The data for m z and m z were

The data for m/z 10 and m/z 11 were collected simultaneously on L3 and H3 Faraday cups. Their on-peak-zero was measured for 27 s before the laser was on. After the laser was powered on, the measurements were taken in 400 cycles for each spot, and the integration time for each cycle was of 0.131 s. The 11B/10B of each spot was calculated after subtracting the on-peak zero on m/z 10 and m/z 11. Corrections for instrumental drift, mass bias between 10B and 11B were both conducted by a “standard-sample-standard” bracketing external standardization technique. One piece of standard tourmaline IAEA B4 with a size of ∼3 × 3 × 2 mm was used for this purpose. Two spots on the IAEA B4 were measured before and after each 10 spots measurements of unknown samples, and the mass bias of the 11B/10B of the unknown samples were calibrated by reference to the mean of the four 11B/10B ratios of the IAEA B4 to its certificate value (Gonfiantini et al., 2003; Le Roux et al., 2004). Off-line data reduction (including selection and integration of background and analyte signals) was performed by a spreadsheet program created in house by the authors. The time-resolved signal of single isotopes and isotope ratios was carefully inspected to verify the presence of perturbations related to inclusions, fractures and mixing of different sample domains. The final result of the B isotopic analyses is expressed in terms of δ11B, which is defined as follows: δ11B (‰) = [(11B/10B)Sample/(11B/10B)Standard-1] × 1000, where the standard was NIST SRM 951 boric prostaglandin synthase from the National Institute of Standard Technology (11B/10BNIST SRM 951 = 4.05003 (Ishikawa and Tera, 1997; Ishikawa et al., 2001). The internal precision for each spot after calculated into δ11B was generally better than 0.1‰ (1σ). Two international tourmaline reference standards (dravite (#108796) and schorl (#112566)) (Dyar et al., 2001; Leeman and Tonarini, 2001) and two laboratory tourmaline standards (IMR RB1 and IMR RB2) (Hou et al., 2010) were repeatedly measured along with the samples to monitor the quality of the measurements. The external precision of the δ11B was generally better than ±0.5‰ (2 SD), and their values are all identical to the reported reference values within analytical errors (Table 3), demonstrating the reliably of the analytical procedure.
Analytical results
Discussion Boron isotope fractionation between tourmaline, aqueous fluid and granitic melt is a highly controversial topic that has received much discussion in the literature (e.g., Palmer et al., 1992; Jiang and Palmer, 1998; Hervig et al., 2002; Tonarini et al., 2003b; Meyer et al., 2008; Trumbull et al., 2008; Marschall et al., 2009b; London, 2011; Marschall and Jiang, 2011). Based on studies on volcanic glasses, Tonarini et al. (2003b) suggest that B isotope fractionation is mostly related to the relative amount of trigonal and tetrahedral boron sites in the glass network rather than to other processes, including the speciation of hydrous species in the glass structure. A basic consensus is that the nature and extent of isotopic fractionation between tourmaline and fluids is different from that between tourmaline and melts (Leeman and Sisson, 1996; Marschall and Jiang, 2011). Trigonally co-ordinated B(OH)3 is the only significant species that occurs in a solution or fluid due to their low pH (Palmer et al., 1992), but B generally occurs as tetrahedral co-ordinated B(OH)4− in a melt (Jiang and Palmer, 1998). In general, 11B is preferentially partitioned into aqueous fluids during heating and decomposition of hydrous minerals (clay minerals and white mica) or during fluid–rock interaction (e.g., Jiang and Palmer, 1998; Marschall and Jiang, 2011). Thus, tourmaline directly crystallizing from a fluid has higher δ11B values than the clays and micas from which the fluid originated (Marschall and Jiang, 2011). However, if granitic tourmaline crystallized directly from a melt, there should be relatively little isotopic fractionation as the change in the boron symmetry during tourmaline crystallization is small; i.e., the δ11B values of granitic tourmalines would be similar to those of the melt (Jiang and Palmer, 1998).