Evolution of the Volcanic Rocks of Santorini

The analytical data for basement phyllites cancel those rocks out as source material for the volcanic calcalkaline series. It is assumed that the magmas were generated in that part of the mantle which lies above the subducting plate by partial melting induced by an aqueous phase, enriched in incompatible elements, which was generated form the subducted sea floor.

Papers on the volcanic rocks of Santorini prove that they were formed by three eruptive cycles, each starting from high alumina basalt and finishing with dacite or rhyodacite respectively (Pichler & Kussmaul, 1972; Nicholls 1971, 1978; Pichler et al 1972, 1978). The phase of latest volcanic activity since about 197 B.C., which is visible only in the islands of Palaea and Nea Kameni, provides access to dacitic lavas only, although more basic xenoliths are found occasionally as inclusions.

Since all volcanic cycles followed similar lines it seemed justified to study the chemical development of all volcanic rocks of the Santorini group together in order to elucidate the mode of magma development.  

The sample locations are given on figures 1 and 2. The rocks have been analysed by wet chemical methods (major elements), XRF (Sr, Ba, Ni, Cr, Zr), ionsensitive electrode (F) and instrumental neutron activation - INAA -(Rb, Cs, Co, Cr, Sc, La, Ce, Sm, Eu, Tb, Dy, Yb, Lu, Hf, Ta, Th, U). The parameters of the INAA are given in Puchelt & Kramar 1978 b. The results are listed in Tables 1 - 5.

In order to visualize the interelement correlation, Harker diagrams have been drawn for the major elements (fig. 3). Grouping of the rocks was done according to the scheme of Taylor (1969) (cf. fig. 4), and the generalized trend of magmatic development can be seen from the A F M diagram (fig. 5). Further information can be obtained from correlation plots for trace elements vs. silica (fig. 6 a+b) and the plots of chondrite -normalized REE concentrations of six characteristic samples (fig. 7).

The correlation diagrams show a steady decrease of Al₂O₃, CaO, and MgO with increasing silica. A steady increase is observed in the same direction for Na₂O and K₂O, while total iron (as Fe₂O₃) starts with a horizontal or slightly increasing trend and decreases above about 60 % of SiO₂. This very trend is more pronouned for titania with the maximum value in the same region of SiO₂.

With trace elements a steady decrease of different slope is observed for Co, Sc and Sr, while the data for Ni and Cr show a completely different picture: high but scattering values are observed in the range <60 % SiO₂, but consistently low values are found for SiO₂ concentrations above 60 % SiO₂.

The trace elements, which increase with silica, still have varying enrichment factors (averaged concentration at highest SiO₂ divided by averaged concentration at lowest SiO₂) :

Rb~7; La~6; Lu~3; Zr~7; Hf~4; Th~8; Ba~5;

(Co~0.1; Sc,~0.2; Sr~0.5)

Chondrite normalized (Puchelt & Emermann 1977) REE data sets for characteristic lavas (and one Christiani pumice) show the development from low absolute values with an almost horizontal distribution (OI 70/3) to patterns which are strongly enriched in light rare earths (LREE) without (OI 70/4) or with (Christ. pumice) a negative Eu anomaly (fig. 7).

DISCUSSION:

The consistency of the various element data in the SiO₂ plots indicate a rather coherently developing series of the respective magmas. Starting from an initial high aluminum basalt, dacitic compositions can be reached by differential crystallisation of olivine (ol) orthopyroxene (opx), clinopyroxene (opx) titaniferous magnetite (mt) and plagioclase (plg). Phenocrysts of these mineral phases have been found and microprobed by Nicholls (1971).

If the concentrations of the elements in a crystallizing melt are governed by the degree of incorporation (distribution coefficient: concentration in crystal/concentration in the remaining melt) into a mineral phase, an increase with increasing SiO₂ points to D values below unity and vice versa. Distribution coefficients for ol, cpx, opx, mt and plg have been determined by Puchelt (1976) and Puchelt & Kramar (1978a) for calcalkaline rocks from Santorini and their phenocrysts.

Table 6 gives a rough survey of those D values with the range of changes according to the varying chemistry of melt (andesite to rhyolite) and phenocryst.

Distribution coefficients for Santorini pumice only were reported by Schock (1977). From Table 6 it can be seen that high D's are prevailing for Co. and Cr in olivine. The same is true for Co, Sc, and Cr, in opx and cpx. While olivine incorporates REE's only to a very small degree, cpx incorporates the REE's differently but mostly below unity: LREE's have lower D's but heavy REE's. Thus crystallizing of ol, opx and cpx decreases the content of Ni, Co, Cr, Sc in the melt and increases the concentration of the other elements. If the oxidation conditions in a magma are such that titanomagnetite can form, this phase may serve as a new scavenger for Ti (cf the change of slope in the TiO₂ SiO₂ plot in Fig. 3), Co, Ni, Cr and Sc (cf. Leeman et al. 1978). Magnetite may also extract Zr, Hf, Ta, Th and U under favourable compositions of the melt.

The development of the lavas on Santorini can be explained by a model of crystal differentiation using the indicated distribution coefficients, but it has to be considered that the D's of each mineral vary with its composition and with that of the parent magma (Puchelt & Kramar 1978a).

Crystallisation and separation of plg increases the REE concentration in the melt but LREE's are concentrated to a lesser degree in the melt. Separation of plg under suitable conditions of pO₂ leads to an enrichment of Eu in the feldspar and a related negative Eu anomaly in the melt.

Calcutation for the amount of crystallization from an initial magma of high-Al-basalt composition to a thyodacite using the above-mentioned D values give about 80 solidification and separation, which includes approximately 6 - 9% olivine, 30 - 35% pyroxene, 32 - 34% plagioclase, and 3 - 5% titanomagnetite.

Different source rocks have been discussed for the calcalkaline volcanics of Santorini, but only those assumptions which allow for the presence of high Al basalt (Nicholls 1978) deserve further consideration, since a differentiation sequence for all volcanics of the Santorini group must be assumed. While high Al-basalts cannot be generated from oceanic tholeiites or their metamorphic equivalents by partial melting of the subducted seafloor, they can be formed by partial melting of mantle material under hydrous conditions and proper pressure (Kushiro 1974; Nicholls 1974). The necessary water must be provided by reactions of metamorphosis in the downgoing slab of the underthrusted plate.

Since the degree of partial mantle melting which is calculated by Nicholls (1978) does not provide the necessary concentrations of Rb, Cs, Zr, Hf, Th, U it has to be assumed that these elements are extracted by the hydrous solutions from the subducted tholeiite-eclogite plate and/or additional part of the mantle and introduced into the newly generated magmas. These high Al-basalt magmas start to differentiate in higher levels and thus form the different observed calcalkaline rocks.

A connection of the phyllitic basement and the generation of the calcalkaline magma can be ruled out, as the few phyllite analyses of Table 7 and investigations of Davis and Bastas (1978) demonstrate.

-------------------------------------------

---------------------------------------