Middle Triassic arc magmatism along the northeastern margin of the Tibet: U–Pb and Lu–Hf zircon characterization of the Gangcha complex in the West Qinling terrane, central China

The tectonic setting of Mesozoic magmatic complexes in the northeastern margin of the Tibet plateau is disputed, and hence gives rise to uncertainty concerning the tectonic evolution of the northeastern Tibet Plateau and the timing of the closure of the Palaeo-Tethys ocean. The Gangcha complex is typical of these complexes, consisting of andesite, dacite, gabbro, gabbroic diorite, granodiorite, quartz diorite, and diorite with typical chemical traits of continental margin arc rocks. Andesite, gabbroic diorite, and mineralization-associated potassic-altered diorite yield weighted mean 206Pb/238U ages of 242.1 ± 1.2 Ma, 243.8 ± 1.0 Ma and 234.0 ± 0.6 Ma respectively. Zircon ϵHf(t) for magmatic grains ranges from −3.5 to +5.7, interpreted to demonstrate that the Gangcha complex contains crustally contaminated mantle magmas. Inherited zircons in the complex yield similar U–Pb ages (777–310 Ma) to the A’nyemaqen composite ophiolite assemblage with ϵHf(t) of −17.4 to +11.6. This suggests that components of this older ophiolite melted and contributed to the Gangcha complex magmas. Hence the Gangcha complex is considered to have formed as a continental margin arc in northeastern Tibet by northward subduction during consumption of the Palaeo-Tethys ocean. Regionally, it corresponds to the arc magmatism along the eastern and western Kunlun sutures to the west and the Mianlue suture to the east. Supplementary material: Geochemical data, and zircon U–Pb and Lu–Hf data for the Gangcha complex are available at www.geolsoc.org.uk/SUP18521.

The West Qinling terrane (Fig. 1a) in the northeastern margin of the Tibet plateau is an important part of the Chinese Central Orogenic Belt between the North China and Yangtze Cratons. This terrane contains the convergence of the Qinling, Qilian, Kunlun and Songpan-Ganze orogens (Fig. 1b), in the northernmost part of the west Qinling-Songpan-Ganze region, central China (Zhang et al. 2004). It is important for understanding the tectonic evolution of northeastern Tibet, but little is known of the region because of severe superimposed deformation and extensive Mesozoic magmatic modification here. Abundant Cu-Au, Cu-Mo, Au, and Pb-Zn deposits have been explored in this area in recent years, but there are no comprehensive geochemical or geochronological data for the igneous rocks associated with these mineral deposits. Undoubtedly, the tectonic settings and ages of the magmatic rocks and associated mineral deposits are important, not only to understand the evolution of the Chinese Central Orogenic Belt but also to elucidate the tectonic history of the Tibet plateau and east Asia.
The Gangcha complex (Fig. 1c) is a typical Middle Triassic magmatic assemblage in the West Qinling terrane, which hosts several porphyry-skarn-type deposits. The associated Xiekeng Cu-Au-Fe deposit has been explored intensively. Previously, the tectonic setting of the Gangcha complex and other associated volcanic associations in the West Qinling terrane and Chinese Central Orogenic Belt have been interpreted as formed either in post-collisional settings (e.g. Sun et al. 2002;Zhang et al. 2005;Zhang et al. 2006a, b) or in magmatic arc settings (e.g. Jiang et al. 2010;Guo et al. 2011). These two interpretations are still disputed, which seriously restricts understanding of the tectonics of northeastern Tibet and the history of the Palaeo-Tethys ocean.
In this paper, we report new results for the Gangcha complex associated with the Xiekeng deposit ( Fig. 1c) in the West Qinling terrane. The presented major and trace element geochemistry, zircon U-Pb ages, and Hf isotope data for a suite of gabbro, gabbroic diorite, andesite, and dacite strongly suggest that the Gangcha granitoids as well as andesite and dacite assemblages in the West Qinling terrane were formed in a Middle Triassic continental arc environment, coupled with northward subduction of the Palaeo-Tethys ocean crust along the northeastern margin of the Tibet plateau.
of the Triassic clastic rocks in the Huashixia area (Yan et al. 2008) and tectonic setting of the Triassic granitoid plutons (Luo et al. 1999) in the southern margin of the Kunlun Mountains further support this inference. In addition, some mafic-ultramafic pillow lava, gabbro and chert blocks along the southwestern margin of the Qinghai Lake probably represent Carboniferous oceanic crust (Wang et al. 2001;Guo et al. 2009), which separated the West Qinling terrane in the south from the Qilian orogen to the north. However, its continuation eastwards is uncertain because of widespread and voluminous Triassic clastic rocks and plutons. Xiao et al. (2009) interpreted the Qilian orogen as an Altai-type orogenic belt (Sengör & Natal'in 1996) that formed from the early Palaeozoic  Feng & Zhang 2002). (c) Regional geological map of the Gangcha complex. (d) Geological map of the Xiekeng Cu-Au-Fe deposit (after Dong et al. 2010). The stars in (c) represent the ore deposits. 1, Jiangligou Cu-Mo deposit; 2, Langmujia Cu deposit; 3, Shuangpengxi Au-Cu deposit; 4, Tiewu Fe-Au-Cu deposit; 5, Dehelongwa Au-Cu deposit; 6, Xiekeng Cu-Au-Fe deposit; 7, Xiechangzhigou Cu deposit; 8, Hongqika Au-Cu deposit.
to Devonian, consisting of collisional terranes to the south and accretionary belts to the north and west.
The Devonian-Permian sediments in the West Qinling terrane were suggested to be forearc deposits based on their provenance features (Yan et al. 2006a(Yan et al. , b, 2007Chen et al. 2010). Permian pillow lava, ultramafic rocks and limestone crop out in the Longwuxia and Bajiaocheng areas. Generally, they are dismembered and are in fault contact with the Triassic sedimentary rocks. Petrological and geochemical data for these volcanic blocks indicate that they are consistent with a suprasubduction-zone (SSZ)-style ophiolite sequence derived from a mature island arc environment (Zhang et al. 2007;Wang et al. 2009). These NW-SE-striking rock assemblages were intruded by the Gangcha complex, which also straddles the boundary between Permian limestone and Lower Triassic clastic rocks (Fig. 1c). The Triassic clastic rocks with abundant andesite, dacite and rhyolite interlayers in this area were also interpreted as deposited in a closing ocean basin (e.g. Jiang et al. 1996) or a remnant basin (e.g. Wang 2009). The Triassic granitoids in the West Qinling terrane have previously been interpreted as the result of post-collision magmatism (e.g. Sun et al. 2002;Zhang et al. 2005Zhang et al. , 2006bQin et al. 2010;Zhu et al. 2011), and host porphyry-and skarn-type Cu-Mo and Cu-Au(-Fe) deposits, such as the Wenquan deposit, c. 250 km SE of Tongren, and the Xiekeng deposit.
The Xiekeng Cu-Au-Fe deposit is hosted by the Gangcha complex and Permian lenticular limestones ( Fig. 1c and d). The Gangcha complex mainly consists of gabbro, gabbroic diorite, dacite and andesite . Dacite and andesite were intruded by potassic-altered diorite. The temporal relationship between the dacites and andesites is unknown. Gabbroic diorites cut gabbros, and thus are younger. Copper and gold veinlets (Fig. 2a) are mainly developed within the potassic-altered diorite. A distinct potassic zone developed within the altered diorite, which consists of K-feldspar, biotite and quartz, indicating a porphyry-type deposit, together with the propylitic alteration (Fig. 2b). Moreover, garnet + diopside + epidote + actinolite assemblages ( Fig. 2c and d) developed within the contact zone between the diorite and the limestone, giving a typical calcite-potassic wall-rock alteration zone. The Permian limestones were strongly disrupted by faults, with abundant pyrite and gold-bearing quartz veinlets distributed along faults or fractures ( Fig. 2e and f). Gold and copper grades are 1.14-31.14 g t− 1 and 0.4-15.7 wt% (Dong et al. 2010), respectively. Lenticular magnetite-chalcopyrite bodies mainly occur within the fractured zone of limestone; the total iron (FeO T ) grade of the eight magnetite ore bodies ranges from 38.1 to 64 wt% (Dong et al. 2010).

Sample description
Thin sections of 45 Gangcha complex samples were screened to reject those that are too strongly altered for geochemical and geochronological analysis. The remaining suite of 17 samples covered the full lithological range of gabbro, gabbroic diorite, andesite and dacite.
Gabbros have a 1-3 mm grain size, are overall equigranular, and are composed of pyroxene (50%) and plagioclase (40%) with minor biotite, amphibole, epidote, chlorite and olivine (Fig. 3b). The dominant accessory minerals are Fe-Ti oxides. Generally, clinopyroxene is about 70% of the pyroxene and is 2-3 mm in size and the orthopyroxene grains are 1-3 mm in size, and most of them are strongly altered to amphibole and chlorite.

Analytical methods
The 17 representative and relatively fresh samples for geochemical analyses were cleaned, and all weathered and veined surfaces were removed. Each sample was powdered using an agate mill, and fusedglass beads were prepared for major element analysis utilizing a Phillips PW 4400 X-ray fluorescence (XRF) spectrometer with a rhodium X-ray source. FeO was determined by digestion with HF and H 2 SO 4 and titration with standard potassium dichromate solution, using the diphenylamine sulphonate indicator. Loss on ignition (LOI) was measured by weighing before and after 1 h in a furnace at 1000 °C. All major and trace elements were analysed at the National Research Center, Chinese Academy of Geological Sciences, Beijing. International standards GSR1, GSR2 and GSR3 were used to monitor analytical quality control. Detection limits for major elements were <0.01% (except for TiO 2 and MnO, which had detection limits <0.001%). Detection limits for minor and trace elements were 1-0.05 ppm. Based on the analyses of international reference materials, the analytical precision for all major oxides by XRF is estimated to be better than 1%. Trace element concentrations were determined by inductively coupled plasma mass spectrometry (ICP-MS) using a VG Elemental PQII Plus system. All measurements of the trace elements were corrected for instrumental drift using the peak intensities of 115 In and 185 Re internal monitors. The results of standard analyses were consistent with their reference values within the published error ranges, respectively, with the differences for trace elements, including REE, within 5-10%. Details of trace element analytical procedures have been described by Qi et al. (2000).
Zircons from three samples were dated using U-Pb techniques to constrain the timing of magmatism and to detect any pre-magmatic components. Zircon grains were separated using standard heavy liquid and magnetic separation techniques. U-Th-Pb analyses were carried out by laser ablation-multicollector (LA-MC)-ICP-MS using the Neptune instrument at the Institute of Mineral Resources, Chinese Academy of Geological Sciences, Beijing. Handpicked zircons were mounted in epoxy resin, ground to present grain half-sections, and then polished. Reflected and transmitted light photographs and cathodoluminescence (CL) images were acquired to obtain information on the internal structures of the grains and to target specific areas therein. U-Th-Pb isotope ratios were determined relative to the Pl and GJ-1 standard zircons (Jackson et al. 2004;Sláma et al. 2008), and their concentrations were calibrated relative to the M127 reference zircon (Nasdala et al. 2008). The instrumental techniques are similar to those described by Hou et al. (2009). Data were reduced according to the procedure of Liu et al. (2010), and assessed using Isoplot 3 (Ludwig 2003). The analytical data are presented as 1σ on the concordia plots. Uncertainties in mean ages are quoted at the 95% confidence level. In situ zircon Hf isotope analysis was carried out using a New Wave UP213 LA microprobe, attached to a Neptune MC-ICP-MS system at the Institute of Mineral Resources, Chinese Academy of Geological Sciences, Beijing, employing a 55 μm spot. Helium was the carrier gas to transport the ablated sample from the LA cell to a mixing chamber where argon was added, prior to introduction to the ICP-MS system. Zircon GJ-1was used as the reference standard, with a weighted mean ratio for 176 Hf/ 177 Hf of 0.281995 ± 11 (2σ, n = 13) during acquisition of the data presented here. Instrumental conditions and data acquisition were comprehensively described by Wu et al. (2006) and Hou et al. (2007).

Geochemistry
Major-element results indicate an increase in SiO 2 and Al 2 O 3 from gabbro, gabbroic diorite, andesite to dacite, whereas TiO 2 , FeO T , MnO, MgO and CaO generally decrease, as well as Mg-number values (Table 1). On the K 2 O v. SiO 2 binary diagram proposed by Rickwood (1989), the studied rocks fall within the calc-alkaline and high-K calc-alkaline domains (Fig. 4a). On a molar A/NK v. molar A/CNK diagram (Maniar & Piccoli 1989), the chemical compositions reflect the metaluminous nature of the Gangcha complex (A/NK = 1.30-2.62; A/CNK = 0.45-1.07; Fig. 4b). The dacite is different from other samples in these plots because of relatively low K 2 O and high Al 2 O 3 concentrations, which in this case probably reflects sericite alteration.
Apart from three gabbroic diorite samples (09XK39, 40, 41) with similar Nb, Ta and La concentrations to andesite and dacite, the gabbro and gabbroic diorite samples have high Cr, Co, Ni, Sc, Y and Yb, and low Nb, Ta, Ba, Sr and La concentrations ( Table 2). The average primitive mantle-normalized trace element patterns of the studied rocks exhibit extreme large ion lithophile element (LILE) enrichment with respect to high field strength elements (HFSE) and show negative Nb, Ta, P and Ti, and positive Pb and Sr anomalies (Fig. 5a). Their chondrite-normalized REE patterns are very similar and display apparent light REE (LREE) enrichment and heavy REE (HREE) depletion (La N /Yb N = 2.16-10.19) and small but distinct Eu anomaly (Fig. 5b).

Zircon U-Pb data
Samples of andesite, gabbroic diorite and mineralization-associated potassic-altered diorite were used for zircon U-Pb dating. Typical zircon CL images are shown in Figure 6. The concordance  of each analysis spot is higher than 95% (except 09XK30-2). All zircons are transparent and colourless, with variable shape, size and internal structure. Most zircons from the andesite (sample 09XK30) exhibit welldeveloped long prismatic crystal faces with distinct magmatic oscillatory zoning (Fig. 6a), and range from 150 to 300 μm long with aspect ratios of 3:1 to 4:1. Some grains have a core-rim structure consisting of an inherited core with distinct oscillatory and patchy zoning (Fig. 6a, spots 24 and 25) and a thin rim (150-200 μm across; aspect ratio of 2:1). Generally, the width of the rim is too small for single-spot analysis. Two inherited cores yield 206 Pb/ 238 U ages of 310 ± 4 (spot 2) and 777 ± 7 Ma (spot 25) with Th/U ratios of 0.44 and 1.65, respectively. Four cores yield a 206 Pb/ 238 U weighted mean age of 452.7 ± 3.6 Ma with Th/U ratios from 0.54 to 0.96 (spots 10, 23, 24 and 31). Twenty-five grains with typical magmatic oscillatory zoning yield a similar 206 Pb/ 238 U age ranging from 236 to 247 Ma, with a weighted mean age of 242.1 ± 1.2 Ma (MSWD = 0.68; Fig. 7a), which represents the magma emplacement age. Th (54-282 ppm) and U contents (91-503 ppm) vary considerably, with generally high Th/U ratios, from 0.29 to 0.9 (average 0.61).
The zircons from the mineralization-associated potassic-altered diorite (sample 10XKTK3) are colourless, with subhedraleuhedral prismatic habit (c. 200-350 μm in diameter) with aspect ratios of 2.5:1-4:1. CL imaging shows that most of these grains have distinct magmatic oscillatory or band zoning (Fig. 6c), and others are homogeneous. Several grains have a core-rim texture consisting of a detrital core with clear and faint oscillatory zoning and patchy zonation and a very thin overgrowth rim. Nineteen spot analyses were obtained from the magmatic oscillatory or band zoned zircon. They yield similar 206 Pb/ 238 U ages ranging from 232   6. Representative CL images of zircon grains from magmatic rocks of the Gangcha complex. (a) Long prismatic zircons from the andesite sample 09XK30 with oscillatory zoning (spots 1, 11, 12, 21 and 28) and the inherited zircons with core-rim structure consisting of an inherited core with distinct oscillatory (spot 25) and patchy zoning (spot 24) and a thin rim. (b) Zircons from the gabbroic diorite sample 09XK47 mainly consist of prismatic grains that are homogeneous in CL images (spots 4, 6, 23 and 26) and equigranular grains with banded (spots 10 and 16) and sector (spots 14 and 16) zoning in CL images. (c) Long prismatic zircon grains from the potassic-altered diorite sample 10XKTK3 with distinct magmatic oscillatory or band zoning (spots 4, 5, 7, 8 and 14), core-rim texture (spot 10), and homogeneous (spots 3 and 11) in CL images. The continuous-line circles and dashed circles represent locations of U-Pb dating and Lu-Hf analysis, respectively.
Two spot analyses (spots 7 and 10) on the zircon cores with magmatic oscillatory zoning yield a weighted mean age of 419.3 ± 3.3 Ma (MSWD = 0.08), with Th/U ratios 0.14 and 0.93, respectively.

Zircon Lu-Hf data
Lu-Hf isotopes were measured on all three U-Pb dated samples. The ε Hf (t) variation with zircon age is shown in Figure 8, where the dotted lines represent the possible source for the Triassic magmatic rocks.
The 236-247 Ma zircons from andesite 09XK30 show a range of initial 176 Hf/ 177 Hf ratios from 0.282601 to 0.282784, and slightly negative to mainly positive ε Hf (t) values of −0.8 to +5.7. The Hf T DM1 model ages for these zircons are 651-941 Ma. Zircon 10XK30-2 is distinct, with an initial 176 Hf/ 177 Hf ratio of 0. Zircons in potassic-altered diorite of 10XKTK3 can be divided into three groups based on the U-Pb ages and Hf isotopic composition. Zircon 10XKTK3-7 is characterized by an initial 176 Hf/ 177 Hf ratio of 0.282717, an ε Hf (t) value of +7.3 and Hf T DM2

Petrogenesis
The high MgO and Cr contents and Mg-number preclude a purely crustal origin for any of the rocks in this study (Wedepohl 1995). The similar chondrite-normalized REE patterns and primitive mantle-normalized trace element abundances indicate that the various rocks of the Gangcha complex formed in the same environment, probably an arc system. However, higher LREE/HREE ratios and SiO 2 contents, and lower Ni, Co and Cr abundances of the dacite and andesite in contrast to the gabbroic diorite and gabbro suggest that dacite and andesite could be derived from a more fractionated magma source than the gabbros.
The relationship between the major and trace element concentrations and zircon Lu-Hf isotopic data show that the Gangcha complex originated from hybridization of mantle magmas by crustal contaminants. The negative correlation between SiO 2 and Ni, Cr and Co suggests that the gabbro, gabbroic diorite, andesite and dacite are cogenetic, probably dominated by olivine, spinel or clinopyroxene fractionation (Green 1980). High concentrations of LILE (Rb, Ba, Th and K) and low concentrations of HFSE (Nb, Ta and Ti) are the basis for geochemical schemes to distinguish subduction-related magmas from those generated in other tectonic settings (Pearce & Cann 1973;Pearce 1983;Hawkesworth et al. 1993;Macdonald et al. 2000), generated predominantly by the fusion-fluxing of the depleted mantle wedge (McCulloch & Gamble 1991;Macdonald et al. 2000). Seven inherited zircons yield U-Pb ages of 777, 451-450, 419-420, and 310 Ma with negative (−17.4 to −3.7) and positive (+2.9 to +11.6) ε Hf values, indicating that the 245-235 Ma magmatism incorporated pre-Triassic material of both evolved and juvenile characteristics. Spatially, the andesite and dacite are closely associated with gabbro and gabbroic diorite. The close age relationship of the volcanic and magmatic elements of the Gangcha complex is similar to that of continental arc volcano-plutonic complexes, such as in the Andes and Mexico (Bagby et al. 1981;Gutscher 2002).
The U-Pb zircon dating shows that with an age of 234 ± 0.6 Ma, the potassic-altered diorite formed 8-9 Ma after the andesites and gabbroic diorites. In porphyry-type deposits, there is late release of a metal-charged aqueous phase along with residual magmas from the cooling parental chambers, with the precursor plutons and porphyry Cu stocks generally separated by as little as 1-2 Ma (Sillitoe 2010). Based on Pb and S isotopes, the ore-forming fluids of the Shuangpengxi Cu-Au and Dehelongwa Au-Cu deposits, which are also hosted in the Gangcha complex, are of mixed crust-mantle mixed origin . This is consistent with the zircon Hf isotope compositions of the potassic-altered diorite, which is the host rock of copper mineralization. Therefore, the Cu mineralization of the Xiekeng deposit is likely to be contemporary with intrusion of the potassic-altered diorite.
The Triassic zircon populations from the Gangcha complex yield slightly negative to mainly positive ε Hf values (09XK30, −0.8 to +5.7; 09XK47, −3.5 to +4.8; 10XKTK3, −0.7 to +5.0), suggesting that the magmatism was dominated by juvenile material. The positive ε Hf values of 2.9 (310 Ma), 7.3 (420 Ma), 10.3 (450 Ma) and 2.9-11.6 (451 Ma), with corresponding Hf T DM model ages of 1139, 944, 779 and 695-980 Ma, indicate that crust formed in the Meso-to Neoproterozoic contributed to the Triassic magmatism. Two inherited zircon grains with negative ε Hf values of −17.4 (777 Ma) and −3.7 (419 Ma) have Hf T DM model ages of 2783 and 1643 Ma. These integrated U-Pb and Hf data indicate the contribution of some much older crust to the magma system. The U-Pb ages of these inherited zircons agree with the two generations of ophiolite complex along the A'nyemaqen fault (Bian et al. 2004), indicating that melting of these A'nymaqen rocks provided some contribution to the Gangcha complex.
Regionally, abundant geochemical and isotopic data indicate that an Andean-type margin developed along the southern margin of the North China plate during the late Silurian-early Devonian (e.g. Lerch et al. 1995;Yan et al. 2006aYan et al. , b, 2007Zhang et al. 2006a). On the other hand, petrological, geochemical and isotopic data for mafic and ultramafic blocks, limestone, and cherts from the Tongren and Bajiaocheng areas in the West Qinling terrane (Fig.  1c) indicate that Carboniferous-Permian ophiolite units that formed in a suprasubduction-zone setting are present (Zhang et al. 2007;Wang et al. 2009), which might correspond to the Guoke ophiolite in the southwestern margin of the Qinghai Lake (Wang et al. 2001;Guo et al. 2009). To the south, abundant Triassic flysch with the A'nyemaqen composite ophiolite complex, together with Triassic arc magmatism within the Kunlun Mountains (Luo et al. 1999), indicate northward subduction-accretion at the edge of the Palaeo-Tethys ocean (Xu et al. 1996;Bian et al. 2004;Yan et al. 2008). Recently, northward subduction-accretion of the Palaeo-Tethys ocean was proposed, based on the early to middle Triassic granitoids along the Mianlue suture in the Qinling orogenic belt, which connects with the A'nyemaqen suture to the west and possibly even to the Western Kunlun sutures (Xiao et al. 2002(Xiao et al. , 2005Jiang et al. 2010;Dong et al. 2011). This indicates that an active margin could have developed along the southern margin of the Palaeo-Tethys ocean during the early to middle Triassic.
We therefore propose that along the northern Tibet plateau, northward subduction-accretion at the margin of the Palaeo-Tethys ocean occurred in the Triassic (244-235 Ma). The ascending magmas with calc-alkaline geochemical affinity were contaminated by Neoproterozoic-Carboniferous (777-310 Ma) crust as recorded in inherited zircons and zircon Hf isotopic signatures, sourced from a Palaeozoic subduction-accretion complex along the northern Tibet Plateau.

Conclusions
(1) The Gangcha complex, which is associated with the Xiekeng deposit of the West Qinling terrane, is composed of sub-alkaline, metaluminous gabbro, gabbroic diorite, andesite and dacite. The geochemical characteristics of rocks within this complex are typical of subduction-related arc magmas. Detection of older inherited zircon cores and the spread in zircon initial Hf isotopic signatures show that the mantle magmas were contaminated by older crust during their ascent.
(2) Zircon U-Pb ages of andesite and gabbroic diorite of 242.1 ± 1.2 and 243.8 ± 1.0 Ma indicate that the Gangcha complex formed in the Middle Triassic. The Cu-Au mineralization took place in the late stage of the magma evolution synchronous with the 234.0 ± 0.6 Ma potassic-altered diorite.
(3) The Gangcha complex in the West Qinling terrane belongs to a Middle Triassic active continental arc, which was formed by northward subduction of the Palaeo-Tethys ocean along the northeastern margin of Tibet. This study is supported by the National Science Foundation of China (41172178) and the Geological Survey Project of China (1212010911033, 1212011120159). Y. Ling and Y. Chun are acknowledged for the assistance in making mounts and CL imaging. A. Nutman of the University of Wollongong is thanked for improving the English. Special thanks are due to W. Xiao, M. Flowerdew and subject editor T. Rooney for their valuable and constructive suggestions and comments, which greatly improved the paper.