The challenge: analysis of tiny inclusions
The glassy, partially crystallized or nanogranite inclusions from the case studies analysed so far are small, especially at Ronda, where their diameters are often <5µm. The small size represents an analytical challenge as it is near or below the limits of the spatial resolution of several conventional microanalytical techniques. It should be pointed out that the primary difficulty for the analysis is sample preparation, as uncovering and polishing the inclusions for SEM and EMP characterization is very difficult and often results in the mechanical removal of the inclusion content.
Concerning the major element analysis of crystals, the electron microprobe often provides compositions that are contaminated by adjacent or underlying phases. This problem could be solved by using the new generation of FEG-based microprobes, as well as standardized EDS analysis by SEM. An additional and important problem is the loss of Na from the inclusion material, that increases by focussing the beam size and increasing the beam current. Sodium loss is particularly important during the analysis of hydrous felsic glasses (Morgan and London, 1996, 2005). In the absence of a N2-cooled EMP cryostage (e.g. Clemens, 2009), our approach for the analysis of silicate glass in inclusions has been that suggested by Morgan and London (1996, 2005): to correct the data for Na loss by comparison with the behaviour, at the same analytical conditions, of rhyolitic glass standards with variable H2O contents (e.g. Cesare et al., 2009).
It is the bulk composition of the melt that produced the nanogranite that is of major petrological and geochemical interest, rather than the composition of the individual mineral phases that constitute the nanogranite. Therefore, nanogranites and partially crystallized inclusions are re-melted, and the glass obtained upon rapid quenching is analysed with the techniques described above. Remelting has been performed at ambient pressure using the high-temperature stage, a routine technique in igneous petrology (Frezzotti, 2001); although in general the composition of the remelted nanogranite is comparable with that of preserved glassy inclusions found in the same crystal (see below, and Cesare et al., 2009), the results are not entirely satisfactory as remelting by the high-temperature stage often induces decrepitation accompanied by volatile loss and interaction with host mineral (Figure 14C). As a consequence nanogranites have been also remelted in a piston cylinder. Experiments at 5 kbar with the samples from Ronda have produced the complete remelting of crystalline inclusions, without decrepitation or production of peritectic phases (Figure 14D). This technique, although time-consuming, appears to be more suitable than the heating stage for restoring the bulk composition of the primary melt in the inclusion. In this regard, perfect remelting without the growth of peritectic phases or the retention of mineral residues supports the idea expressed above that the melting of preexisting solid inclusions suggested by Vernon (2007) may be realistic, but requires: 1) the presence of nanogranite as inclusions - i.e. the rock must have already undergone a previous partial melting event with growth of peritectic phases and MI entrapment; and, 2) the onset, in the new anatectic event, of the same PT conditions as those which led to the nanogranite formation, to obtain a perfect remelting of the inclusion.
If the MI really behave as closed systems, remelting without MI decrepitation potentially allows any exsolved fluid to be re-dissolved into the melt phase, and the primary fluid content of the anatectic melt to be measured after quenching. A first-order approximate indication of the melt volatile content can be obtained by the difference between 100 and the EMP total, if the analytical setup follows the recommendations of Morgan and London (1996, 2005). Because H2O is the main volatile in S-type felsic magmas, a more precise quantification of the volatile concentrations in glassy or remelted inclusions can be performed by SIMS, IR- or Raman spectroscopy; given the small size of the MI, the most promising method among these appears to be Raman spectroscopy (Thomas et al., 2006).