Differential growth of the northern Tibetan margin: evidence for oblique stepwise rise of the Tibetan Plateau

Models of how high elevations formed across Tibet predict: (a) the continuous thickening of a “viscous sheet”; (b) time-dependent, oblique stepwise growth; and (c) synchronous deformation across Tibet that accompanied collision. Our new observations may shed light on this issue. Here, we use 40Ar/39Ar and (U-Th)/He thermochronology from massifs in the hanging walls of thrust structures along the Kunlun Belt, the first-order orogenic range at the northern Tibetan margin, to elucidate the exhumation history. The results show that these massifs, and hence the plateau margin, were subject to slow, steady exhumation during the Early Cenozoic, followed by a pulse of accelerated exhumation during 40–35 Ma. The exhumation rate increases westward (from ~0.22 to 0.34 and 0.5 mm/yr). The two-fold increase in exhumation in the western part (0.5 mm/yr) compared to the eastern part suggests westward increases in exhumation and compressional stress along the Kunlun Belt. We relate these observations to the mechanisms responsible for the oblique stepwise rise of Tibet. After collision, oblique subduction beneath Kunlun caused stronger compressional deformation in the western part than in the eastern part, resulting in differential growth and lateral extrusion.

Scientific RepoRts | 7:41164 | DOI: 10.1038/srep41164 been reported from the eastern part of the Kunlun Belt 5,21,[24][25][26] , where a cooling history between 350-60 °C was extracted. A rapid exhumation event and therefore the onset of deformation during the Cenozoic were then concluded to have occurred at approximately 30 Ma 24,25 or 40 Ma 5,10 in the eastern part of the Kunlun Belt. In contrast, low temperature geochronological studies are lacking, and existing constraints on the exhumation and cooling history of the western part of the Kunlun Belt are very poor. Therefore, although it is the first-order orogenic belt of the northern Tibetan margin and has great potential to reveal the evolution of the plateau, systematic research and direct evidence of the couplings between tectonics and surface processes are still lacking, and the modes of uplift and exhumation needs to be better quantitatively constrained.
To evaluate the timing and history of mountain-building in the Kunlun Belt, we date and model samples from three new vertical age-elevation transects from different parts of the Kunlun Belt (transect 1 comes from the middle part; transects 2 and 3 are found in the western part) (Fig. 2, Table S1). To facilitate comparisons with the eastern part of the Kunlun Belt, the original data from two of our previous transects in the eastern part (transects 4 and 5, Fig. 1) 10 are also processed (Table S1) and modeled in this study. All of these transects are distributed evenly over a distance of ~600 km along the Kunlun Belt and are located in the immediate hanging wall of the North Kunlun Fault (Fig. 1). Thus, the exhumation history of the range provides an important constraint on the manner of mountain-building along the northern Tibetan margin and can potentially inform models of the growth of the plateau.

Results
We present results using the K-feldspar 40 Ar/ 39 Ar and apatite (U-Th)/He thermochronological systems in Figs 3 and 4 and Supplementary Tables S3 and S4, respectively. A summary of the ages, including their sample locations, is provided in Table S1. Furthermore, the kinetics of argon and helium diffusion in these systems span a temperature window ranging from ~350 °C down to ~60 °C 27,28 and allow us to investigate the thermal history of this part of the plateau margin in unprecedented detail. 40 Ar/ 39 Ar analysis results. The results of K-feldspar 40 Ar/ 39 Ar analysis of two samples from transects 1 (15kl03) and 2 (1335-2), respectively, are shown in Fig. 3 and Supplementary Table S3. Two K-feldspar samples exhibit flat age spectra (Fig. 3), defining good plateaus that account for > 80% of the total 39 Ar released at high temperature steps, implying that rapid cooling occurred at these times (see Supplementary Information for details). The plateau ages are 232.5 ± 1.4 Ma (15kl03) and 234.6 ± 1.2 Ma (1335-2), which are similar to those reported by previous studies in the Kunlun Belt 11,14,15 . These massifs were rapidly exhumed during the early Late Triassic and are attributed to the orogeny of the Kunlun Belt, which was probably driven by the collision between the Songpan-Ganzi and Kunlun blocks 11,29 .

K-feldspar
Conversely, the two age spectra display staircase shapes at the low temperature steps of the step-heating (Fig. 3). Although less than 20% of the total released 39 Ar resides within these low-temperature domains (Fig. 3), Scientific RepoRts | 7:41164 | DOI: 10.1038/srep41164 the behavior 30-33 of multi-diffusion domains within K-feldspar suggests that these staircase spectra are the result of slow cooling and, therefore, record cooling ages between ~350-150 °C. Their minimum ages of 198.8 ± 2.4 and 90.3 ± 2.1 Ma (Fig. 3, Table S1) reflect the final time when the samples passed the ~150 °C isotherm.
The 40 Ar/ 39 Ar thermochronologic data from K-feldspars imply that rocks now at the surface resided at or above temperature of ~350 °C during the early Late Triassic and ~150 °C during the late Cretaceous. These results place a maximum bound on total exhumation of ~11 km since the middle Triassic and ~5 km since the late Cretaceous, assuming a gradient of 30 °C/km 5 .
Apatite (U-Th)/He dating results. Ages from the apatite (U-Th)/He thermochronological system provide insight into the Cenozoic cooling history. Single apatite grains are analyzed by using the (U-Th)/He dating technique. The results are presented in Fig. 4 Fig. 4a), exhibiting a distinct age-elevation trend. The pattern of the trend shows an increase in the gradient at approximately 35 Ma, implying that accelerated exhumation began at this time. This observation is similar to those from our previous transects in the eastern part of the Kunlun Belt near Nuomuhong 10 , where a rapid exhumation starting around 35 Ma was also recognized. In contrast, (U-Th)/He ages from transect 2 present a steep age-elevation relationship: the ages ranging from 39.8 ± 3.0 to 36.6 ± 3.2 Ma are dispersed over an elevation range  (Fig. 4c). The relatively small elevation ranges (~500-600 m) from transect 2 and 3, which are    partially responsible for the age clustering, are insufficient to permit examination of age-elevation correlations for these massifs. Even so, given that the seven samples do not show any trend reflective of a slow exhumation in the age-elevation correlation suggest that transects 2 and 3 reflect parts of accelerated exhumation trend, although we do not see the break where it starts. This can be demonstrated using our new (U-Th)/He dates from detrital apatites (39.5-43.2 Ma, authors' paper under preparation) from Miocene growth strata in the Kumkuli intramontane basin, which suggest that the apatites eroded from the surrounding mountains (the western part of the Kunlun Belt and Qimantagh) have the same ages as transects 1 and 2. In addition, transect 2 exhibits very slightly negative correlation between age and depth (Fig. 5, Table S1), implying that this massif could have been tectonically tilted slightly. However, such an event is difficult to interpret on the basis of just this marginal correlation. Therefore, we conclude that the patterns of transects 2 and 3 most likely represent a monotonic cooling or an accelerated exhumation between ~38 and ~43 Ma. These results clearly reveal a picture of differential evolution of exhumation history along the Kunlun Belt. To compare directly, samples from each transect are normalized to the local elevation of a remnant erosion surface resolved from a 30-m resolution Digital Elevation Model (DEM) (Figs S1 and S2, see Supplementary Information), and plotted as age versus structural depth (Fig. S3, Fig. 5). To permit comparisons between different parts of the Kunlun Belt, two of our previous transects near Nuomuhong 10 , which are labeled as transects 4 and 5 in Fig. 1, are normalized to the local erosion surface (Table S1) and plotted and modeled in Fig. 5 as well. The five transects form three distinct age-depth trends in the age-depth plot (Fig. 5). Given various scenarios, forward thermal modeling was carried out on each trend, and the solutions are shown in Fig. 5 and Fig. S6. The data and modeling results demonstrate a tendency for uplift and exhumation to strengthen from east to west along the Kunlun Belt during Eocene time. The gentle slopes of transects 1, 4 and 5 imply that a very slow, constant exhumation (0.01-0.02 mm/yr) history was predominant in the Kunlun Belt during the late Cretaceous to early Paleogene (Fig. 5). Abrupt changes in the slopes of transects 1 (from 0.02 to 0.34 mm/yr), 4 and 5 (from 0.01 to 0.22 mm/yr) appear at 35 Ma, indicating that the onset of accelerated exhumation occurred 35 Ma ago in the eastern and middle parts of the Kunlun Belt. Similarly, modeled results that form a steep slope at ~40 Ma in transects 2 and 3 demonstrate that rapid exhumation occurred at this time at a rate of 0.5 mm/yr (Fig. 5). Although the difference (~5 Myr) in starting times of accelerated exhumation between the eastern and western parts of the Kunlun Belt is small and can be obscured by the uncertainties of the dates, the overall trend illustrated by all samples from transects 2 and 3 (where no samples yielded ages < 35 Ma) suggests that this difference may be meaningful and reflects the "true" situation. More evidences are discussed later that show that this inference is reasonable. The data from transects 2 and 3 do not record the abrupt change in slope, most likely because of the rapid exhumation which removed overlying rocks that would have recorded rapid slope changes in those transects. Nevertheless, the uppermost sample of this steep slope indicates the lower limit of rapid slope change in depth, which is 900 m and 1400 m higher than those of transect 1 and transects 4 and 5, respectively (Fig. 5). According to He kinetic and diffusion theory, the rapid slope change in the helium age-depth curve defines the start of the rapid motion of rocks towards the Earth's surface when pass the closure isotherm (~60 °C) 28 . Therefore, our observations suggest that rapid uplift and exhumation occurred earlier in the western part than in the eastern part of the Kunlun Belt during the early Cenozoic.

Discussions
Our new data suggest that the Kunlun Belt grew differentially from west to east during the Cenozoic. Specifically, uplift initiated at approximately 40 Ma in the western part and at approximately 35 Ma in the eastern part. Exhumation rates of 0.22 mm/yr in the eastern part, 0.34 mm/yr in the middle part and 0.5 mm/yr in the western part suggest a westward increase in uplift, under the logical assumption that the different parts of the Kunlun Belt experienced the same climatic conditions. Given a uniform geologic setting, the westward strengthening of uplift and exhumation suggest that lateral compression increased in strength from east to west along the Kunlun Belt. The modeled exhumation rates (Fig. 5) imply that the western part experienced compressive forces twice as strong as those experienced by the eastern part.
A growing body of evidence indicates that the regional deformation was stronger and initiated earlier in the western part than in the eastern part along the Kunlun Belt during early Cenozoic. Rapid exhumation of the eastern part of the Kunlun Belt at approximately 30 Ma is inferred from accelerated cooling determined from feldspar multi-domain diffusion modeling 24,25 . Previous (U-Th)/He data revealed a rapid exhumation event beginning at approximately 35 Ma in the eastern part of the Kunlun Belt 5 . Apatite and zircon (U-Th)/He and apatite fission-track ages along the strike of the eastern part of the Kunlun Fault show that exhumation rates increase eastward, from 30-25 Ma near Golmud to 20-15 Ma near Xiangride 34 (Fig. 1).
The initiation of sedimentation in the Kumkuli basin since the middle Eocene suggests that rapid exhumation started ~40 Ma in the western part of the Kunlun Belt 35 35 , suggesting that the provenance of sedimentation moved from west to east in the surrounding mountains. Regional seismic reflection profiles across the Qaidam Basin reveal that a progressive shift in crustal thickening mechanisms: crustal shortening decreases from > 48% in the west to < 1% in the east during the early Cenozoic 42 . In the Hoh Xil basin, immediately south of the study area, there was rapid sedimentation at approximately 40-35 Ma 4 . A recent review that investigated various lines of evidence, including deformation, faulting, the results of basin research, paleomagnetic declination anomalies and thermochronology, concluded that growth and deformation of the northern Tibetan margin began at 50-40 Ma 11,12 .
Two end-member models of how the Tibetan Plateau formed are: (a) diffuse crustal thickening of a viscous sheet, with shear on vertical planes playing a subsidiary role (i.e., the viscous sheet model) 3 , and (b) localized shear between rigid lithospheric blocks (i.e., oblique stepwise growth model) 1 . Both models predict time-dependent, northward progression of strain and deformation that is initially concentrated at the plate boundary and then propagates in time away from that boundary as topographic stresses begin to favor deformation at greater distances from the plate boundary.
Recently, a new mechanism for Tibetan Plateau growth has been proposed, namely the central Tibet growth model, although when and how are still in dispute. The main feature of this model is that the region of high topography initiated in central Tibet and the lithospheric stresses propagated north and south from this high topography outward 4,6,7,8,11 . Although the onset time of growth remains controversial, most studies suggest a scenario in which the high topography in central Tibet and the northern Tibetan margin was built up by [40][41][42][43][44][45][46][47][48][49][50]6,7,8,11,12 . The timing of collision between India and Eurasia has been disputed over the past several decades but is crucial for understanding the far-field effects of the collision. The commonly suggested age of collision of 47-55 Ma 2,13,15 implies an immediate response to the collision in central and northern Tibet. However, several recent studies suggest that the collision between India and Eurasia may have occurred during the early Paleocene (~65-60 Ma) 16-18 , 30-20 Myr earlier than the initiation of regional scale deformation across the central and northern Tibet. These latest observations imply that the propagation of the far-field effects of the collision to the northern margin was much slower than previously thought 2,5 . (U-Th)/He ages from the North Qilian Shan suggest that rapid exhumation extended slowly northward, reaching as far as the Hexi Corridor by the late Miocene (~10 Ma 43 ), and apatite fission track ages from the Liupan Shan indicate that rapid exhumation did not begin until ~8 Ma 44  The viscous sheet model predicts progressive northward migration of the associated crustal thickening and therefore synchronous uplift and uniform force at the frontline in the propagation direction after the rigid "indenter" represented by India converged with the weaker Eurasian lithosphere. However, this is not consistent with our observations of diachronous deformation with different intensities across the northern Tibetan margin. In addition, the strain generated by edge tractions in the thin viscous sheet model is insufficient to generate faulting in northern Tibet 5 .
Our thermochronologic data and modeling results from the Kunlun Belt support the oblique step-wise growth mechanism. Although how and when the stress propagated across the Tibet is called into question by new deformation evidence from central Tibet 4,[6][7][8]12 , the notion of the oblique stepwise growth model 2 , i.e., the oblique rise of Tibet and lateral extrusion of the thickened crust, is consistent with our thermochronologic results along the Kunlun Belt. This model describes a picture of oblique subduction after collision. Oblique subduction of the Asian lithospheric mantle plays a key role that involves both lateral extrusion and crustal thickening 2 . It can account for the markedly asymmetric growth of relief toward the east and the eastward extrusion observed in northeastern and eastern Tibet 45 . Far-field impacts on the northern Tibetan margin initially developed approximately Scientific RepoRts | 7:41164 | DOI: 10.1038/srep41164 40-35 Ma. Oblique subduction of the Qaidam block beneath the Kunlun Belt to depth of approximately 200 to 300 km 2 resulted in sinistral slip along the Kunlun and North Kunlun Faults and crustal thickening. A stronger stress was created in the western than in the eastern part of the Kunlun Belt due to the northeast-southwest oblique compression, resulting in an eastward decrease in uplift and exhumation along the Kunlun Belt (Fig. 5). The movement along lithospheric faults resulted in eastward extrusion of the Tibetan Plateau, followed by a major compression event in the late Eocene. Consequently, the present-day Kunlun Belt contains a highland with low relief in the western part and a lowland with high relief in the eastern part.
Our results support the temporal development predicted by the central Tibetan growth model. This model suggests a scenario of plateau growth in central Tibet and propagation to the northern Tibetan margin by [40][41][42][43][44][45][46][47][48][49][50] Ma that is consistent with our observations along the Kunlun Belt 4, [6][7][8]12 . The growth of central Tibet and northward stress propagation are associated with folding and thrusting that resulted from subduction of the Indian lithosphere underneath the Tibetan plateau 4,6-8 . Northward thrusting from central Tibet is obliquely oriented 2,4,6,8 , causing an eastward decrease in compressional stress along the Kunlun Belt. 40 Ar/ 39 Ar geochronology. Characterized by complicated microstructures that serves as domains with different sizes, argon retention properties and closure temperatures 30,31,46 , K-feldspars are potentially able to record 40 Ar/ 39 Ar ages of closure temperatures ranging from 350 down to 150 °C as they cool 27,47 , and thermal histories can consequently be extracted [30][31][32][33]48 . A high resolution (36-40 steps) step-heating technique was applied from 450 to 1300 °C in a 40 Ar/ 39 Ar analysis to reveal the argon distributions within the K-feldspar grains in as much detail as possible (Table S3). K-feldspars were carefully inspected to meet the criteria and requirements of thermochronology using the 40 Ar/ 39 Ar method 27,48 . The sample processing and laboratory procedures involved in 40 Ar/ 39 Ar analysis are described by Wang et al. 31 . Ages were calculated against the international standard YBCs sanidine (29.286 ± 0.045 Ma 49 ). The experiments were conducted on an MM5400 mass spectrometer; plateau ages were calculated using adjacent ages of step-heating that agreed within a range of 2σ . A complete discussion of sample preparation and analytical procedures can be found in the Supplementary Information.

Methods
(U-Th)/He geochronology. See the Supplementary Information for sample preparation. In principle, clear, euhedral apatite grains longer than 130 μ m and wider than 75 μ m were used for (U-Th)/He dating. Each apatite grain was inspected carefully under a high-power microscope to eliminate those containing impurities and inclusions. Needle-like, fragmentary, subhedral, rounded or zoned apatite grains were excluded as well. After this selection process, each selected apatite grain was enclosed in a 1 mm × 1 mm platinum capsule and moved one after another into a well in a stainless-steel disk for He measurement using an Alphachron MK II noble gas mass spectrometer. Each grain was heated twice at 900 °C by a diode laser, each time for 10 min, to extract the He completely as possible. The abundance of 4 He was determined using the isotope dilution technique. The 3 He spike used is calibrated daily against an independent 4 He standard tank. The uncertainty of the 4 He measurements averaged less than 2%. After completion of the He measurements, the degassed apatite grains were removed and prepared for U and Th measurements. After being removed from the capsule and placed in a beaker, each apatite was dissolved in 25 μ L of a liquid reagent made up of 50% HNO 3 spiked with 235 U and 230 Th. The beakers were subjected to ultrasound vibrations until the apatite crystals had dissolved completely. Finally, the liquid reagent was diluted to 5% HNO 3 and analyzed for U and Th abundances by using a Thermo Fisher X-Series II ICP-MS. As determined by averaging replicate analyses of spiked standard solutions, the analytical precisions of the measured 235 U/ 238 U and 230 Th/ 232 Th ratios were 0.8% and 0.5%, respectively. A series of Durango standard grains was analyzed between measurements of sample grains for monitoring the whole experimental protocol. Application of (U-Th)/He methods at IGGCAS yielded an averaged Durango apatite age of 32.24 ± 1.01 Ma 50 with an internal precision (1σ ) of 1.5%. This age is quite consistent with the recommended age 10,50 . A detailed description of analytical procedures can be found in the Supplementary Information.

Forward modeling.
A one-dimensional thermal model was utilized for forward modeling of the thermal response to exhumation. Various exhumation histories were imposed, subject to the constraint imposed by a thermal gradient of 30 °C/km (see the Supplementary Information), and subject to a constant surface temperature of 10 °C. Thermal history solutions for each transect were obtained using a constant temperature offset (30 °C/ km) model in QTQt v.5.3.0 51 , and the median thermal history was used for the transect. The detailed modeling procedure can be found in Gallagher et al. 51 . Samples along a transect are used to calculate model helium ages using the helium diffusion kinetics from Flowers et al. 52 (radiation damage model RDAMM). We seek a series of preferred model fits that faithfully represent most of the sample ages, i.e., the basic trend of the data. A detailed description of the modeling procedure can be found in the Supplementary Information.