Re-evaluating the geochronology of the Permian Tarim magmatic province: implications for temporal evolution of magmatism

The Permian Tarim magmatic province has 118 published ages, ranging from 358 to 205 Ma, but the timing of mafic magmatism is not well constrained. We report two new secondary ion mass spectrometry U–Pb zircon dates on the Halahatang trachydacite and Wajilitag olivine clinopyroxenite, which are 287.2 ± 2.0 Ma and 283.2 ± 2.0 Ma, respectively. The trachydacite overlies the uppermost basalt and constrains the latest eruption age of basalt in northern Tarim. The latter is the first high-resolution date for the Wajilitag mafic layered intrusion. By screening all published ages, we identified 22 robust ages, ranging from 290.9 ± 4.1 to 261.7 ± 1.8 Ma. The robust ages together with our new data reveal a protracted period of mafic magmatism at c. 283 and c. 267 Ma. Silicic magmatism occurred from 291 to 272 Ma. Although the current known volume of Tarim basalt is too small to qualify as a large igneous province, the eroded and intrusive components, as well as pyroclastic deposits and silicic lavas, may increase the estimated volume. Further work is required to refine the duration of magmatism and the volume estimate of the province. Supplementary material: A list of all the published dates, raw data of published 40Ar–39Ar dates, raw data of published U–Pb dates, published geochemistry data used for calculation and complete details of the data evaluation are available at http://www.geolsoc.org.uk/SUP18862.

Extensive Permian magmatism is found in the Tarim Basin in NW China (Fig. 1), and many researchers have interpreted it to represent a large igneous province (LIP) based on the area of volcanic rocks exposed in outcrop and found in drill cores (e.g. Yang et al. 2007;Tian et al. 2010;Xu et al. 2014). The magmatism can be divided into two main components ( Fig. 1): (1) extrusive rocks, consisting of basaltic, andesitic, dacitic and rhyolitic lavas, and mafic and silicic pyroclastic deposits; (2) intrusive rocks, including mafic-ultramafic layered intrusions, syenite intrusion, mafic and silicic dykes, ultramafic breccia pipes and granite intrusions (e.g. C.L. Tian et al. 2010;Wei & Xu 2011, 2013Yu et al. 2011b;and references therein).
The details of the temporal evolution of Tarim magmatism are poorly constrained, despite the large number of geochronological studies available. There are 118 published geochronology dates on Tarim magmatism, on a wide range of lithological units by many techniques (whole-rock K-Ar, whole-rock 40 Ar-39 Ar, mineral 40 Ar-39 Ar, mineral U-Pb dates by laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS), secondary ion mass spectrometry (SIMS), sensitive high-resolution ion microprobe (SHRIMP) and chemical abrasion thermal ionization mass spectrometry (CA-TIMS), whole-rock Rb-Sr and whole-rock Sm-Nd), which range from 358 to 205 Ma, with a span of 153 myr ( Fig. 2a  and b). Flood basalts, the most significant extrusive component, have an age range from 297.4 ± 5.6 Ma (basalt zircon U-Pb date, Zhang et al. 2012) to 252.3 ± 3.5 Ma (whole-rock 40 Ar-39 Ar date, Liu et al. 2012), a 45 myr time span. Whether this large time period represents a continuum or distinct multiple pulses of magmatism, or simply issues with the geochronology results, is not clear. This raises questions about whether the Tarim meets the classification of a large igneous province (LIP), defined by Ernst (2014) as emplacing substantial amounts of mafic magma (areal extent >0.1 Mkm 2 and igneous volume >0.1 Mkm 3 ) within a short duration pulse or multiple pulses (less than 1-5 myr), with a maximum lifespan of up to 50 myr. Published geochronological data are often taken at face value without careful consideration of their reliability or whether the geological context of the sample is known. There have been some previous efforts to re-evaluate published geochronology results (e.g. Li et al. 2011;Wei et al. 2014), but on rather limited datasets. Some of the potential issues associated with data quality and validity include the following: (1) the effects of alteration on K-Ar and 40 Ar-39 Ar dates; (2) dating xenocrysts or antecrysts that may not represent the crystallization age of their host rocks (e.g. perovskite in kimberlite, zircon in basalt). Thus, it is appropriate to comprehensively re-evaluate the previous geochronology work, and also to carry out new geochronology studies on samples selected to answer specific issues with the Tarim magmatism chronology.
This paper addresses both of these issues by providing two new SIMS U-Pb zircon ages on a trachydacite lava from Halahatang and an olivine clinopyroxenite from the Wajilitag mafic-ultramafic layered intrusion, and a comprehensive re-evaluation of the published ages. The trachydacite sample was chosen as it overlies the uppermost basalt lavas in the Halahatang region (Fig. 3a), and thus constrains the age of the youngest mafic volcanism in this area. The age of the mafic intrusive component of Tarim magmatism and its temporal relationship to the extrusive magmatism is not well constrained either (Fig. 2b). Therefore, the olivine clinopyroxenite sample was analysed to provide the first high-resolution date for the Wajilitag mafic-ultramafic layered intrusion. By carefully examining sample locations, experimental procedures and data interpretation, combined with both the established and newly developed evaluation approaches for geochronology data, we have identified a subset of 22 previously published ages that we consider to be robust (Fig. 2c). These high-quality ages, combined with the two new ages, are then used to develop a revised model for the temporal evolution of magmatism in the Tarim during the Permian.

Geological background
Bounded by the Tianshan Mountains to the north and NW, the Kunlun Mountains to the SW and Altun Mountains to the SE (Fig. 2a), the Tarim Basin is located in northwestern China and occupies an area of c. 600 000 km 2 ). The floor of  the basin is a complex Precambrian basement, which is believed to be a fragment of the Rodinian supercontinent (Li et al. 2003;Lu et al. 2008) and is overlain by thick sequences of Ordovician, Permian and Cretaceous strata (Jia 1997;Zhang 2003). A large volume of Permian extrusive rocks, including flood basalt and rhyolite lavas, as well as mafic-ultramafic intrusive complexes and mafic dykes are widely reported in the Tarim Basin (e.g. Jia 1997;Yang et al. 2006Yang et al. , 2007Fig. 1). Subsurface seismic and drilling data indicate that the Tarim basalt lavas have an estimated areal extent of 265 000 km 2 and a volume of 40 000 km 3 (Pan 2011;Pan et al. 2013;Fig. 2). The distribution of outcrop and subsurface data suggests that the bulk of the province (>99%) is covered by the Taklamakan Desert sands and Quaternary sediments (Fig. 1a). The main outcrop of flood basalt is located in Keping, which exposes c. 100 km of lateral section (Fig. 1c), and a cumulative thickness of c. 400 m. The Keping outcrop consists of two packages of basaltic lavas and volcaniclastic rocks with a c. 800 m sedimentary package between them. The lower basalts and intercalated clastic deposits are called the Kupukuziman Formation, and the upper basalts and intercalated clastic deposits are called the Kaipaizileike Formation. The thick intermediate unit has not been formally described, but is informally referred to as the FP Formation (Shangguan et al. 2012). Other smaller basalt outcrops have been reported from the Damusi and Qimugan (Yang et al. 2006;Li et al. 2013). Thick sequences of subsurface basalt lavas (>1200 m) were also reported from drilling data in northern Tarim (Tian et al. 2010). Felsic lavas have limited exposures in the Xiaotiekanlike-Wenquan outcrop , and multiple felsic lava layers are also reported coexisting with the basalts from drilling data in northern Tarim (Tian et al. 2010). Intrusive complexes, including the Wajilitag and the Piqiang mafic-ultramafic intrusions, the Xiaohaizi syenite body and the Piqiang granites are all located around the NW margin of the basin (e.g. Wei & Xu 2013;Zhang & Zou 2013). Owing to the lack of seismic data, the actual dimensions of the intrusive complexes are still unclear. Mafic dykes are widespread in the Bachu area and Keping, and some intrude the syenite body and the Keping basalt. The thickness of Bachu dykes ranges from 0.6 to 4 m (Z.L. , and the Keping dykes have a mean thickness of 3.8 m (Chen et al. 2014 relationships have been observed between the breccia pipes (dykes) and the Wajilitag intrusion.

Analytical procedures
Zircon crystals were extracted from HA2-5379 and W08 by density and magnetic separation, and selected grains were handpicked for analysis. Zircons were mounted in epoxy together with the standard Plesovice zircon (Sláma et al. 2008) and polished. Cathodoluminescence (CL) and transmitted and reflected light images of all zircons were obtained to guide SIMS U-Pb measurement. CL images were obtained using a FEI PHILIPS XL30 scanning electron microscope in Peking University, with analytical conditions of 12 kV and 12 μA.
U-Pb ages were measured by SIMS at the State Key Laboratory of Lithospheric Evolution, Institute of Geology and Geophysics, Chinese Academy of Sciences. Zircon analytical procedures by SIMS follow those of Li et al. (2009). The measured Pb isotopic compositions were corrected for common Pb using non-radiogenic 204 Pb. Corrections were small enough to be insensitive to the choice for the common Pb composition. An average of present-day crustal compositions (Stacey & Kramers 1975;Wiedenbeck et al. 1995) was used for the common Pb. Data processing was done using the Isoplot 3.6 software (Ludwig 2008). Errors of single analyses are quoted for 1σ uncertainty. Ages are quoted as weighted mean values at 95% confidence levels.

Halahatang trachydacite (HA2-5379)
A total of 18 spot analyses were obtained (Table 1). Zircon grains are euhedral cylindrical, with lengths of 50-150 μm and a length/ width ratio of 4:1 to 1.5:1 (Fig. 5), and show clear and dense concentric zoning. Th/U ratios range from 0.48 to 0.89, with U contents of 57-167 ppm, which indicate a typical magmatic origin. On the 206 Pb/ 238 U-207 Pb/ 235 U concordia plot, we eliminated one analysis (analysis 4), which plots away from the concordant line. The apparent 206 Pb/ 238 U ages of the remaining 17 concordant analyses ( Fig. 6) lie between 293.3 and 281.3 Ma, yielding a weighted mean 206 Pb/ 238 U age of 287.3 ± 2.0 Ma (MSWD = 0.91, probability = 0.61) that is interpreted as the timing of trachydacite emplacement.

Wajilitag olivine-clinopyroxenite (W08)
A total of 19 spot analyses were obtained (Table 2). Zircon grains show magmatically resorbed textures, with lengths of 50-150 μm and a length/width ratio of 1:1 to 2:1 (  age of 283.2 ± 2.0 Ma (MSWD = 0.36, probability = 0.991). Based on the similarity of zircon textures and narrow distribution of single-grain ages, we interpret this age as a good indication of the crystallization age of the olivine clinopyroxenite.

Re-evaluation of published geochronological data
Since 1991, 118 ages of Permian Tarim mafic and silicic magmatism have been reported. Unfortunately, not all of these published ages present adequate information on geological context or raw  data. The ages without context and/or raw data are not considered further here (a total of 35 ages). For the ages with inconsistent geological context in two or more publications, we review the data but consider the contradictory information an open problem. For the remaining ages with clear geological context and complete raw data, we carefully evaluate each age based on criteria specific to the method used and the particular rock types. Published data are assigned into four categories: group 1, highest quality data, with clear field observations, analytical method descriptions and robust statistical perspectives (Fig. 2c); group 2, data that provide chronological constraints with less precision and accuracy, owing to either inconclusive field relationship or large errors that cannot be attributed to a specific cause or sample material issues; group 3, data with large errors, discordance or RHO value (correlation coefficient; e.g. Schmitz & Schoene 2007) outside the 0-1 range; group 4, data that lack critical information (e.g. sample location, analytical methods) or raw data. Below we briefly summarize the basic approaches taken for data evaluation for the various geochronological methods, and also summarize the results for Tarim samples.
(1) 40 Ar-39 Ar dating. The 40 Ar/ 39 Ar step-heating method uses a series of apparent ages to recover possible crystallization ages for igneous rocks, and the reproducibility of the resulting step ages requires the data to meet a basic statistical hypothesis (Marzoli et al. 1999;Baksi 2003). Alteration in basalt lavas older than a few million years may disturb igneous whole-rock material used for argon dating (e.g. Karoo, Duncan et al. 1997;Deccan, Baksi 2014) and usually results in inaccurate estimates of the crystallization age (Baksi 2007a,b). Therefore statistical tests on plateau sections for validity (Baksi 2003;Sharp & Clague 2006) and the alteration state test on dating material (Baksi 2007a,b) are simultaneously applied here to evaluate the reliability of 40 Ar/ 39 Ar dates. None of the 20 40 Ar-39 Ar dates in Tarim satisfy these statistical and alteration criteria simultaneously, and thus may not yield reliable ages.
(2) Zircons from basalt. Zircons have been found in mantlederived basalts and dated by U-Pb methods (e.g. Katayama et al. 2003;Polat et al. 2009). The question is what the relationship between the zircon and the host basalt magma really is. Experimental petrology suggests that zircons are unlikely to crystallize directly from a basalt magma (Boehnke et al. 2013), and it is probable that the zircons are crustal-derived xenocrysts entrained during either ascent or emplacement on the surface. Therefore, these zircons are rarely used to represent the age of the host rock. In the case of the zircons from the Keping basalt dated by D.Y.  and Yu et al. (2011b), difference in Hf isotope compositions of the zircons and the host basalt suggests a xenocrystic origin.
(3) Silicic extrusive and intrusive rocks. Advances in zircon TIMS and SIMS analyses have revealed a protracted history of zircon crystallization prior to eruption, of potentially more than several million years, in many silicic magmatic units (e.g. Miller et al. 2007;. Hence, the weighted mean age of single-grain U-Pb dates will represent an average age for zircon crystallization, which can be older than the eruption or emplacement age of the host rock. To address this, we replot every date of silicic rocks to try to confirm the presence of distinct populations. The zircons from silicic extrusive rocks show age scatter to variable extent, but the continuous single-grain age distribution and the relatively low resolution of the dating technique impede further interpretation. The only TIMS dating , of which the expected precision should be better than 0.1% (Mattinson 2005), reveals distinct populations of zircon, and we take the weighted mean of the youngest age population as the best approximation of the eruption age of the host rhyolite. For the Xiaohaizi syenite intrusion, the 6.2 myr age span from different rock units could result from analytical error, but is more probably attributed to a long-lived syenitic centre. Scattered single-grain ages are also observed in granite plutons, and the extent of scatter can be linked to Zr saturation temperature.
(4) Detrital zircons. Detrital zircon ages can be used as a tool to evaluate the maximum depositional age of the host sedimentary rock (e.g. Brown & Gehrels 2007). Dickinson & Gehrels (2009) proposed four measures to infer the youngest detrital zircon ages present in a sample. They are, from least to most statistically robust, (a) youngest single-grain age (YSG), (b) youngest graphical age peak controlled by more than one grain age (YPP), (c) mean age of the youngest two or more grains that overlap in age at 1σ (YC1σ(2+)), and (d) mean age of the youngest three or more grains that overlap in age at 2σ (YC2σ(3+)). The YC2σ(3+) ages of three samples from Zou et al (2013) conservatively constrain the sixth, fourth and first basalt layers of the KZ basalt at Keping to be younger than 263.7 ± 3.4, 267.0 ± 3.4 and 287.6 ± 3.8 Ma, respectively. Li et al. (2013) gave the YC2σ(3+) age of underlying sandstone of the Damusi basalt as 284.1 ± 4.7 Ma.
(5) U-Pb dating of mafic dykes. Zircon crystallizes in a differentiated gabbroic environment and can be used to represent the crystallization age of host rock (e.g. Kaczmarek et al. 2008). However, zircons in a mafic environment usually experience a complex history, which makes it necessary to identify zircon origins and group populations by variable approaches such as zircon chemistry and morphology (e.g. Grimes et al. 2009). The U-Pb dating of single mafic dyke samples of Tarim shows a large age span (e.g. Wajilitag gabbro, Zhang et al. 2009), scattered single-grain age distribution (e.g. Xiaohaizi diabase, Li et al. 2007), a large number of discordant ages (e.g. Yijianfang diabase, Zhang et al. 2009) and a large portion of xenocrystic zircons (e.g. Xiaohaizi diabase, Wei & Xu 2013). All these issues hinder a proper interpretation of previous mafic dyke dates. Furthermore, the geographically isolated dykes are used as a horizontal marker in Tarim, to chronologically relate different rock units (Li et al. 2011;Yang et al. 2013). LIP-related dyke swarms usually have an average thickness >10 m (Ernst et al. 1995) and maximum thickness >70 m (Abbott & Isley 2002), and may have extents greater than 1000 km (Ernst 2014). The Keping dykes are the thickest in Tarim, and have an average thickness of 3.8 m, maximum thickness of 21.4 m and maximum exposed length of 17.1 km (Chen et al. 2014), which are significantly smaller in dimension compared with typical LIP-related dyke swarms. Thus the dykes found in Tarim cannot be used as a horizontal marker.
(6) Kimberlite mineral U-Pb dating. Perovskite and baddeleyite from kimberlitic rocks commonly show significant compositional heterogeneity (e.g . Heaman 1989;Heaman & Kjarsgaard 2000;Sarkar et al. 2011;Schärer et al. 2011). Kimberlite eruptions can be characterized by multiple pulses sharing the same magma conduit and entrain xenocrystic material at depths throughout the lithosphere (Sparks 2013). Therefore, perovskites and baddeleyites may belong to different generations, and may not define the kimberlite eruption age. Different perovskite and baddeleyite populations may be recognized by differences in size, morphology and composition (e.g. Sarkar et al. 2011;Schärer et al. 2011), and presumably on a weighted mean U-Pb age plot.
The assumption that dated perovskites are cogenetic with the host kimberlite can also be tested using partition coefficient data from Beyer et al. (2013) to calculate melt compositions in equilibrium with the perovskites. In the study by , the large diversity of Th concentrations and large age span among single grains of both perovskite and baddeleyite, and the disequilibrium condition between dated perovskites and the Bachu kimberlite, as well as age clusters of baddeleyites, hinder a proper interpretation of these U-Pb ages.
In summary, 24 ages are highest quality, including our two new U-Pb SIMS ages (group 1; Fig. 2c; Table 3), 12 can be used for reference (group 2), and nine data have significant errors (group 3).

Re-evaluation of the temporal evolution of Permian Tarim magmatism
Several temporal evolution models for Tarim magmatism have been proposed by researchers (e.g. Li et al. 2011;Xu et al. 2014). Based on 27 published dates, Xu et al. (2014) proposed that the Permian Tarim magmatism comprises three main episodes: c. 300 Ma kimberlites, c. 290 Ma flood basalts and c. 280 Ma ultramafic-mafic-felsic intrusions and dyke swarms. The first episode is defined by three SIMS dates on perovskite and baddeleyite , which we conclude here are not robust. The second episode is defined by 40 Ar-39 Ar dates (e.g. Yang et al. 2006;Wei et al. 2014) and basalt zircon dates (e.g. Yu et al. 2011b), which were not statistically robust and lack clear geochronological significance, respectively. The third episode, consisting of ultramafic-mafic-felsic intrusions and a dyke swarm, is defined by 13 ages. The mafic dykes lack reliable age information, and the robust ages of granitic plutons, syenite intrusion, silicic dykes and ultramafic intrusions yield a >20 myr time span from 285.6 ± 3.6 Ma (Sun et al. 2008) to 261.7 ± 1.8 Ma (Zhang & Zou 2013). Based on our new ages and reassessment of a larger database of published ages (summarized in Fig. 2), this model for the temporal evolution of Permian Tarim magmatism needs to be re-evaluated.

New ages reveal the pulsed nature of mafic magmatism
The most critical problem in establishing a temporal model for Permian Tarim magmatism is the poor age constraints on the flood basalt and mafic intrusive rocks. The Halahatang trachydacite layer we dated directly overlies the uppermost basalt in the drill core, and provides a robust constraint on the minimum emplacement age (287.2 ± 2.0 Ma) of the basalt flows in northern Tarim. Our revised interpretations of published detrital zircon data provide reliable constraints on the maximum depositional age of the host sandstone and in turn restrict the maximum emplacement age of the upper part of the Keping basalt and Qimugan basalt, which overlie the sandstones, to 267.0 ± 3.4 Ma and 284.1 ± 4.7 Ma, respectively. Thus, the basalt in Halahatang is temporally distinct from the upper part of the KZ basalt at Keping, which indicates a minimum of c. 20 myr time gap between basalt eruptions. This is further supported by petrochemically distinct basalts from Halahatang, Keping and Qimugan (Shangguan 2015), and they probably represent different pulses of basalt eruptions.
The Wajilitag layered intrusion has trace element patterns and εNd(t) values indicating an ocean island basalt (OIB)-like, asthenospheric mantle source, and it is similar to the Bachu mafic dykes but compositionally distinct from the Keping basalts . The new U-Pb age of 283.2 ± 2.0 Ma is a good estimate of its emplacement age. The Piqiang ultramafic-mafic complex, 150 km NW of the Wajilitag layered intrusion, was emplaced at 262.3 ± 2.1 Ma (Zhang & Zou 2013), which is c. 21 myr later than the Wajilitag layered intrusion. Thus the two mafic-ultramafic intrusions also represent distinct magmatic pulses and are evidence for a protracted magmatic history in the Tarim Basin. Although such large mafic layered intrusions are generally inferred to be part of the plumbing system of an LIP (e.g. Cawthorn 1996), the comagmatic volcanic piles of these two intrusions have yet to be identified.
The temporally and compositionally distinct episodes of basalt eruption and mafic-ultramafic intrusion emplacement rule out the possibility of a continuous activity of the Tarim magmatism (Fig. 7). Refining the comprehensive temporal history and chemical relationship needs to be further explored though more detailed geochronological and geochemical studies. There are many examples of LIPs that show distinct pulses of magmatic activity; in particular, those with broad age ranges (e.g. >20 myr) were usually emplaced in multiple shorter duration pluses of c. 1-5 Ma rather than as a continuous longer-lasting magmatic event (e.g. Courtillot & Renne 2003;Prokoph et al. 2004;Bryan & Ernst 2008). The North Atlantic LIP shows an initial magmatic pulse at c. 60 Ma and a second magmatic pulse at c. 55 Ma (e.g. Saunders et al. 1997). Another example of a multiple pulse event is the protracted Matachewan LIP with ages of c. 2490, 2475 and 2446 Ma, represented by dyke swarms, layered intrusions and volcanic rocks, respectively (event 206 of Ernst & Buchan 2002). Our two new ages (287.2 ± 2.0 and 283.2 ± 2.0 Ma) overlap with a probability of 31.7%, which indicates that the Halahatang basalt and the Wajilitag layered intrusion may have been emplaced at the same time. With the age constraints of critical units, at least two mafic magmatic pulses can be recognized in the Tarim province: (1) the Halahatang basalt (>287.2 ± 2.0 Ma) and Wajilitag layered intrusion (283.2 ± 2.0 Ma); (2) the upper KZ basalt (<267.0 ± 3.4 Ma) and Piqiang complex (267.0 ± 3.4 Ma) (Fig. 7). Although the ages of these pulses may overlap slightly, they can also be differentiated by separate geographical location of magmatism and different chemical composition (Shangguan 2015). Determination of the duration of the Tarim province requires additional high-resolution geochronological studies, which may subdivide the currently identified pulses, fill the gaps between pulses, or broaden the currently recognized age span of magmatism.

The temporal evolution of silicic magmatism
There are 11 robust ages on silicic extrusive rocks, which are exposed in the northern margin and buried in northern Tarim. They are evenly distributed from 290.9 ± 4.1 Ma (YM30-1, Tian et al. 2010) to 271.7 ± 2.2 Ma (MN1-1, Tian et al. 2010). Three samples (HA-5379, 287.2 ± 2.0 Ma, this paper; YM30-1, 290.9 ± 4.1 Ma, and YM 5-8, 286.6 ± 3.3 Ma, Tian et al. 2010) are intercalated with basaltic lavas. Eight others are not coexistent with the basalt, and range from 286.8 ± 0.5 to 271.7 ± 2.2 Ma (Fig. 7). Silicic volcanism accompanies the first pulse of mafic volcanism in northern Tarim, and continued for at least 20 myr (Fig. 7). The eight samples that are not coexistent with the basalt and are broadly distributed in northern Tarim reasonably indicate that mafic volcanism is absent in the whole northern Tarim after c. 287 Ma.
The syenite-granite intrusions are located at the NW periphery of the Tarim magmatic province. Four ages, on different units within the Xiaohaizi syenite body, range from 285.2 ± 3.6 to 279.7 ± 2.0 Ma, and show a positive correlation with SiO 2 content, indicating a long-lived syenitic centre. Ages on four granitic plutons range from 274.6 ± 2.2 Ma (Halajun 1, C.L.  to 268.6 ± 2.0 Ma (Halajun 3, Zhang & Zou 2013), and do not overlap in age with the syenite intrusion (Fig. 7). Studies have shown that LIP-related syenitic complexes can have significant age spans of c. 10 myr (e.g. Kangerlussuaq Alkaline Complex, Greenland, Riishuus et al. 2006), and the timeframe of granitic pluton assembly is potentially several million years (e.g. >5 myr for Mt. Stuart, WA, USA; >8 myr for Tuolumne, CA, USA; Miller et al. 2007). Thus, the age ranges observed here probably reflect a protracted magmatic history. The syenite and granite plutons represent two discrete, episodic phases of magmatism, at c. 285-280 Ma and c. 275-269 Ma, respectively (Fig. 7). The silicic intrusive rocks are roughly synchronous with the extrusive silicic rocks, with some older ages found in extrusive silicic units (Fig. 7) (Cohen et al. 2013). Inset in the lower left corner shows the probability density plot of the two new ages. It should be noted that the two ages overlap with a probability of 31.7%. rocks are associated with most continental LIPs (Bryan et al. 2002;Bryan 2007), and can predate, be coincident with, or postdate the main phase of basalt magmatism (Bryan et al. 2002;Ernst 2014).
Based on our new database, the silicic magmatism in Tarim province temporally overlaps with the first pulse of the mafic magmatism (Fig. 7).

P a l e o t e t h y s
Qiangtang Dyke swarm P a n g e a P a n g e a Australia  Zhai et al. (2013), Domeier & Torsvik (2014) and Xiao et al. (2014). The dark areas show the recognized LIP-scale or sub-LIP-scale magmatic provinces between 300 and 250 Ma. Small map in the upper right corner shows the present-day positions of these magmatic provinces.

Is the Permian Tarim magmatic province an LIP?
Large igneous provinces are commonly defined by area and volume, as well as duration of magmatism: >100 000 km 2 , 100 000 km 3 and a maximum lifespan of up to 50 myr, respectively (Bryan & Ernst 2008;Ernst 2014). The estimates for current Tarim areal extent range from 150 000 km 2 (Chen et al. 1997a,b) to 265 000 km 2 , with a volume of 40 000 km 3 , based on surface exposures and subsurface seismic and drill core data (Pan 2011;Pan et al. 2013). These estimates ignore the effect of erosion and the intrusive component of the province, as well as pyroclastic deposits and silicic lavas. Mafic layered intrusions exposed in NW Tarim indicate that there may have been extensive erosion of surface deposits. There are no available volume estimates for Tarim intrusions, and their equivalent in other LIPs can be volumetrically significant (e.g. the Dufek complex in the Ferrar LIP is 60 000 km 3 , Elliot et al. 1999).
Pre-Mesozoic LIPs can be intensely fragmented by continental break-up such that portions of a single LIP are dispersed on different crustal blocks (Ernst 2014), and Tarim may represent a remnant of an LIP, similar to the Coppermine River basalts in Canada (Baragar et al. 1996). In central Asia there are three mafic provinces with similar ages: the Panjal traps in northern India (c. 290 Ma, Shellnutt et al. 2011Shellnutt et al. , 2014, Qiangtang dyke swarm in Tibet (c. 283 Ma; Zhu et al. 2010;Zhai et al. 2013), and mafic-ultramafic intrusions (284.0 ± 2.0 to 278.6 ± 1.2 Ma, Qin et al. 2011) in the Beishan accretionary belt. The Panjal and Qiangtang provinces are roughly coeval with early Tarim mafic magmatism; however, Permian palaeogeography suggests that they are not likely to be cogenetic, given that they are separated by the Palaeotethys at c. 290 Ma (Fig. 8; Domeier & Torsvik 2014). Instead, the approximately coeval mafic-ultramafic intrusions (284.0 ± 2.0 to 278.6 ± 1.2 Ma; Qin et al. 2011;Xiao et al. 2014) in the Beishan accretionary belt may potentially relate to the Tarim magmatic province, given they were adjacent in the early Permian (Fig. 8).

Conclusions
(1) The new SIMS U-Pb zircon dating of the Halahatang trachydacite layer constrains the youngest emplacement age of the basalt to be 287.2 ± 2.0 Ma in the northern Tarim. The olivine clinopyroxenite from the Wajilitag layered intrusion constrains the crystallization age of the layered intrusion to be 283.2 ± 2.0 Ma.
(2) One hundred and eighteen published geochronological ages were evaluated. Twenty-four of them are robust, including our two new ages.
(4) The silicic magmatism spans at least 20 myr, from 290.9 ± 4.1 to 271.7 ± 2.2 Ma, and temporally overlaps with the first pulse of mafic magmatism.
(5) At the current stage of knowledge, the Tarim magmatic province is of insufficient volume to qualify as an LIP based on the criteria of Ernst (2014). The coeval mafic-ultramafic intrusions in the Beishan accretionary belt may potentially relate to the Tarim magmatic province. Inclusion of silicic volcanic rocks and intrusive complexes, plus more accurate estimates of erosion, will all increase the estimated original emplacement volume.