Fig. 3.17. Results of trace element modelling. (a) Shows the samples selected for modelling and their SiO2 concentrations and Mg#’s. (b) The probability that cumulate-bearing lavas are derived through fractional crystallisation of Sumb81. Small numbers represent corresponding XLFRAC Least Squares Model. Elements which provide a satisfactory solution are closer to 1and the grey line represents within 20% error. (c) The probability that lava trace element concentrations can be produced through fractional crystall- isation of Sumb81. (d) The probability that trace element concentrations in the pyroclastic deposits can be produced through fractional crystallisation of Sumb81. (e) A model involving other potential parent lava compositions. (f) Modelling between two pyroclastic deposits. Mineral phases and degree of crystallisation used from least square models above. Partition coefficients used from selected basaltic-andesites and andesites from http://earthref.org/GERM/index.html.
from the XLFRAC models, which are probably unrealistic considering the small variations in SiO2 and Mg# values (figure 3.17a).
Figure 3.17c shows models 4-7 which produce slightly better solutions, but with variable predicted HFSE concentrations. Again, there may be a number of reasons for this. In particular, kd values for HFSE can vary dramatically depending on residual phases such as amphibole, in which HFSE may be more compatible. However, this problem can be applied to all of the models, most of which show variable HFSE. This is highlighted in figure 3.17e (8 and 9), which shows relatively good model solutions for all elements other than Nb and Ta.
Figure 3.17d shows the trace element modelling between lava Sumb81 and the pyroclastic deposits. This shows particularly poor solutions for Sr, which is higher in the samples than calculated by the models; and HFSE which are significantly lower than predicted by the models. Similar variations in Sr and HFSE are noted between pyroclastic deposits and lavas in fig. 3.11. Ba is also low in these rocks, which suggests that the addition (or accumulation) of plagioclase may be responsible. Figure 3.17f (12) shows a similar profile between the high SiO2 pyroclastic sample (parent composition) and low SiO2 pyroclastic sample; albeit mostly
above the 1 line (compare profiles of 3.17d (10 and 11) with 3.17f (12)).
Major and Trace Element Summary between groups
A combination of observational evidence, geochemistry and modelling suggest that differentiation is driving the Pyroclastic Deposits towards lower SiO2 with higher 87Sr/86Sr contents, accumulating plagioclase and clinopyroxene.
Fractional crystallisation cannot account for the difference between the lavas and pyroclastic samples, or within group variations. Sr and to a lesser extent Ba are higher in these samples than would be expected from simple fractional crystallisation, and HFSE‟s are lower.
Lavas produce slightly better modelling solutions, although fractional crystallisation is unable to account for variations between the CBL and other lava samples. Some trace element model profiles resemble those for the pyroclastic rocks.
3.5.1.2. Differentiation of magmas 2: Contamination of the Arc Crust
Major and trace element modelling shows that additions, as well as removal, of mineral phases is important at Sumbing; and that simple closed-system evolution produced by fractional crystallisation does not fully account for the geochemistry of samples. While this type of semi-quantitative modelling provides some important information about magma differentiation, it does not distinguish between fractional crystallisation, magma mixing and assimilation or assimilation and fractional crystallisation (AFC). Therefore, other types of differentiation must be examined.
It was noted in section 3.5.1, that simple binary mixing should produce linear relationships between elements and hyperbolic arrays between elements and ratios which are not obvious in any of the Sumbing groups. Therefore, this section will address crustal contamination. For simplicity, the lavas will be discussed as one group because they contain very similar isotopic ratios compared to the pyroclastic rocks.
Continental xenoliths have not been discovered at Sumbing to provide direct evidence of the lithologies through which the magmas may interact at lower and mid-crustal levels. However, recently erupted calc-silicate xenoliths at Merapi provide considerable evidence for interaction between the magmas and local carbonate-rich upper crust (Chadwick et al., 2007; Deegan et al., 2011; Troll et al., in press). The xenoliths contain skarn-type mineralogy of wollastonite and diopside with vesicular reaction rims from interaction with the basaltic andesite (Deegan et al., 2011). Similar magma-carbonate interactions have also been reported from Vesuvius and the Alban Hills, Italy (Del Moro et al., 2001; Gilg et al., 2001; Dallai et al., 2004), and Popocatepetl, Mexico (Goff et al., 2001; Schaaf et al., 2005), and is often accompanied by high CO2 fluxes and explosive eruptions similar to those at Merapi (Deegan
et al., 2010; Troll et al., in press).
The upper crust beneath Merapi is reported to be a 2 km thick succession of limestone, marl and volcaniclastic units which may extend to depths of up to 11 km in the Kendeng basin, (van Bemmelen, 1949; Smyth, 2005; Deegan et al., 2010). This is likely to be similar for Sumbing, which is located about 50 km west of Merapi and bordered to the south by carbonates, marls and volcanic remnants from the West Progo Beds. So the magmas have, at least the potential to interact with similar rocks at shallow levels.
Experimental studies suggest that basaltic magma which incorporates sedimentary carbonates will result in silica-depleted (and under-saturated) residual melt, as a consequence of favoured Ca-Al rich silicic phases such as Ca-rich plagioclase and clinopyroxene, at the expense of other mineral phases. Increasing carbonate interaction will also increase the degree of crystallisation in a magma and generate CO2-rich fluid phases in the reactions:
CaCO3 + MgOmelt + 2SiO2melt → CaMgSi2O6cpx + CO2fluid
2CaCO3 + 3SiO2melt + Mg2SiO4olivine → 2CaMgSi2O6cpx + 2CO2fluid
This provides one possible explanation for the XLFRAC modelling results and mineralogy of pyroclastic deposits, particularly the idea of magmas becoming progressively silica-depleted and the highly crystalline, plagioclase and clinopyroxene-rich, nature of the rocks. It also poses an interesting question about the volcanic nature of the deposits, as they are all composed of pyroclastic (i.e. more explosive) samples, while the lavas represent more effusive lava flows?
Although considerable study at Merapi provides much better temporal magmatic constraints (Hammer et al., 2000; Gertisser & Keller, 2003), the similarity in chemistry between Merapi and Sumbing, and their close proximity and trench position, suggest that similar processes may be involved in their petrogenesis. Unfortunately no age constraints can be placed on the Sumbing magmas, so the only distinction which can be made is by petrology and chemistry.
To test the hypothesis that Sumbing magmas have undergone assimilation by carbonate- bearing rocks, two approaches are taken here:
1) Silica saturation modelling, using CIPW normative calculations, and 2) Trace element modelling, using Sr/HFSE ratios.
Silica saturation and CIPW normative calculations
To test the idea of subduction-related magmas becoming progressively silica undersaturated, CIPW normative contents of quartz, hypersthene and diopside are shown in Table 3B, together with selected major and trace elements, trace element ratios and isotope ratios for the lavas and pyroclastic deposits.
Normative minerals are hypothetical end-member solutions, where silica is allocated relative to the oxides in the rock. These calculations can be utilised to examine the degree of silica saturation by reassigning silica into the end-member minerals (Best, 2003). They also enable a user to differentiate between magmas of varying alkali contents for a particular value of SiO2 (wt.%). For example, using silica as a standard index for differentiation, the reader is
unable to differentiate between basalt and a phonotephrite, even though the latter is a more silica undersaturated rock.
Normative minerals for Sumbing are calculated assuming 20% of Fe as Fe+3, a conservative estimate for the Fe-oxidation state of subduction zone magmas (Kelly & Cottrell, 2009). All samples for Sumbing are plotted on a Ne-Ol-Di-Hy-Qz tetrohedron, together with a field for Merapi, in figure 3.18a. The two planes, between Di and Hy and Di and Ol, represent divisions for silica saturated rocks and silica undersaturated rocks repectively. The lavas show a strongly silica saturated character, with larger proportions of normative quartz and hypersthene. In contrast, the pyroclastic deposits move into a field for Merapi, and appears to becoming progressively less silica-saturated (i.e. away from quartz-hypersthene and towards diopside-hypersthene). These samples move into a field for Merapi, which are mostly within the triad for silica saturated rocks.
For changes in silica saturation at Sumbing to be attributed to interaction with carbonate at crustal levels, the variations must be controlled by differentiation. For example, suites of highly silica undersaturated and silica saturated lavas at Muriah are not thought to be a function of fractionation, rather mixing between parental magmas of alkaline and silica- saturated affinities, and/or variable degrees of partial melting (Edwards et al., 1991). To address this issue, figure 3.18b shows SiO2, Sr/Nb and 87Sr/87Sr against the sum of normative
diopside (Di) and hypersthene (Hy) and normative quartz (Qz).
If silica saturation is controlled by differentiation, the concentrations and ratios should change with progressive increases or decreases in the normative minerals. Sr/Nb was choosen because firstly, calculated trace element models show that the pyroclastic magmas contain elevated Sr and low HFSE‟s; and secondly, that carbonates (and calc-silicates) contain high concentrations of Sr and very little Nb (Chadwick et al., 2007). Therefore, Sr/Nb ratios of carbonate-rich sediments are higher than silicic or volcanogenic sediments (Plank & Ludden, 1992; Vroon et al., 1995; Gasparon & Varne, 1998).