Metallogenesis at Santorini a Subduction-zone Related Process. I: Deposition of Hydrothermal Sediments

The sediments consist of lava detritus, pumice, opaline silica, siderite, pyrite and a gel-like fraction of poorly-defined iron hydroxides and x-ray amorphous iron sulphides. The deposits can be modelled as simple mixtures of pumice and hydrothermal spring contributions. The flux rates of matter from the hot springs indicate that most metals do not settle in the bays, but are removed; Si, Ba and Mn, in particular, are selectively lost to the open caldera or the Aegean Sea. These losses are probably most significant during stormy periods.

The size of the spring deposits and the contents of Fe are much higher than in continental hot-spring deposits, but resemble exhalative sediments at other young calc-alkaline island arcs. The results support the hypothesis that hot springs associated with young subduction zones differ from hot springs on the continents or on older island arcs with more silicic rock components and in which iron-rich exhalative sediments are missing.

INTRODUCTION

Many submarine hot springs at spreading centres in the ocean are major sources for Fe, Mn, Ba, Cu, etc., and have produced extensive metalliferous deposits (Boström 1980; Rona et al. 1983; Rona 1988). Most hot springs on major land areas, on the other hand, generally seem to be metal-poor, as springs in Yellowstone National Park, Iceland and New Zealand show, and large metal-rich exhalative-sedimentary deposits are conspicuously absent there (Allen and Day 1935; Barth 1950; Waring 1965; White et al. 1988). However, hydrothermal systems near subduction zones such as at Santorini and other volcanic centres on typical island arcs (Zelenov 1972) show distinct metallogenetic tendencies, indicated by Fe-rich solutions and muds, although they are not as rich in trace metals as at springs in the deep sea.

The reason for this distribution of hot-spring types is poorly understood. More light could be shed on this problem by studies of the easily accessible hot-spring system at Santorini, located in conjunction with the Hellenic subduction zone.

In this study we present additional sediment data for trace elements and immobile components like Al, Ti and Zr, which are needed for a better genetic modelling of the deposits. In the mineralogical studies we mainly address those aspects that explain the distribution of trace elements, whereas possible equilibria between the sedimentary constituents are not discussed.

GENERAL FIELD RELATIONS

Santorini is located on a major arc-shaped swell in the Aegean Sea, stretching from Methana and Aigina over Milos and Santorini to Nisyros. All these islands show volcanic rocks (Pichler and Kussmaul 1972) and some have hot-spring deposits of varying ages. These features, and the presence of earthquake data, suggest that this arc is associated with a northward-sloping subduction zone (Berckhemer 1978; Meulenkamp et al. 1988; Spakman et al. 1988).

Santorini (Thera) represents a fractured caldera (Fig. 1), the last stage of which was created around 1600-1400 BC. Later volcanism in the middle of the caldera formed the islands Palaea, Mikra and Nea Kameni, which first appeared in AD 197, 1573, and 1711 respectively (Fouqué 1879). Mikra and Nea Kameni later merged and are presently the centre of present volcanic and fumarolic activity; eruptions and the out-pouring of new lavas now take place irregularly a few times per century. Hydrothermal activities and the deposition of iron-silica muds in embayments and straits have long been known at these islands (Fouqué 1879; Brun 1911). The best-developed sedimentation occurs primarily in two bays, one on Nea Kameni (Butuzova 1969, 1978) and the other on Palaea Kameni (Puchelt 1973), shown as A and P on Fig. 1. Field studies show that shores near very active springs have intense ferric hydroxide staining (A, B, C, L, O, P, Q, R and T in Fig. 1), while weak hydrothermal activity results in a faint staining (Fig. 1, bays D, E, F, G, I, J and S); some bays lack staining (K, M). Furthermore, very active springs produce distinct colour changes in the sea water from the normal deep-blue colour of the open caldera to a pale greenish shade; fronts between normal sea water and debouching spring-affected waters are generally easy to spot. In the innermost part of some bays waters may even turn yellowish-brown. Further details of the hot-spring waters and their chemical compositions are given in Boström et al. 1990a.

HYDROTHERMAL SEDIMENTS; DISTRIBUTION, SAMPLING AND DESCRIPTION

Pronounced hydrothermal sedimentation is primarily developed in the A and P bays, whereas most other bays that are well exposed to the swell and surf from open-caldera waves appear almost barren of sediments, as coring attempts, diving, and inspection with glass boxes revealed. Sediment sampling took place in 1984 and 1986, using manually-pushed plastic tubes (diameter 10 cm, length 50 cm); thanks to this large diameter, the sediments were only slightly disturbed. A few samples were obtained in deeper water, using the plastic tubes as barrels in a gravity corer. The distribution of samples in the Nea Kameni (NK) bay is shown in Fig. 2. The Palaea Kameni (PK) core was procured half-way between the little chapel and the innermost part of the bay, see Fig. 4 in Puchelt 1973. Upon recovery, the sediments were capped and carried upright as hand luggage to the laboratory to prevent transport damage. To complete the geochemical results in this study we also incorporate data for NK 5, 6 and 8, which were obtained earlier (Boström and Widenfalk 1984).

In the laboratory the cores were cut in half and placed in an inert atmosphere glove compartment for sub-sampling. The sediments NK 2, 4, 6 and 7 were very watery and, particularly in the top layers, gel-like in appearance, while cores NK 1 and 5 were fairly sandy and rich in large pumice fragments, some up to 2 cm in diameter. The remaining cores fall between these extremes, appearing to be mixtures of silty pumice material and a gel-like phase.

The colours of the sediments ranged from tan to reddish-brown in the oxidized tops to purplish-white to black, or brownish-greenish-black, in the reduced bottom parts.

MINERALOGICAL RELATIONS

Studies on 22 samples from cores NK 1, 2, 3, 4, 7, 9, 10 and the PK core confirm the mineralogical results obtained by Schroll (1978). The x-ray diffraction studies reveal that 16 samples contain siderite, but the distribution is uneven. Thus, the x-ray data and carbonate determinations in a Poisson apparatus show that cores PK and NK 3 contain 25-40% siderite, whereas other cores have less than 7%. The d-spacings for this siderite are larger than reported in JCPDS (1974), the d-values for (104) and (113) being 2.81 and 2.-15 respectively, instead of 2.79 and 2.13. Analyses of separated crystals showed the composition to be (Fe_0.9.Ca_0.08Mg_0.02Mn_0.001) CO₃. The content of large Ca ions (8 mole %) could explain the d-spacing shift if this is a linear function between the d-spacings of pure calcite and siderite. The MnO content corresponds to 0.09% which is significantly higher than for any bulk sediment; see further below.

The lower parts of the cores showed small x-ray diffraction peaks for pyrite, and pyrite framboids less than 0.02 mm in diameter were identified under the microscope after the samples had been treated with hot HCI and hydroxylamine-chloride, core PK showing the highest concentrations. Panning and magnetic separation of such samples removed pumice fragments and magnetite crystals, after which pyrite could be chemically analysed, revealing a content of 10 ppm Co, 49 ppm Cu and 42 ppm Zn. The low concentrations of pyrite and magnetite in all cores suggests that most iron is present as x-ray amorphous iron-oxyhydroxides, but in some cases goethite was identified. Vivianite was not found. Microscopy revealed the presence of much opaline silica as diatom frustules.

GEOCHEMICAL STUDIES: METHODS AND RESULTS

Chemical analyses of 102 sediments were done on materials dried at 95°C, using a sequentially reading ARL 3520 atomic emission spectrometer with inductively coupled plasma for excitation (ICP-AES); the procedures and the quality of the results are discussed in Burman et al. (1978, 1979) and Boström and Widenfalk (1984). Studies of the precision and accuracy in an interlaboratory study confirm the reliability of the methods (B. Boström 1987). Furthermore, analyses of seventeen Santorini caldera samples by our laboratory and by SGAB analytical laboratory in Luleå showed excellent agreements in the data; not the least for the sum of oxides and for refractory components like TiO₂ and Al₂O₃ (Boström et al. 1990c). To remove data-scatter due to variations in loss of ignition, all analyses were recalculated to an oxide-sum of 100%.

The chemical compositions (see Table 1), particularly the contents of iron and aluminum, reveal that the sediments form well-defined groups. This grouping agrees with the petrography of the samples and with the field observation that sediments are coarse at the shore, except where the most active springs debouch south of coring-site 2 (see Fig. 2).

Plots of the data from Table 1 reveal that concentrations of P₂O₅, Fe₂O₃ and Na₂O are highest in the most extreme hydrothermal sediments (HS-1, Fig. 3a), whereas the PK sediments and the pumice-rich cores 1 and 5 show much lower values. A similar trend (Fig. 3b) exists for V and Ni, except that the largest values are found in the less extreme hydrothermal sediments HS-2. However, much Na₂O may also be added by admixed sea water. 

Opposite trends are shown by MnO, CaO and Zn, which show low values in HS-1 and HS-2, and the highest in core 5 (Fig. 3b). The components SiO₂, TiO₂, Al₂O₃, Zr, and Ba (Fig. 3d), finally, have a concentration pattern opposite to that of phosphorus, iron and sodium.

Ternary compositional plots (Fig. 4) confirm the trends observed in Fig. 3. The sediment compositions fall along trend lines, with the extreme hydrothermal sediments (cores 2, 4, 6 and 7) forming one end-component and sediment 1 the other. The patterns in Fig. 3 and 4 confirm that the sediments largely represent mixtures of pumice and hydrothermal matter, as shown in plots by Boström and Widenfalk (1984). To corroborate this hypothesis we modelled the sediments as mixtures of pumice and an extreme hydrothermal deposit, represented by the mean of the NK 4 and NK 6 sediments. The optimal models are shown in Table 2. A 100% hydrothermal input explains sediments NK 2, 4, 6 and 7, whereas an input of 70% pumice is needed to explain NK 1; NK 5 and PK each require a pumice input of 20%. The fit of the models is demonstrated by Fig. 5. In the graphs, full lines represent the anticipated one-to-one correspondence between real and model sediments, and the broken lines show the deviations of a factor of two (i.e. between +100% and -50%).

The data in Table 2 and in Fig. 2-5 show that hydrothermal matter makes up a larger fraction of the sediments in the inner part of bay A, than elsewhere, and that the NK sediments become progressively richer in pumice towards the open strait between Palaea and Nea Kameni. It should be recalled, however, that the exact hydrothermal end-member is hard to define from the analyses of NK 4, 6, since small traces of pumice also occur in these sediments.

These results are suggested also by a comparison of the extreme hydrothermal sediments (NK 4, 6) with the hydrothermal spring waters (HSW, Fig. 6; Boström et al. 1990a, 1990b). The graphs in Fig. 6 suggest, however, that only 3-10% of all spring-delivered silica, barium and manganese form deposits near the springs; these elements are brought from here to the surrounding caldera or further out into the Aegean Sea.

The data thus show that most P and Fe are present in a hydrothermal product, whereas Al, Ti and Zr primarily are lithogenic, most deriving from pumice. However, since the mixing components have different compositions, even a small admixture of one component may account for most of a given element. Thus, the iron content in the hydrothermal component is much higher than in the lithogenic component and therefore even a 6% admixture of hydrothermal matter may explain 50% of the iron in a given mixture (Fig. 7), whereas the hydrothermal phase must exceed 97% to explain more than 50% of all Ti. Other components derive equally from both sources, e.g. Cu, V, Y, Mg and Zn.

DISCUSSION

The metalliferous processes at Santorini described here corroborate the results of Butuzova (1969, 1978), Puchelt (1973), Puchelt et al. (1973), Smith and Cronan (1978, 1983) and Boström and Widenfalk (1984). The geological evolution of these bays and the plate-tectonic setting of these metalliferous springs, however, has been little discussed.

     Evolution of the bay sedimentation:

The origin of bay A dates back to 1866,when two sub-parallel flows of the Georgios lavas formed the shorelines; lava-flows from later eruptions have hardly affected the depths in the bay (Fouqué 1879; Huijsmans 1984). Fouqué reported the depths in the innermost 60 metres of the bay to be about 2-4 m; these depths have hardly changed, since present depths there range mainly between 1.5 and 3.8 m. This suggests a very slow accumulation of sediments, but it appears likely that sedimentation is fast.

Studies of the springs on the south shore of the innermost basin in bay A show that the likely debouching-rates of the waters are 3-12 m per minute; a geometric mean of these estimates and the size of the spring openings indicate an outflow of 400L per minute. It is also likely that other weaker springs in the bay account for some 200L per minute. These estimates and the compositions of the spring waters (Table 3) show that the springs deliver about 8000 kg Si, 4000 kg Fe, 400 kg Mn and 16 kg Ba annually. The exhalative-rich sediments, furthermore, have an Fe-content of about 50%, that is, 4000 kg Fe would correspond to 8000 kg deposit, if all the iron precipitates. Watery gelatinous sediments generally have very low dry-matter densities; assuming a realistic value of 0.5 g per cubic centimetre, one finds that an 8-ton deposit takes up a volume of 16 x 10⁶ cm³.This sediment volume would be equivalent to a sediment build-up of a little more than 3 cm per year, since the inner basin has a deep area of about 500 m². Furthermore, taking into account some pumice admixture one could expect the sediment build-up to be some 4-5 cm per year, or about 5 metres since Fouqué (1879) described the bay.

These hydrothermal supplies are very approximate, but could well be too low, since past hydrothermal activity was probably more intense than at present. The most likely conclusion, therefore, is that hydrothermal sediments are lost from the bay. This conclusion is supported by the fact that pumice lines occur on the shore boulders some 70 cm above the water surface, suggesting that during stormy periods the waters in the bay are thoroughly reworked. Such a process could remove the sediments deposited during calmer periods. This explains some of the accumulated hydrothermal phases in the caldera sediments (Boström et al. 1990c).

     Plate tectonic control of hot-spring distributions:

Geothermal areas with hot-springs occur in three major geotectonic environments, namely at deep-sea spreading centres, at island arcs and in continental areas, primarily in major caldera and fracture systems. Hydrothermal springs at deep-sea spreading centres have been discussed in Boström (1980), Rona et al. (1983) and Rona (1988). The metalliferous nature of such springs is hardly contested any more, and the origin of the metal-enriched solutions is now partly understood (Rosenbauer and Bischoff 1983).

It appears less widely known that many springs at calc-alkaline island arcs are metalliferous, as has been observed in the Kuriles (Gorshkov 1958; Naboko 1963; Fyfe et al. 1978; Zelenov 1964), Tjiater and Banu Wuhu, Indonesia (van Bemmelen 1949; Zelenov 1964, 1972), Matupi Harbour (Ferguson and Lambert 1972), Deception Island (Elderfield 1972), and Kick 'em Jenny off Grenada (Lesser Antilles). These springs produce Fe-rich muds, but not as trace-metal rich as those in the deep sea. Similar conclusions were presented by Zelenov (1972), based on a study of hot-spring related metalliferous processes at well-developed young island arcs in eastern and south-eastern Asia. However, even large island arc deposits, such as those at Ebeko in the Kuriles (Gorshkov 1958; Naboko 1963; Zelenov 1972), are smaller than the biggest deep-sea deposits.

The chemical composition of formed deposits, e.g. on Java and the Kuriles (Zelenov 1972), show strong resemblances with the sediment data presented in Table 1. Likewise, data in the literature for spring waters from these island-arc occurrences show a distinct resemblance with the hot-spring waters at Nea Kameni and those in a drill hole on Palaea Kameni (Boström et al. 1990a, 1990b).

In contrast with the island-arc related hot-spring deposits, metalliferous springs are strikingly rare in continental areas, such as the USA, or in older island arcs, where silicic volcanism and plutonism is more widespread, e.g. in Japan and in New Zealand. Allen and Day (1935, 357-359) explicitly discuss the paucity of metalliferous deposits in the Yellowstone area. The exceptions, the Chocolate Pots, are considerably smaller in volume than those in the Palaea Kameni bay alone (Puchelt 1973); the total volume of hydrothermal matter discharged at the Kameni islands, which partly covers also the caldera floor, is several order of magnitudes larger (Petersen and Müller 1978; Boström et al. 1990c). Field studies carried out at several springs in the Yellowstone National Park by one of us (KB) on three occasions during 1964 and 1976 revealed no exhalative-sedimentary deposits, although they were looked for. Even many deposits with a spectacular appearance, e.g. at West Thumb, have proved to be poor in iron (Allen and Day 1935). These conclusions agree with later data for the Yellowstone hot-spring waters, which tend to show iron contents lower than those reported in Allen and Day (1935) and also suggest that the ratio Al/Fe in many springs exceeds 1.0 (Rowe et al. 1973; White et al. 1988). These results are surprising in view of the total size of the Yellowstone hydrothermal system, which is probably the biggest land system in the world, with a very large total heat-flow (White et al. 1988).

Likewise, extensive studies of most New Zealand springs show a distinct paucity of iron and aluminum (Weissberg et al. 1979). Furthermore, the monograph by Waring (1965) lists no major metalliferous deposits in continental settings.

The most intense hydrothermal processes occur at spreading centres on the ocean floor, and strong metal-producing processes at sub-aerial volcanoes seem to be associated with young calc-alkaline island arcs, which are all located at subduction zones suggesting the following general pattern:

A special classification problem is posed by springs at major continental fault-zones, and on large mid-oceanic islands, like Iceland (Stefansson 1983). However, in spite of the fact that some metal-rich brines have been found in such settings, especially at Salton Sea (Muffler and White 1969), no metalliferous surface deposits have resulted from such springs, suggesting a minor flux of elements to the earth's surface.

The reason for this plate-tectonically controlled subdivision of hot-spring occurrences is poorly understood, but is probably caused by several factors. The supply rate of water may appear to be the single best explanation, but this is probably not the case. Thus, the hot springs at Reykjanes (Barth 1950; Kristmannsdottir 1983) are fed by sea water, but produce no exhalative deposits. Sea water under very high pressure may, however, be pressured into close contact with hot magmatic zones, which could explain the metal-rich nature of deep-sea springs, in contrast with more shallow ones (Boström 1980; Stefansson 1983).

The mechanism does not explain the differences between springs in calc-alkaline settings and those in more sialic crusts, nor are these differences easily explained as simple functions of pH, temperature or carbon dioxide pressure. Thus, all acid spring waters in Yellowstone, for instance, contain less iron than the moderately acid waters in Santorini or Banu Wuhu (Boström et al. 1990a). Furthermore, it is obvious that carbon dioxide alone can hardly explain the preferential release of iron in the hydrothermal system at Santorini, although alternative views have been presented by Harder (1964). Thus, large emissions of carbon dioxide, such as at Methana (Greece), Pamukale (Turkey) and Mammoth Springs (Yellowstone National Park), have not resulted in noticeable metalliferous deposits (Waring 1965). The same conclusion holds for the Reykjanes springs (Kristmannsdottir 1983). Indeed, it is well known that many volcanic emanations are rich in carbon dioxide, yet are distinctly metal-poor (White and Waring 1963). Furthermore, the metal-rich solutions on Santorini have much lower temperatures than the metal-poor solutions in Yellowstone or in Iceland (White et al. 1988; Kristmannsdottir 1983).

This enigma as to the origin of the metal-rich hot-spring waters is still unsolved (White et al. 1988). We do not claim to have a definite solution, but the results above suggest that the subduction process may be related to the genesis of metal-rich solutions, which probably form at very large depths and pressures and also have ample time to overcome reaction-rate constraining relations (Burnham 1979; Sillitoe 1972; Boström 1981).

A more thorough thermodynamic or kinetic model may therefore required to explain the predominance of Fe and Si in the Kameni spring waters, in contrast with their low contents of Al, Mn and many trace elements (Boström et al. 1990a). This problem is not unique, however; alteration studies of basaltic matter form artificial hydrothermal solutions that are significantly poorer in trace elements (e.g. Ba and Zn) than is characteristic in nature (Rosenbauer and Bischoff 1983). Another problem is the variability of Mn content in island-arc related deposits. Thus, the deposits in the Kameni bays and the Santorini caldera are poor in Mn compared to other caldera hot-spring deposits on Milos, located on the very same arc-system as Santorini; these problems are discussed elsewhere (Galanopoulos et al. 1989).

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

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