoic accretionary and collisional processes, including two representative epithermal gold systems: the Almalyk district in eastern Uzbekistan (overprinting a giant porphyry copper system) and the Tulasu volcanic basin in NW China (Fig. 1b;Seltmann et al. 2014;Zhao et al. 2014a). However, these pre-Mesozoic epithermal gold deposits should probably have been wiped out by the complex and episodic Mesozoic to Cenozoic exhumation and erosion processes, as revealed by previous thermochronological studies (Glorie and De Grave 2016;Jepson et al. 2018;Zhang et al. 2019). How did these pre-Mesozoic epithermal gold deposits survive the exhumation and erosion processes of the ancient orogenic belt? This question remains enigmatic and requires a thorough exploration.
The Tulasu volcanic basin is located c. 30 km NE of Yining City (Fig. 1b), and it hosts several important Paleozoic epithermal gold deposits, i.e., Axi, Tawuerbieke and Qiabukanzhuota (Hart et al. 2003;Pirajno et al. 2011;Zhao et al. 2014a;Gu et al. 2016). The basin's exploration history can be traced back to the late 1980s when the No. 1 Geological Team of the Xinjiang Bureau of Geology and Mineral Resources carried out a 1:50 000 litho-stratigraphic mapping and geochemical investigation to successively discover the three deposits (XBGMR 1993). Since then, multidisciplinary research projects have sought to unravel the metallogenesis of this large epithermal gold district through the study of the deposits' geology and geochemistry (e.g. Zhai et al. 2009;Zhao et al. 2014b;Gu et al. 2016;An and Zhu 2018), petrology (e.g. Tang et al. 2009;An et al. 2013;Yu et al. 2016;Zhao et al. 2020), mineralogy (e.g. Liu et al. 2018;Zhang et al. 2018) and geochronology (e.g. Zhao et al. 2014c;Peng et al. 2017;Liu et al. 2019). So far, these studies have failed to address the crucial question of the epigenetic district's post-mineralization exhumation history and its preservation mechanism.
Apatite fission track (AFT) dating is one of the most widely used low-temperature thermochronological techniques, and it has been applied in economic geology to reconstruct the post-mineralization history of a deposit (McInnes et al. 2005). Due to its thermal sensitivity to low temperatures (c. 60-120°C; Gallagher et al. 1998), AFT dating can help to elucidate the low-temperature thermal history and exhumation process of an epithermal deposit, as well as help to estimate its preservation potential (McInnes et al. 2005). In this paper, we applied the AFT technique in combination with both the 40 Ar/ 39 Ar dating method and numerical modelling to better quantify the cooling and exhumation history of the Late Paleozoic Tulasu epithermal gold district and to provide new insights into the preservation of pre-Mesozoic epithermal gold deposits in ancient orogenic belts.

Geological setting
The 2000 km long western Tianshan orogeny experienced multiple Paleozoic subduction, accretion and collision events that involved complex geodynamic interactions between several branches of the Paleo-Asian Ocean, including the Terskey, Turkestan and Junggar Oceans (Windley et al. 2007;Biske and Seltmann 2010;Xiao et al. 2013Xiao et al. , 2015Han and Zhao 2018). During the Mesozoic and Cenozoic, the ancestral orogen was modified by the intense reactivation from intracontinental deformation, episodic uplift and exhumation processes that reworked this magnificent mountain range (Jolivet et al. 2010;Glorie and De Grave 2016;Jepson et al. 2018;Wang et al. 2018;Zhang et al. 2019). Bounded by the Nikolaev Line-North Nalati and the Turkestan-Atbashi-Inylchek fault zones, the western Tianshan Orogen is generally divided into the North, Middle and South Tianshan tectonic units (Fig. 1a;Zonenshain et al. 1990;Biske and Seltmann 2010;Xue et al. 2014). These units are dextrally crosscut by the Talas-Fergana fault.
The North Tianshan comprises mainly Precambrian metamorphic basement with overlying Paleozoic volcanic and sedimentary formations (Mikolaichuk et al. 1997;Gao et al. 1998;Xue et al. 2014). The northward subduction of the Terskey Ocean beneath the North Tianshan occurred in the Late Ordovician, followed by the accretion of Middle Tianshan to North Tianshan, which gave rise to the Nikolaev Line-North Nalati fault zone (Lomize et al. 1997;Glorie et al. 2010;Xue et al. 2014). To the west of the Talas-Fergana fault, the Uzbekistan Middle Tianshan is represented by the Late Paleozoic Beltau-Kurama arc belt, which developed on the Precambrian basement and overlying Early Paleozoic clastic rocks (Seltmann et al. 2014). The Middle Tianshan narrows to the east and contains a Precambrian basement and overlying Paleozoic volcanic-sedimentary sequence (Han and Zhao 2018;Zhao et al. 2019). Intense magmatism occurred in the Middle Tianshan that concentrated in the Silurian-Early Devonian and Late Carboniferous-Early Permian in response to the northward subduction and closure of the Turkestan Ocean, respectively Han and Zhao 2018). The South Tianshan represents a Late Paleozoic accretionary belt that has experienced intense thrusting and folding during the closure of the Turkestan Ocean (Biske and Seltmann 2010;Xiao et al. 2013).
Tectonically, the Tulasu epithermal gold district occurs in the western part of the Chinese North Tianshan (Fig. 1b). The gold deposits are hosted by a fault-controlled volcanic-sedimentary basin that developed on a continental arc which formed from the southward subduction of the Junggar Ocean beneath the North Tianshan during the Late Paleozoic (Chen et al. 2011;Xiao et al. 2013;Zhao et al. 2014a). The Junggar Ocean is suggested to have opened as early as the Late Cambrian (Ren et al. 2014(Ren et al. , 2018b, and the Bayingou and Lucaogou ophiolitic mélanges along the northern edge of Chinese North Tianshan are thought to be the remnants of the Junggar oceanic crust (Xu et al. 2006). The ocean is believed to have been still active in the Early Carboniferous, as evidenced by the Late Devonian conodonts and Early Carboniferous radiolarians found in the siliciclastic rocks of the Bayingou mélange (Xiao et al. 1992), and by zircon U-Pb ages of 325 ± 7 Ma and 344 ± 3 Ma found for the cumulate gabbro and plagiogranite (Xu et al. 2006), respectively. The southward subduction of the Junggar Ocean led to the development of the Devono-Carboniferous northern Yili arc . The closure of the ocean led to the formation of the North Tianshan Fault, and is dated by the Sikeshu stitching intrusion which gives a SHRIMP zircon U-Pb age of 316 ± 3 Ma . During the Permian, intense intracontinental dextral shearing occurred along the North Tianshan Fault , and an Early Permian exhumation event is recognized by recent AFT data in the North Chinese Tianshan (Wang et al. 2018). The Mesozoic and Cenozoic in the North Tianshan are characterized by multiple crustal uplift and exhumation events that mainly occurred in the Late Triassic-Cretaceous and Late Cenozoic (Glorie et al. 2010;De Grave et al. 2013;Zhang et al. 2019).

Geology of the Tulasu epithermal gold district
The Tulasu volcanic basin hosts the large Axi and Jingxi-Yelmend gold deposits, the small Tawuerbieke and Qiabukanzhuota gold deposits, as well as the Tabei lead-zinc and the Tieliekesayi dickite prospects that constitute a Late Paleozoic epithermal gold system (Zhao et al. 2014a;Gu et al. 2016). The system is bounded by the regional NW-SE-striking North Yili Basin Fault and WNW-ESEstriking South Keguqin Fault, and its structural framework is characterized by a number of faults with different orientations (e.g. N-S, NW-SE and NE-SW). In the basin, the Ordovician-Silurian marine clastic rock and carbonate rock are locally exposed with interlayered volcanoclastic rock. They are unconformably overlain by an Lower Devonian molasse succession (Fig. 2), which is unconformably overlain in turn by terrestrial arc-related volcanic rocks of the Lower Devonian to Lower Carboniferous Dahalajunshan Formation (Fig. 3), shallow marine carbonates and clastic rocks of the Aqialehe Formation (c. 840-2180 m thick) and volcanoclastic rocks of the Upper Carboniferous Naogaitu Formation (c. 1410-1920 m thick; XBGMR 1993), respectively.
The Dahalajunshan Volcanic Formation (DVF) are the main host rocks of the Tulasu epithermal gold system (Fig. 2). The strata are commonly gently to medium dipping and they are divided into five lithological members (Fig. 3 (Tang et al. 2009) and SHRIMP U-Pb ages of 360.5 ± 3.4 Ma (Zhao et al. 2014b) and 367.1 ± 3.2 Ma (Peng et al. 2017). An Early Carboniferous granitic porphyry in the Tawuerbieke deposit has given a zircon LA-ICP-MS U-Pb age of 349 ± 2 Ma (Tang et al. 2013) and a SHRIMP U-Pb age of 355.4 ± 4.0 Ma (Zhao et al. 2014a). In addition, a monzonite porphyry enclave found in the andesite at Tawuerbieke has yielded a zircon SHRIMP U-Pb age of 356.2 ± 4.3 Ma, and it is believed to be homologous with the andesite (Zhao et al. 2014c). An et al. (2013) reported zircon SHRIMP U-Pb ages of 386.4 ± 9.3 and 416.7 ± 2.6 Ma, respectively, for rhyolite and tuff from the bottom of the DVF at the NW margin of the Tulasu basin (around Jingxi-Yelmend), whereas subvolcanic porphyritic andesite from this area has a U-Pb SHRIMP age of 370.5 ± 2.1 Ma   (Fig. 3).
Systematic descriptions for the main characteristics of gold deposits in the Tulasu epithermal gold district have been summarized in Zhao et al. (2014a), for which a brief introduction is presented below and listed in Table 1. The large Axi deposit is the most representative epithermal gold deposit in the Tianshan Orogen, containing c. 70 t gold at an average grade of c. 5.57 g t -1 (Rui et al. 2002). It was the first major gold discovery at Tulasu in 1989, followed later by the Tawuerbieke and Qiabukanzhuota gold deposits (XBGMR 1993). Exploration at Axi was completed by 1995 with subsequent open-pit mining and then underground mining in 2004. The Axi deposit is hosted by volcanic rocks of the Upper Andesite member of the DVF, and its lenticular orebody is dominated by a series of annular and radial faults, especially a north-south-striking fault (Fig. 2). The deposit shows a clear alteration zonation from the center of the orebody outwards, including silicification, phyllic and propylitization zones (Zhai et al. 2009). Recently, Liu et al. (2019) reported two Re-Os isochron ages of 350 ± 10 and 332 ± 8 Ma for pyrites from the sericitequartzaltered ores and from later grey quartz veins, respectively, which indicate two distinct stages of gold mineralization at Axi. Gold orebodies at Tawuerbieke occur as veins and lenses in the granitic porphyry and andesite of the Upper Andesite member of the DVF. The early hydrothermal alteration stage consisted of silicification, sericitization and carbonatization processes and led to the formation of altered rock ores, on which were superimposed the late calcite-quartz vein ores. Re-Os dating on gold-bearing pyrite from the auriferous veins yielded an isochron age of 323 ± 11 Ma (Zhao et al. 2014b). The Qiabukanzhuota gold prospect is hosted in dacite and tuff of the DVF, and the orebody occurs mainly as lenses along interlayer fracture zones to form the breccia and massive ores. The Tabei lead-zinc prospect is hosted in the Upper Andesite member of the DVF and it is also dominated by the north-south-striking fault (Gu et al. 2016). Mineralization at Tabei can be divided into three types, namely breccia, massive and disseminated ores (Zhao et al. 2014a).

Sample preparation
A total of 10 unaltered samples of the DVF were collected from different stratigraphic elevations in the Tulasu volcanic basin (Table 2; Figs 2 and 3), including andesite (# N-1, S-2, S-5 and S-6), basaltic andesite (# D-1 and Y-1), tuff (# N-2 and QBK-1), dacite (# S-7) and rhyolite (# Y-11). Apatite grains were hand-picked out from these samples for fission track observation, measurement and dating. One sample from a monzonite porphyry enclave (# XT4-8) in andesite, which underwent later sericite alteration (Zhao et al. 2014c), was collected from the Tawuerbieke deposit, and K-feldspar grains were chosen for 40 Ar/ 39 Ar dating (Table 2; Fig. 2). The K-feldspar and apatite grains were separated from the crushed rocks using a combination of magnetic and heavy liquid separation techniques, and then individual grains were hand-picked under a binocular microscope. Table 2 gives a summary of lithology, sampling coordinates and elevation, and dated mineral for the samples.

40
Ar/ 39 Ar dating The 40 Ar/ 39 Ar analysis was conducted at the Institute of Geology, Chinese Academy of Geological Sciences. The sample grains were ultrasonically cleaned with ethanol, and then sealed in a quartz bottle for irradiation in the Swimming Pool Reactor, Chinese Institute of Atomic Energy, Beijing, along with the ZBH-25 biotite standard with an age of 132.7 ± 1.2 Ma at 1σ and a potassium content of 7.6 wt.% (Wang 1983). The duration of irradiation was 1440 minutes with a neutron flux of 2.65 × 10 13 n cm -2 and an integrated neutron flux of 2.29 × 10 18 n cm -2 . After irradiation, the sample was stepwise heated from 700°C to 1400°C with each step lasting for 10 minutes, followed by a 20 minute purification. Finally, the isotopic analysis was performed using a Helix multiple collector mass spectrometer. The measured isotopic ratios were corrected for system blanks, mass discrimination, interference isotopes and atmospheric argon. The interference isotope correction factors were calculated from coirradiation of pure salts of K 2 SO4 and CaF 2 , i.e., ( 40 Ar/ 39 Ar) K = 0.004782, ( 39 Ar/ 37 Ar) Ca = 0.000806 and ( 36 Ar/ 37 Ar) Ca = 0.0002398. All 37 Ar abundances were corrected for radiogenic decay. The decay constant was 5.543 × 10 −10 yr −1 (Steiger and Jäger 1977). The 40 Ar/ 39 Ar plateau age with its ±2σ error was calculated using the ArArCALC program (Koppers 2002). The detailed procedure has been reported by Chen et al. (2002).

Fission track measurement
The apatite fission track measurement was conducted at the Institute of Geology, China Earthquake Administration. The detailed analytical procedure is given by Pang et al. (2013) and is briefly introduced here. Apatite grains without fractures and inclusions were dated using a uranium-poor muscovite sheet as an external detector (Gleadow 1981). The apatite grains were coated with either Teflon or epoxy resin, followed by polishing. Subsequently, the Teflon-coated grains were etched in a KOH-NaOH eutectic melt at 228°C for 19 hours and the epoxy resin-coated grains were etched in a 40% 5 mol L -1 HNO 3 solution at room temperature for 18 minutes. The neutron flux was monitored using CN5 uranium dosimeter glasses. Next, the tracks on the muscovite detectors were revealed by etching in a concentrated HF solution at room temperature. The track densities of the natural apatite grains (ρ s ), the induced fission track populations on the muscovite external detector (ρ i ) and the monitor dosimeter glasses (ρ d ) were measured at 1000× magnification on screen photos from an Olympus microscope, as were the lengths of confined tracks parallel to the Caxis. The fission track ages, which are reported with ±1σ error, were determined based on the zeta (ζ) calibration using a standard weighted mean zeta of 356.6 ± 10 (Hurford and Green 1983). Only horizontally confined tracks were measured for the grains that are confined in the grain and running parallel to the grain's surface. The chi-square (χ 2 ) test was used to quantify the homogeneity of the ages (Galbraith 1981). Samples with a chi-squared probability of >5%, i.e., P(χ 2 ) > 5%, indicate a single-age population. For samples with a chi-squared probability of <5%, the binomial peak fitting method was applied to decompose the observed age distribution into its best-fitting grain age components.

K-feldspar 40 Ar/ 39 Ar age
The K-feldspar 40 Ar/ 39 Ar age dating results from sample XT4-8 are given in Table 3 and the corresponding plateau-age spectrum is shown in Figure 4. Seven high-temperature steps from 820°C to 1140°C yield a well-defined K-feldspar 40 Ar/ 39 Ar plateau age of 313.5 ± 2.2 Ma (2σ; MSWD = 0.82) with c. 54.6% of the released 39 Ar accumulated (Table 3; Fig. 4).

AFT data
The AFT ages are presented in Table 4, and the distributions of the single grain ages for each sample are depicted on radial plots in Figure 5 (Galbraith and Laslett 1993). For all samples, 35 apatite grains were dated, except for sample Y-1 where 19 apatite grains were dated. Sample S-2 failed the χ 2 test, giving a P(χ 2 ) value of 1.2%, which is <5.0% (Table 4), and the central age is chosen as its AFT age. All other samples passed the χ 2 test (Table 4), indicating a single cooling history, and the pooled grain ages are interpreted to be the respective sample AFT ages. Thus, 10 samples yielded Late Jurassic to Early Cretaceous AFT ages, ranging from 156 ± 11 to 109 ± 8 Ma (1σ) ( Table 4). The mean track lengths of the samples ranged from 11.8 ± 2.1 to 13.2 ± 2.4 μm (Table 4; Fig. 6), which are less than the original track lengths of 16.3 ± 0.9 μm (Gleadow et al. 1986) and the mean length of rapid cooling tracks of 14.5-15.5 μm (Green et al. 1989), indicating slow cooling with a protracted residence in the apatite partial annealing zone (APAZ) for the samples.

Inverse thermal modelling
The AFT age and length data were analyzed using the HeFTy inverse modeling program (version 1.9.1; Ketcham 2005) to better understand the thermal history of the Tulasu epithermal gold system. Samples with less than 40 confined fission tracks may cause ambiguous thermal models (Barbarand et al. 2003), and thus samples Y-1 and Y-11 were excluded from the modeling. Sample N-1 was also excluded due to its inadequate D par (average etch pit diameter of fission tracks) data. The apatite annealing equations of Ketcham et al. (2007) were preferentially chosen as the annealing model, and the accordant option was automatically selected for the C-axis projection. The initial mean track length was set to be 16.3 μm by default (Gleadow et al. 1986), and the D par values of fission tracks from each sample were used as kinetic parameters for the modelling. The modelling relies on the Monte Carlo simulation to generate statistically acceptable time-temperature paths (Ketcham 2005). Before running the simulation, it requires several vital timetemperature constraints to better define the thermal history. Since all the samples yielded Late Jurassic to Early Cretaceous AFT ages that are distinctly younger than the Late Devonian to Early Carboniferous eruptive ages, the samples must have experienced at least one episode of total annealing after eruption, which in turn suggests that all samples must have been totally annealed at temperatures above 120°C before the apatite grain with oldest fission track age entered the APAZ. Ages older than the oldest fission track age of each sample and temperatures higher than 120°C were set as the initial time and temperature constraints for modelling. Thus, the AFT ages of the samples were required to be in the APAZ range (c. 60-120°C; Gallagher et al. 1998), which were set as another pair of time and temperature constraints. Further, only paths within the APAZ can be validly interpreted because the fission tracks in apatite anneal instantaneously at temperatures above c. 120°C and the annealing rate becomes extremely slow for temperatures below c. 60°C (Jolivet et al. 2010). The terminal time and temperature constraints of the modelling were set as 0 Ma and current surface temperature, assumed to be c. 20°C, because all the samples were collected from the surface. The results of the modelling are expressed as time-temperature path envelopes, as shown in Figure 7. The purple and pale purple envelopes encompass all the good paths with a goodness-of-fit of >0.50 and all of the acceptable paths with a goodness-of-fit of >0.05, respectively. The weighted mean path (coloured line) is estimated to be the likeliest thermal history for each sample (Fig. 7). In general, the modelled time-temperature paths indicate a similar protracted cooling history (Fig. 7). Sample D-1 crossed the 120°C  Sample n ρ s (10 5 cm -2 ) (N s ) ρ i (10 5 cm -2 ) (N i ) ρ d (10 5 cm -2 ) (N d )   (Fig. 7).

Discussion
Cooling and exhumation of the Tulasu epithermal gold system Our new K-feldspar 40 Ar/ 39 Ar and AFT data are combined with previous geochronological ages to reconstruct the cooling and exhumation history of the Tulasu epithermal gold system (Fig. 8).
The peak of the volcanism (DVF) hosting epithermal gold deposits in the Tulasu basin concentrated to the Late Devonian-Early Carboniferous according to previous zircon U-Pb data (c. 367.1-347.2 Ma; Tang et al. 2009;Zhai et al. 2009;An et al. 2013;Zhao et al. 2014cZhao et al. , 2020Peng et al. 2017). Once the volcanic rocks erupted to the surface, they cooled instantaneously from c. 900°C (the closure temperature of the zircon U-Pb system; Cherniak and Watson 2001) to the paleo-surface temperature of c. 20°C (Fig. 8).
The K-feldspar 40 Ar/ 39 Ar system has a broad range of closure temperatures between 350°C and 150°C because of its poorly quantified diffusion kinetics that result from complex internal microstructures in the K-feldspar crystal (McDougall and Harrison 1999). Hence, our 40 Ar/ 39 Ar age of K-feldspar for the monzonite porphyry enclave in andesite reveals a phase of reheating from c. 357 to c. 315 Ma (Fig. 8) as the samples were gradually reheated to 150-350°C in the Late Carboniferous, which resulted in the total annealing process of fission tracks in the apatite grains. A later magmatic event cannot be responsible for the reheating because there is no evidence for post-Carboniferous magmatism in the Tulasu basin. Two possible mechanisms responsible for the reheating process are basin subsidence with quick burial, or a hydrothermal event.
The DVF is immediately unconformably overlain by the c. 2.4-4.1 km thick Aqialehe and Naogaitu formations after a short period of erosion (Fig. 8). However, this thickness of cover rocks alone cannot reheat the samples to 150-350°C if we adopt a geothermal gradient of c. 25°C km -1 which is the norm for a sedimentary basin. A factor to be taken into account is basin subsidence with exposure to higher thermal gradients as is widely reported from post-orogenic transtensional to extensional strike-slip pull-apart molasse basins. Similarly, Wainwright et al. (2017) reported from the Oyu Tolgoi (Mongolia) porphyry district the occurrence of Upper Devonian sedimentary rocks positioned stratigraphically above c. 370 Ma copper-gold mineralization that preserve a record of uplift, exhumation, and subsequent burial shortly after mineralization. Preservation examples related to rapid subsidence and burial with high heat flow forming anthracitic coal in the Permo-Carboniferous Erzgebirge Basin (Germany) and with protracted exhumation across the Variscan molasse basins are discussed by Wolff et al. (2015), Schneider and Romer (2010), Schneider et al. (2005) and Ventura and Lisker (2003).
Hot hydrothermal fluids flowing from depth are another probable cause of reheating. The sample collected at Tawuerbieke (# XT4-8) had indeed undergone later sericite alteration as described in Zhao et al. (2014c), which supports the occurrence of later hydrothermal event. Further, a previous study of fluid inclusions in the goldbearing quartz veins from the Axi gold deposit indicated that the ore-forming fluids originated from meteoric water, and that fluid boiling occurred during the epithermal mineralization (Zhai et al.  5. Radial plots of single-grain apatite fission track ages for each sample in the Tulasu epithermal gold district. X-axis, relative errors of single grain ages (inversely related to precision, expressed as percentages); Y-axis, standard estimates on single grain ages; arc axis, single grain ages; n, number of apatite grains dated. 2009). The meteoric water circulation drove hot hydrothermal fluids upwards through faults and thus reheated the samples to a temperature of 150-350°C. Since the Carboniferous, the samples underwent a protracted cooling prior to cooling through the APAZ (c. 60-120°C; Gallagher et al. 1998) during the Jurassic and Early Cretaceous before eventually reaching the present surface temperature (c. 20°C; Fig. 8).
Our AFT data and thermal models reveal a Mesozoic exhumation history for the Tulasu epithermal gold system, which indicates that the samples underwent a phase of rapid exhumation at rates of c. 0.029-0.091 mm yr -1 during the Jurassic to Early Cretaceous (Fig. 7). This is also consistent with the geological evidence of thick Jurassic to Early Cretaceous detrital sediments occurring along the northern Yili basin (Glorie et al. 2010;Jolivet et al. 2010Jolivet et al. , 2013. The AFT ages of c. 155-109 Ma fall in the range of previous Mesozoic AFT ages of mostly c. 180-100 Ma in the western Tianshan, and they are thought to record a significant phase of cooling related to the far-field effects of the Cimmerian orogeny along the southern margin of the Eurasian continent and/or the closure of the Mongol-Okhotsk Ocean (De Grave et al. 2007;Glorie et al. 2010;Jolivet et al. 2010;Jepson et al. 2018;Zhang et al. 2019). Available data suggest an eastward-propagating, scissor-like closure of the Mongol-Okhotsk Ocean over a period from the Late Triassic to Early Cretaceous (Yang et al. 2015;Ren et al. 2018a). The closure of the western segment of the ocean is thought to be responsible for the Middle to Late Triassic/Early Jurassic cooling event in the western Tianshan (Glorie and De Grave 2016;Wang et al. 2018), while the closure of its eastern segment is    Jolivet et al. (2010) argued that the Late Jurassic to Early Cretaceous exhumation in the western Tianshan cannot be correlated to the final closure of the Mongol-Okhotsk Ocean because the deformation events during this period do not occur in the Gobi Altai in Mongolia, which is spatially closer to the Mongol-Okhotsk Ocean. The Cimmerian orogeny involves successive evolution of the Tethyan oceanic basins from the Late Triassic to Early Cretaceous, leading to a series of accretionary and collisional events of several Cimmerian blocks, e.g. the Qiangtang and Lhasa blocks (Yin and Harrison 2000;De Grave et al. 2007;Glorie et al. 2010). In particular, the Late Jurassic to Early Cretaceous accretion of the Lhasa block into the southern Eurasian margin is more likely to become a far-field driving force for the Mesozoic exhumation at Tulasu due to their consistency in time. The Late Triassic to Early Jurassic accretion of the Cimmerian Qiangtang block is often interpreted as the lure of the Middle to Late Triassic/Early Jurassic cooling event in the western Tianshan (Jolivet et al. 2010;Glorie and De Grave 2016;Jepson et al. 2018). Therefore, we suggest that the rapid exhumation in the Tulasu basin during the Jurassic to Early Cretaceous is most likely formed from far-field effects of the Cimmerian orogeny.
It is noteworthy that the sample QBK-1 yielded the youngest AFT age (109 ± 8 Ma) which corresponds to the highest elevation (2011 m; Table 2; Fig. 9). Theoretically, when there exists a regional horizontal isothermal surface, the AFT ages of the samples should be positively correlated with their elevations, which means that the sample with older AFT age should pass through the APAZ earlier to reach a higher elevation (Gleadow 1981). However, this is not the case at the Tulasu basin. After discarding this sample (# QBK-1), the other samples still display a slight negative correlation between AFT age and elevation (red dashed line in Fig. 9). This negative linear relationship indicates that the isothermal surface was not horizontal at Tulasu, and differential distributions of thermal states may have existed to affect the AFT ages in the area (Gleadow 1981).
No Cenozoic AFT ages were obtained, indicating that no significant exhumation occurred in the basin, and our modelling showed a steady and protracted phase of slow exhumation at rates of c. 0.013-0.018 mm yr -1 after this rapid Mid-Mesozoic exhumation (Fig. 8). The western Tianshan indeed remained tectonically stable in the Late Cretaceous, which induced extensive planation (De Grave et al. 2007;Jolivet et al. 2013;Chen et al. 2017). However, significant Cenozoic rapid cooling occurred in the western Tianshan as a remote response to India-Eurasia collision, and it has been commonly recorded by numerous thermochronological studies (Glorie et al. 2010;Glorie and De Grave 2016;Jepson et al. 2018;Zhang et al. 2019). Specifically, in the Keguqin mountain region, which was separated from the Tulasu basin by the South Keguqin Fault, AFT dating of granites has indicated Oligocene to Miocene exhumation (Wang et al. 2018). The South Keguqin Fault was formed in the Mesoproterozoic and it was reactivated by Early Mesozoic normal faulting and by Late Cenozoic thrusting, which were responsible for Late Triassic to Late Jurassic exhumation and rapid Oligocene to Miocene exhumation, respectively (XBGMR 1993;Wang et al. 2018). Thus, multiple reactivations of this basement fault may be responsible for differential exhumation between the Keguqin mountain and Tulasu basin.

Implications for the preservation of pre-Mesozoic epithermal deposits
The epithermal gold mineralization in the Tulasu basin is thought to be Early Carboniferous from direct Re-Os dating on gold-bearing pyrite (Zhao et al. 2014b;Liu et al. 2019). Previous microthermometric data for fluid inclusions in the gold-bearing quartz veins revealed low temperatures (c. 120-240°C) and shallow mineralization depths (c. 700 m) for the gold mineralization, which are also supported by field geological evidence (Zhai et al. 2009;An and Zhu 2018;Liu et al. 2018). Such conditions seem to be very unfavorable for preservation of the mineralization, especially since erosion occurred immediately after gold mineralization as shown by the unconformity between the Dahalajunshan and Aqialehe Formations, although typomorphic characteristics of pyrite at Axi indicate minimal erosion (Wei et al. 2011). Our thermochronological study provides important information on the post- Fig. 9. Apatite fission track (AFT) age v. elevation plot for the volcanic rock samples in the Tulasu basin. The red dashed line represents a negative linear relation between AFT ages and elevations after discarding the sample QBK-1 with much younger AFT age. mineralization history at Tulasu to help our understanding of its preservation mechanism. Since the genesis of the epithermal mineralization, the Tulasu basin has undergone a phase of burial reheating until c. 315 Ma despite the minor erosion by the unconformity event. The mineralization was covered by the Lower Carboniferous Aqialehe Formation and Upper Carboniferous sediments with total thickness of at least c. 3.0 km (XBGMR 1993). Thick cover rocks drove the samples deep to generate the total annealing of AFTs and, more importantly, they protected the gold deposits from being eroded away. Although a phase of rapid exhumation at rates of c. 0.029-0.091 mm yr -1 occurred in the basin during the Jurassic to Early Cretaceous (Fig. 8), these rates are relatively slow over a geological time scale, and the resulting erosion of an estimated thickness of c. 2.1 km was not enough to remove the deposits. Therefore, we suggest that rapid post-mineralization burial by a thick sequence of cover rocks and their subsequent relatively slow exhumation favored the preservation of the Late Paleozoic Tulasu epithermal gold system.
The world's epithermal gold systems are concentrated in the circum-Pacific, Tethyan-Himalayan and Paleo-Asian metallogenic domains. Most of these systems are known to occur in Mesozoic or Cenozoic terranes (Cooke and Simmons 2000;Kerrich et al. 2000;Simmons et al. 2005), with a few pre-Mesozoic systems that are mainly found in the western Tianshan (Seltmann et al. 2014), Tasman (Corbett and Leach 1998) and Uralian (Lehmann et al. 1999) Orogens. In the Tasman Orogen, the Late Devonian Yerranderie silver-gold-lead district is hosted in the Early Devonian Bindook Volcanic Complex and it was covered by the Upper Devonian and Permian sediments after its deposition (Downes 2007). The Carboniferous Pajingo gold deposit was buried by overlying Permian-Triassic sediments following its epithermal mineralization in Late Devonian-Carboniferous volcanic rocks (Bobis et al. 1995). In the southern Ural Orogen, the Bereznjakovskoje gold trend is hosted in Late Devonian volcanicsubvolcanic rocks, which are overlain by an Upper Devonian to Lower Carboniferous volcano-sedimentary sequence (Lehmann et al. 1999). Thus, it seems that a thick sequence of cover rocks was a vital factor in the preservation of these pre-Mesozoic epithermal deposits by protecting the mineralization from erosion. Our study on the Tulasu epithermal gold system suggests further that a protracted slow exhumation with coeval tectonic stability also played important roles in preventing such epithermal deposits from being destroyed.

Conclusion
New 40 Ar/ 39 Ar and fission track data, when combined with previous geochronological data and inverse thermal modelling, constrain the cooling and exhumation history of the Tulasu epithermal gold system. The AFT ages of 10 volcanic rock samples range from 156 ± 11 to 109 ± 8 Ma, and 40 Ar/ 39 Ar dating of K-feldspar in a monzonite porphyry enclave in andesite yields a plateau age of 313.5 ± 2.2 Ma. Thermal modelling shows a phase of reheating of the Early Carboniferous epithermal gold mineralization prior to c. 315 Ma in the Tulasu basin that was likely caused by burial by younger Carboniferous sediments. The Tulasu basin then underwent a phase of rapid cooling at rates of c. 0.9-2.7°C Ma -1 during the Jurassic to Early Cretaceous, followed by slow cooling at c. 0.4-0.5°C Ma -1 . Rapid exhumation at rates of c. 0.029-0.091 mm yr -1 occurred in the Jurassic to Early Cretaceous, leading to an estimated c. 2.1 km of erosion that was not enough to destroy the known epithermal deposits. We propose that syn-to post-mineralization basin subsidence and the rapid burial by thick cover rocks and the subsequent protracted slow exhumation appear to be pivotal factors in the preservation of the Tulasu epithermal gold system, similar to most of the pre-Mesozoic epithermal deposits worldwide.