Journal of Volcanology and Geothermal Research

Transcription

1 Journal of Volcanology and Geothermal Research 173 (2008) Contents lists available at ScienceDirect Journal of Volcanology and Geothermal Research journal homepage: Chemical transport in geothermal systems in Iceland Evidence from hydrothermal alteration Hjalti Franzson a,, Robert Zierenberg b, Peter Schiffman b a Iceland GeoSurvey, 9 Grensásvegur, 108 Reykjavík, Iceland b Department of Geology, University of California, Davis, One Shields Avenue, 95616, CA, USA article info abstract Article history: Received 1 July 2007 Accepted 28 January 2008 Available online 4 March 2008 Keywords: Iceland basalt geothermal systems hydrothermal alteration chemical transport isocon method This study focuses on the chemical changes in basaltic rocks in fossil low- and high-temperature hydrothermal systems in Iceland. The method used takes into account the amount of dilution caused by vesicle and vein fillings in the rocks. The amount of dilution allows a calculation of the primary concentration of the immobile element Zr, and by multiplying the composition of the altered rock by the ratio of Zr (protolith)/zr (altered rock) one can compute the mass addition caused by the dilution of the void fillings, and also make a direct comparison with the likely protoliths from the same areas. The samples were divided into three groups; two from Tertiary fossil high-temperature systems (Hafnarfjall, Geitafell), and the third group from a low temperature, zeolite-altered plateau basalt succession. The results show that hydrothermally altered rocks are enriched in Si, Al, Fe, Mg and Mn, and that Na, K and Ca are mobile but show either depletion or enrichment. The elements that are immobile include Zr, Y, Nb and probably Ti. The two high-temperature systems show quite similar chemical alteration trends, an observation which may apply to Icelandic fresh water high-temperature systems in general. The geochemical data show that the major changes in the altered rocks from Icelandic geothermal systems may be attributed to addition of elements during deposition of porefilling alteration minerals. A comparison with seawater-dominated basalt-hosted hydrothermal systems shows much greater mass flux within the seawater systems, even though both systems have similar alteration assemblages. The secondary mineral assemblages seem to be controlled predominantly by the thermal stability of the alteration phases and secondarily by the composition of the hydrothermal fluids Elsevier B.V. All rights reserved. 1. Introduction The location of Iceland as a subaerial part of the Mid-Atlantic Ridge, with a wealth of geothermal activity, makes it a unique area for the study of water rock interaction. Temperatures of fluid rock interaction range from about 2 to 3 C in the groundwater systems to probable supercritical values in the high-temperature geothermal areas (Elders and Fridleifsson, 2005). The rocks that host the geothermal reservoirs are of igneous origin, with basaltic compositions constituting about 90% of the volume with the remainder having more evolved compositions. The abundant geothermal resources are economically important and widely utilized, and have consequently been extensively studied from geochemical, structural, geophysical and geological points of view (e.g., Arnorsson et al., 1983; Bodvarsson, 1983; Marty et al., 1991; Schiffman and Fridleifsson, 1991; Lonker et al., 1993; Riedel et al., 2001). The geothermal resources range from low-temperature systems, in which shallow groundwater has gained heat in response to the prevalent regional geothermal gradient, to high-temperature Corresponding author. Tel.: ; fax: addresses: hf@isor.is (H. Franzson). systems in which the high thermal gradient is due to shallow crustal magmatic activity. The latter type is mostly confined to active volcanic centres. However, fossil high temperature geothermal systems are exposed by erosion allowing three dimensional access to the subsurface portions of the hydrothermal systems (e.g. Fridleifsson, 1983, 1984). The majority of the hydrothermal systems in Iceland have waters derived from local meteoric water, although seawater-dominated hydrothermal fluids occur in some near coastal geothermal systems (Sveinbjornsdottir et al., 1986). The general similarity of both source fluids and host rocks throughout much of Iceland allows comparison among hydrothermal systems. Rock properties play an important role in the petrophysical parameters of the individual systems and an increasing emphasis has been placed on the study of reservoir characteristics. During the last 13 years, Orkustofnun (the National Energy Authority of Iceland), with the financial support of Orkuveita Reykjavíkur (Reykjavik Energy), has worked on a project with the main aim of defining the reservoir characteristics of the rocks in the geothermal systems in Iceland. While active geothermal systems have the advantage of allowing direct comparison of rock alteration to fluid composition, this comparison only relates to a single stage, i.e., the present, in the evolution of the system. In contrast, sampling of fossil geothermal fields allows /$ see front matter 2008 Elsevier B.V. All rights reserved. doi: /j.jvolgeores

2 218 H. Franzson et al. / Journal of Volcanology and Geothermal Research 173 (2008) examination of the time integrated history of rock alteration, including determination of the temporal evolution of mineralogy and porosity. The sampling strategy utilized in this study emphasized variably eroded fossil geothermal systems, focussing on areas that had been previously mapped with respect to their geological structures and hydrothermal alteration. The assumption is that rocks in these fossil systems are close analogues of those in presently-active geothermal reservoirs. Approximately 500 samples were collected for these studies, a large proportion of which have been analyzed for total and effective porosity, permeability, chemical composition, and petrographic characteristics. This extensive data set has been especially useful in determining the interrelationship between porosity and permeability of Icelandic rocks (e.g., Sigurdsson and Stefansson, 1994; Sigurdsson et al., 2000; Stefansson et al., 1997), and the relationship of these physical properties to the degree of hydrothermal alteration (e.g., Gudmundsson et al., 1995; Franzson et al., 1997, 2000; Franzson, 1999). The main emphasis of this paper is to elucidate the nature of chemical transport in Icelandic geothermal systems by correlating the bulk composition of rock samples with their petrographical characteristics. Using the petrographically determined percentage infilling in the rocks, we will show that it is possible to assess the amount of chemical dilution that has taken place. This allows the deduction of the primary concentration of the immobile elements in the altered rocks, and by correlating these with the relatively fresh volcanic rocks equivalents, we assess the chemical change that has taken place during the hydrothermal alteration. Furthermore, using the isocon method (Grant, 1986) for determining mass transport, we also evaluate the chemical enrichment-immobility-depletion of various chemical components within individual samples. 2. Geological setting This study is based on data from approximately 130 basaltic rock samples and focuses on three locations (Fig. 1) as described below: a) The Hvalfjordur basalt succession consists of ca. 4 3 myr old plateau basalts which have been subjected only to zeolitic alteration (chabazite thomsonite to mesolite scolecite, e.g., as described from Eastern Iceland by Neuhoff et al., 1999). The Hvalfjordur area has been mapped in considerable detail and the rocks are well characterized chemically (Franzson, 1979). Twenty two samples of basaltic lava flows were taken from this area. b) Hafnarfjall central volcano is a myr old, deeply eroded central volcano that has been mapped with respect to its structure, geochemical evolution, and locally, its hydrothermal alteration (Franzson, 1979). The volcanic products range from basalts to rhyolites which have been subjected to variable degrees of hydrothermal alteration. Forty five samples are included in this study, and off these twenty seven are lava flows, five very scoraceous or tuffaceous tops of lava flows and twenty intrusions (mostly fine grained dykes or sills). c) Geitafell central volcano was formed about 6 myr ago, is deeply eroded, and has been mapped in detail, both with respect to volcanic evolution and hydrothermal history (Fridleifsson, 1983). An investigation has also been completed on the geochemical evolution of the volcano (Thorlacius, 1991). Fifty five basalt samples are included in the study, where thirty are lavas, five very tuffaceous or scoraceous lava tops and twenty intrusions (mostly fine grained dykes or sills). The three study areas are all Tertiary in age, and differ from younger volcanic successions that host active geothermal systems principally by their lack of hyaloclastite formations, which occur almost exclusively in formations younger than the onset of glaciation at 3.3 myr (c.f. Fig. 1). These older hydrothermal systems were subjected to an initial phase of low-temperature alteration, followed by the main phase of high-temperature alteration, and subsequently overprinted by lowtemperature alteration formed during the demise of the central volcanic complex and succeeding erosion. It has been demonstrated by extensive petrographic observations that mineral deposition during the high temperature alteration regime in Icelandic hydrothermal systems generally reduces rock permeability and the last low-temperature alteration episode does not greatly affect the bulk rock composition. Petrography can provide important constraints on the relative timing of alteration events. For example, minor late-stage zeolite deposition, present in some of our samples, has been tentatively interpreted as due to deposition from fluids trapped in relatively impermeable rock. In general calcite does not occur within the epidote amphibole zone in presently-active high-temperature systems where temperatures exceed 290 C. However, in the fossil Fig. 1. A simplified geological map of Iceland showing the Tertiary, Plio-Pleistocene and the volcanic zones. The areas sampled for the petrophysical study are dark shaded, and the locations focussed on in this study (Hafnarfjall, Hvalfjordur and Geitafell) are also marked.

3 H. Franzson et al. / Journal of Volcanology and Geothermal Research 173 (2008) Fig. 2. The alteration zones used in Iceland, their dependence on temperature and the main alteration features of the primary basaltic rock components. systems, calcite is occasionally superimposed on epidote actinolite zone mineral assemblages, implying carbonate deposition either during cooling of the high-temperature system by inflow of colder fluids or during the succeeding low-temperature episode. Hence, changes in the chemical composition of rocks relative to their fresh counterparts are believed to have occurred up to the highest temperature stage of alteration that the rock has undergone, with possible effects due to precipitation of late-stage calcite. The samples are all either volcanic rocks which have been gradually buried by accumulation of younger volcanics, or intrusive rocks. The age of the accumulated sequence is progressively older and more altered as it is more deeply buried. However, the same cannot be inferred about the intrusive rocks, as their minimum age is generally unconstrained. This may be of importance when considering the timing of alteration, especially within the most intense alteration zones, as intrusions may occur during any stage of the alteration. Rocks intruded at a later stage of alteration may only show a minor alteration effect compared to those intruded earlier, even though they were subjected to similar temperatures and pressures. Therefore, contrasts in the degree of alteration within intrusive rocks and their surrounding extrusive host rocks should be expected and may depend on factors other than differences in permeability and porosity. 3. Petrographic data The rock samples are all of basaltic composition, ranging from olivine tholeiite to quartz normative tholeiite. They would dominantly be holocrystalline, but with some glass fraction in the most scoracious part of the lavas. The main primary components in basalts are relatively Ca-rich plagioclase, clino-pyroxene (augite) and opaques (magnetite ilmenite), and with subordinate amount of olivine, especially in the more primitive basalts. Hydrothermal alteration is essentially made up of two components: (a) replacement of primary components in the rocks by alteration minerals, and (b) precipitation of alteration minerals into voids in the rock. Hydrothermal alteration in active Icelandic geothermal systems is systematically zoned with respect to temperature (Fig. 2). The top zone, which contains relatively fresh rocks, includes a sequence that does not show any indication of Fig. 3. Range in primary porosity in Icelandic rocks deduced from petrographic study of the rock samples (n=127).

4 220 H. Franzson et al. / Journal of Volcanology and Geothermal Research 173 (2008) Fig. 4. The relation between degree of infilling and the total percentage of alteration minerals (i.e., % alteration minerals divided by the total mode) of a sample. Samples are grouped according to their alteration zone (from Franzson et al., 2000). geothermal interaction and may at the most show minor oxidation due to groundwater circulation. The temperature dependency, as seen in Fig. 2, affects both the sequence of alteration minerals that fill voids, as well as the ones that replace primary minerals. Thus olivine and glass are completely altered near the upper boundary of the mixed layer clay zone. Plagioclase and opaques are more resistant to alteration and may be only partially altered to the minerals shown on Fig. 2, especially when taking into account that some of the rock samples are low porosity intrusions. Pyroxene shows in general similar resistance to alteration as plagioclase. It is mainly seen altering into clays and then actinolite at deeper levels. Petrographic examination of the rock samples collected for this study used point counting (200 points) to quantify primary porosity, i.e. the original open space in rock prior to alteration (dominantly vesicles and minor fractures), and to assess how much of that porosity had been filled by deposition of alteration minerals. Two hundred points were counted on each rock thin section. Fig. 3 shows the distribution of primary porosity in Icelandic rocks, where porosity ranges from zero up to about 70%. As alteration proceeds void space in the rock is progressively filled with alteration minerals. Fig. 4 shows the relationship between the degree of pore filling and the extent of alteration, defined as the percentage of alteration minerals relative to primary minerals. Rocks with a primary porosity, as petrographically determined, below approximately 15% have variable degrees of infilling (Fig. 5). In contrast, rocks that initially exceeded the 15% primary porosity threshold tend to show near complete filling of the primary pore-space. This implies a non-linear relationship between permeability and porosity in the Icelandic basaltic samples. Fig. 5. The rate of deposition of alteration minerals into vesicle filling in relation to alteration zones and primary rock porosity. Note that there is apparently more rapid (i.e., efficient) deposition in rocks with N15% primary porosity, as determined petrographically (from Franzson et al., 2000).

5 H. Franzson et al. / Journal of Volcanology and Geothermal Research 173 (2008) Chemical data The chemical analyses were made by two commercial chemical laboratories, The Caleb Brett Laboratory in England and McGill University in Canada. Both used standardized XRF techniques. Values for samples analyzed by both laboratories are generally within analytical error. The samples were analyzed for major, minor, and several trace elements. Loss on ignition (LOI) was measured in all the samples, and in some samples, CO 2 and Stotal were specifically analyzed. The trace elements analyzed included Zr, Y, Zn, Cu, Rb, Sr, Nb, Ga, Ce, V, Pb, U, Th and As. All the analyses presented here have been recalculated to 100%, without LOI, to allow direct comparison with unaltered protoliths. The compositional range of relatively unaltered rocks has been acquired from other sources for comparison with the data on altered rocks presented here. Analyses describing the evolution trend within the Hafnarfjall central volcano and the Hvalfjordur plateau basalt succession are from Franzson (1979), and analyses from the Geitafell central volcano are from Thorlacius (1991). Chemical analysis from the Krafla central volcano (Karl Gronvold, unpublished data) and the Reykjanes Langjokull volcanic zone (Sveinn Jakobsson, unpublished data, 1999) have been used as reference samples for comparison with the altered samples where relatively unaltered samples are unavailable from the respective areas. Loss on ignition values range from zero up to about 13%. Fig. 6 shows the relationship between the % of remaining primary rock component (i.e. =100-(rock alteration+void filling)) in basaltic lava flows and LOI, where the samples have been grouped into alteration zones. The figure clearly shows a strong correlation between LOI and the extent of alteration in these samples. It also shows an inverse correlation between the percentage of primary component and the minimum LOI for samples as a function of alteration grade. Smectite zeolite grade samples with LOIb1% have a much higher percentage (N80%) of primary components than epidote amphibolite samples with similar LOI values. This is probably due to more abundant of alteration minerals with low-water contents in the higher grade samples. LOI consists mainly of H 2 O + and CO 2. The latter was measured separately in only part of the sample group. The results show that CO 2 is typically b1% in samples with less than 60% alteration intensity with higher CO 2 values (b6.5%) only present where alteration is more intense. A good correlation is between CO 2 and calcite in the samples. 5. Methodology for calculating mass transfer The methodology used in this study is based upon the petrographic characteristics of Icelandic basalts altered in geothermal systems. Fig. 7. A sketch showing a simplified model for the dilution of a chemical rock component with the introduction of vesicle filling. The Y-component in the rock is symbolised in black and becomes relatively reduced when vesicle fillings are added to the sample. Specifically, the methodology arises from the key observation that much of the alteration process entails the filling of primary pores by secondary minerals, effectively diluting the chemical composition of the protolith by mass addition (Fig. 7). This mass addition will dilute most of the components in the original rock and thus whole rock analyses will show lower concentrations of immobile elements even though there has been no mass flux of these elements. The extent of dilution is known through point-counting as discussed above. By identifying components in the rock that have remained immobile during alteration and taking into account this dilution effect, the primary composition of the immobile element in the sample may be calculated. One way of assessing the effect of chemical dilution on the rock composition is to plot the elemental concentration against the rock dilution (Fig. 8). In this diagram the concentration from z to x represents the concentration range measured for element A in unaltered basalt. The tie lines from 100% dilution (point R) to points z and x represent the amount of dilution of z and x, respectively, assuming that the dilutant mineral (or minerals) contains no chemical component A. If element A is immobile and has y concentration in the sample, the dilution will cause the apparent decrease of that component along the line towards R. One may therefore expect that all immobile elements within the basalt range will, with increasing dilution, be contained within the field labelled a. A mobile element Fig. 6. Loss on ignition (LOI) plotted against the % primary rock component (i.e.,100-rock alteration and void filling). Also shown are the best fit lines belonging to each of the alteration groups. Fig. 8. Diagram showing effect of dilution on the concentration of hypothetical chemical component A during alteration of basaltic rocks. Fields labelled a, b, and c : represent, respectively, the range of diluted compositions following no mass transfer, addition of component A, and loss of component A. See text for more details.

6 222 H. Franzson et al. / Journal of Volcanology and Geothermal Research 173 (2008) Fig. 9. A plot showing the relation between SiO 2 % and % mineral deposition (dilution) deduced from petrography for samples grouped into alteration zones. The lines demarcate the basalt compositional field (c.f. lines z R x in Fig. 8). Deviation of the samples outside the basalt field indicates apparent SiO 2 enrichment. may not follow the same trend. If vesicle fillings contain the mobile element, which has been derived from outside the sample, the diluted compositions will shift to the right of the line y R (e.g., field b). If, on the other hand, a chemical component is being removed from the rock, the dilution will move compositions to the left of line y R, and these samples will group within the field indicated by c. Samples for which the extent of addition or depletion of an element is small relative to the variation of that element in the unaltered protolith may not be distinguishable by this method. However, rocks that have undergone significant mass flux should be readily identified. In Fig. 9, the SiO 2 concentrations of many of the altered samples plot to the right of the primary basaltic compositional field, indicating a strong SiO 2 enrichment. This enrichment is seen in samples from all of the alteration zones. Fig. 10 on the other hand, generally shows indications of K 2 O depletion during hydrothermal alteration and dilution. Zircon is an example of an immobile element (Fig. 11), where a majority of the samples are confined within the field of primary basaltic composition. Some samples plotting to the right of the primary basalt compositional field have basaltic andesite compositions and are retained for our purposes as they help define the fractionation trend of the sample group. Fig. 11 shows (a) the strict confinement of the samples within the basalt compositional range, as previously mentioned, and (b) that most samples plot in the left side of the basaltic field, in accordance with the observation that the majority of Fig. 11. A plot showing the relation between Zr and % mineral deposition. Most of the samples fall within the basalt field, as defined by the lines, which indicates that the element has remained relatively immobile. the samples are normal basalts and that more evolved basaltic compositions with higher Zr contents are less abundant. These deductions further imply that dilution must play the largest role in modifying the chemical composition of basaltic rocks within the alteration zones studied here. It is interesting that there are no apparent progressive gradients in chemical transport discernable between samples in the smectite zeolite to the epidote amphibole zones, which may indicate that such changes, if present, are small compared to the chemical dilution effect. An important indication of the immobile elements is that their ratios will remain constant by definition during dilution and depletion of other chemical components (e.g. Barrett and MacLean, 1994; Leitch and Lentz, 1994). Fig. 12 shows the Zr/Y ratios of the basalt samples from successive alteration zones compared with values from the least altered rocks analyzed from the Hvalfjordur and Reykjanes Langjokull volcanic zones (Sveinn Jakobsson unpublished data, and Karl Gronvold unpublished data) and the Geitafell (Thorlacius, 1991) and Hafnarfjall (Franzson, 1979) central volcanoes. The figure shows that the Zr/Y ratios in altered rocks have a similar range to those in their fresh counterparts and further confirms Y as also an immobile element in these systems. If an element is immobile, its original concentration prior to dilution can be found by extrapolating along the dilution line to the original concentration of the element at 0% dilution, as shown along Fig. 10. A plot showing the relation between K 2 O % and % mineral deposition. Lines demarcate the basalt field (c.f. Fig. 8). Note that samples more commonly plot to the left of the basalt field, indicating apparent depletion. Fig. 12. The Zr/Y ratios in the rock sample and a comparison with fresh volcanic rocks (vertical bars) from Reykjanes Langjokull volcanic zone (Sveinn Jakobsson, unpublished data, 1999) and the least altered samples from the Hvalfjordur basalt, Hafnarfjall and Geitafell central volcanoes (Franzson, 1979; Thorlacius, 1991). See text for further explanation.

7 H. Franzson et al. / Journal of Volcanology and Geothermal Research 173 (2008) line R to y in Fig. 8. Barrett and MacLean (1994) suggested that by comparing the concentration of immobile elements prior to and after hydrothermal alteration the mass changes that have taken place during the hydrothermal alteration event can be calculated. The calculation of mass change removes the misleading effects of closure (constant sum of 100%) on the relations between elements in untreated samples (Barrett and MacLean, 1994). After determining the primary concentration of a given immobile element (in this case Zr) by the method described above, we can then calculate the reconstructed composition for each rock component by multiplying the original value by a factor equal to the concentration of Zr in the precursor divided by the concentration of Zr in the altered sample. In this way, we have now resurrected the volume of each chemical component to what it was in the primary rock. If dilution represented the only change in rock composition, then all elements other than those contained in the pore-filling precipitates would reveal their original values by the correction and would plot on the one-to-one line when compared to an unaltered protolith. This allows us to test the hypothesis that mass dilution during pore filling is the dominant chemical change affecting the altered rocks. It is of course not always possible to determine the composition of the unaltered protolith. However, since Zr can be shown to be essentially immobile in these rocks, we can bracket the potential compositional variations in primary composition if we can determine the fractionation relationship between Zr and the element of interest in unaltered rocks. By plotting the corrected Zr concentration against another component and comparing it with the fractionation trend of the respective volcano one can assess whether enrichment, depletion or no apparent change has occurred during alteration. The extensive data available on the composition of fresh basaltic rocks from Iceland provides constraints on the fractionation trends with respect to Zr. The limited range of fractionation, from the spread of Zr concentrations in unaltered basalts, provides constraints on the possible original variations in other elements and allows us to recognize significant mass flux. For this exercise, the samples have been subdivided into 3 groups as described above, i.e. the Hafnarfjall and the Geitafell central volcanoes and the Hvalfjordur plateau basalt succession. Although Icelandic volcanoes follow, in general, a Thingmuli tholeiitic evolutionary trend Fig. 13. Chemical components plotted against Zr. Xs are least altered rocks from Hafnarfjall and Geitafell central volcanoes (Franzson, 1979; Thorlacius, 1991). Diamonds are rocks of 0 33% intensity alteration, squares are rocks of 33 66% alteration and triangles are rocks of % alteration. The line delineates the primary compositional field of the respective volcanoes.

8 224 H. Franzson et al. / Journal of Volcanology and Geothermal Research 173 (2008) Fig. 13 (continued ). (Carmichael, 1964), individual centers may show small deviations from that trend. Grouping the altered samples according to their petrogenesis allows more direct comparison to their primary counterparts and results in a better resolution of compositional changes that have taken place during the hydrothermal alteration. An additional advantage is that this allows a comparison of enrichment/depletion between different geothermal environments, and gives a scope for speculation of heterogeneity in chemical alteration trends in Icelandic geothermal systems. For plotting purposes, the samples have been divided into three groups based on their extent of alteration (Fig. 4) at 0 33%, 33 66% and % alteration (vesicle filling+rock alteration), in order to show gradual chemical changes, if any, from less to more altered samples. Figs. 13 and 14 are examples of some of the chemical plots where the relation between individual components is shown against Zr concentration. Relatively unaltered rocks from the same area are also plotted on these diagrams to indicate the compositional range of the protoliths of the altered rocks. The overall results are summarized in Table 1. It must be emphasized that the results depict the integrated chemical changes that have taken place during the history of the hydrothermal areas. In particular, the central volcanoes have probably undergone several geothermal episodes throughout their lifetime, which may exceed 1 myr. The Hvalfjordur basalt succession, on the other hand, has only been subjected to low-temperature burial alteration with the regional temperature gradient resulting in temperatures ranging from about C, and is apparently devoid of major geothermal anomalies. The results from the central volcanoes can be divided into four main categories: (a) TiO 2, Ce, Nb, Y and Zr show clear indications of being immobile as they fall very near the fractionation line of the least altered samples. (b) The elements that most consistently show signs of enrichment are SiO 2,Al 2 O 3, FeO, and to a lesser extent, MnO, MgO, Na 2 O, CaO and possibly Cu. (c) K 2 O and Sr are enriched in some samples, but are depleted in others. (d) Other elements do not show a clear trend. However, a lack of comparison with fresh rock equivalents prevents an assessment of elements such as Th, U, Pb, As, S, and partly Rb. The hydrothermal chemical changes between the two central volcanoes are similar. The chemical changes observed in the Hvalfjordur plateau basalts are much more subtle, as expected, due

9 H. Franzson et al. / Journal of Volcanology and Geothermal Research 173 (2008) Fig. 14. Chemical components plotted against Zr. + are fresh rocks from the Reykjanes Langjokull volcanic zone (Sveinn Jakobsson, unpublished data, 1999), x are least altered basalt lavas from Hvalfjordur lava succession, squares are lavas from Krafla central volcano (Karl Gronvold, unpublished data), and triangles are the rock samples of this study. to its lower geothermal gradient. These do, however, indicate SiO 2, and K 2 O enrichment and subtle enrichments in Al 2 O 3, FeO, MgO, Na 2 O and possibly Sr. Comparison with some of the trace elements is difficult due to lack of data from fresh rock equivalents. 6. Isocon plots The data presented above allows us to evaluate the general behaviour of individual chemical components within a group of samples for an entire hydrothermal field. The method also allows easy comparison between the magnitude of compositional variation related to igneous fractionation trends relative to those that can be attributed to hydrothermal alteration. The isocon method of determining mass flux has the disadvantage of requiring sample by sample comparison to specific precursor rock compositions. A distinct advantage of the isocon method is that it allows direct evaluation of the mass flux of elements in a specific sample, allowing easier recognition of coupled geochemical behaviour between elements (e.g., K and Na). Our approach in using the isocon plots was to first correct the altered rock geochemical data for the known effects of dilution and to plot these extrapolated values against appropriate precursor compositions to identify mass flux variations that are not due to simple rock dilution by alteration minerals. The isocon diagrams (Fig.15) compare the relationship between the various components within four representative samples, two chosen from each of the central volcanos. Two of the samples have a primary

10 226 H. Franzson et al. / Journal of Volcanology and Geothermal Research 173 (2008) Table 1 Chemical changes in samples subjected to hydrothermal alteration from the Hvalfjordur basalt succession and Hafnarfjall and Geitafell central volcanoes Chemical component Hvalfjordur Hafnarfjall Geitafell Depletion Unchanged Enrichment Depletion Unchanged Enrichment Depletion Unchanged Enrichment SiO 2 X x x TiO 2 X (x) (x) Al 2 O 3 (x) (x) (x) x x FeO(tot) X (x) x x MnO X x (x) (x) MgO X (x) x (x) x x CaO X (x) (x) x x (x) x x Na 2 O X X (x) x x x x K 2 O X (x) x (x) (x) x (x) P 2 O 5 X (x) x x Ce X x x Cu (x) x x (x) V X?? Zn x x x Ga X?? Nb X x x Pb??? Rb X??? (x)? (x)? (x)? Sr (x) (x) x (x)?? Th?? x??? U??? Y X x x Zr X x x As?? S??? x-affirmative, (x)-probable, (x)?-possible,?-uncertain. Zr content of about 150 ppm (H-99, G-62), while the others have about 300 ppm (H-54, G-28) representing relatively primitive and evolved basalt compositions, respectively. The error bars attached to each of the components represent the range of primary compositional values expected at each Zr concentration value within the individual volcano (Fig. 13). Although error limits on the recalculated compositions are difficult to quantify, one may expect that the error would increase with increasing dilution. This is easily observed in Fig. 8 where a small error in petrographic estimation at high dilution could divert the extrapolation line considerably down towards the base line. If the samples have all been properly corrected with respect to mass addition, then the median line should represent the line of no chemical change. Indeed, the immobile elements (Ce, Zr, Y, Nb) tend to lie on this line. An exception is Y in the Hafnarfjall samples, which may possibly be due to a systematic analytical error in the primary trend in the volcano, as deduced from the Zr/Y plot (Fig. 13). In contrast, Fig. 15 indicates a tendency for slight Al 2 O 3 and FeO enrichment in the altered samples, as is also clearly indicated in Fig. 13. MgO can be either enriched or depleted, as are Na 2 O and K 2 O. One might expect that Na 2 O and K 2 O would generally show similar geochemical behaviour and Fig. 15 confirms that the alkalis are either both depleted or both enriched in any given sample. Na 2 O and K 2 O are rather easily leached from rocks during hydrothermal alteration and many rocks are in fact depleted in them. One reason for enrichment may be due to remobilization from nearby more felsic volcanic and intrusive rocks. This hypothesis could be tested by looking at field relations of the altered basalt samples that show alkali enrichment. The isocon plots generally confirm the conclusions that have been attained through plots shown in Figs and support the proposition that changes in chemical composition of the altered rocks is mainly due to addition of mass to the rock by deposition of alteration minerals in primary pores. 7. Discussion Geothermal systems in Iceland have a number of common features. The rocks are dominantly of basaltic composition, with minor amounts of rocks having more evolved compositions within central volcanic complexes. Most of the hydrothermal waters have low salinity with chloride b200 ppm (Arnorsson and Andresdottir, 1995), consistent with heating of local groundwaters. Exceptions are the high-temperature systems on the Reykjanes Peninsula which show salinity approaching that of seawater (Sveinbjornsdottir et al., 1986). Fluid inclusion studies indicate, however, that much of the alteration in the Reykjanes systems may have occurred during an initial freshwater stage during the last glacial period (e.g., Franzson et al., 2002). Studies on the alteration zones in several fields show quite similar distributions of alteration minerals (Fig. 2). There also seems, in general, to be a close correlation between alteration zones and the progressive breakdown of the primary rock and the alteration products formed. The alteration is, to a large extent dependent on temperature, and the maximum temperature at any depth in the explored systems is constrained by the boiling curve. The common occurrence of hyaloclastites in the presently-active systems is not shared by the fossil hydrothermal systems developed in Tertiary rocks, which are dominantly in holocrystalline volcanic rocks erupted in the absence of a glacial icecap. Volcanic glass is very readily altered during geothermal activity, and this will increase the availability of chemical components (e.g., SiO 2 ) into thermal fluids. Therefore the chemical flux in the presently-active geothermal systems may be higher than it was for the Tertiary systems. It must be emphasized that this study describes the overall changes that have taken place during the whole lifetime of the hydrothermal systems. Mineralogical studies of vesicle and vein fillings at Geitafell central volcano show distinct episodes of alteration (Fridleifsson, 1983, 1984). The same pattern of chronologically distinct hydrothermal episodes has been recognized in many of the explored active high-temperature systems (e.g., Lonker et al., 1993). The number of mineral species in individual vesicles and veins in the high-temperature systems at Reykjanes, Nesjavellir, Olkelduhals and Svartsengi (Franzson, 1983, 2000; Franzson et al., 2002), based on about 2000 observations, is very similar. About 70% show two void-filling minerals, about 25% contain three and about 5% show four or more void-filling minerals. The proportion of mono-mineralic veins and vesicles has not been checked specifically. It is interesting to note that many minerals seem

11 H. Franzson et al. / Journal of Volcanology and Geothermal Research 173 (2008) Fig. 15. Isocon plots for four samples from Hafnarfjall (H-54, H-94) and Geitafell (G-28, G-62) central volcanoes. The samples were selected with about 150 and 300 ppm Zr, and from the % alteration group and moderate SiO 2 enrichments. See text for further explanation. to grow to a certain threshold grain size, after which they are succeeded by another mineral (e.g. clays, epidote, prehnite, wairakite, chalcedony and sometimes quartz). Other minerals like calcite, anhydrite (the latter in saline systems), and sometimes quartz, are more rarely constrained by size and tend to fill the available space. Whatever the reason, the limit of mineral size does increase the probability that one could find at least two minerals in any one vesicle. Calcite is the mineral that is most likely to fill available space. However, calcite generally precipitates at a late stage in the mineral sequence after the deposition of the other minerals. The above observations imply that typical rocks in a given high temperature system retain open pore space for sufficient time to record the effects of more than one geothermal episode. Thus the enrichment/depletion trends tend to accumulate over the lifetime of the system. It has also been noted that changes in mineral deposition sequences can be correlated regionally within individual geothermal fields (e.g. Franzson, 2000), which implies that mineral deposition is controlled by large scale hydrological changes in the geothermal system. It is instructive to compare the low salinity Icelandic geothermal systems with basalt-hosted seawater-derived hydrothermal fluids (Alt and Teagle (2000) and references therein and Butterfield (2000) and references therein), as evidenced in ophiolitic rocks and black smoker hydrothermal systems. The overall similarity in the composition of basalts from the seafloor, ophiolites and subaerial Iceland facilitates comparison of these systems. Their physical volcanology does show some differences that could lead to differences in fluid rock interaction. Extensional tectonics and rift zone volcanism are common to each of these settings, but the deep water eruptions, common on mid-ocean ridges and ophiolites, are different from the subareal eruptions in Iceland (Batiza and White, 2000). Seafloor eruptions produce more abundant glassy volcanics and hyaloclastites than subaerial eruptions in the Tertiary, which should enhance the alteration potential of seafloor rocks. However, Pleistocene eruptions in Iceland often occurred beneath ice cover and resulted in extensive formation of hyaloclastite. Seafloor pillow lavas and sheet flows, which characteristically show drain-back features, have very high intrinsic, large scale porosity and permeability, but the permeability drops significantly in older, more deeply buried lavas due to collapse and infilling by later eruptions (Perfit and Chadwick, 1998; Fornari et al., 2004). However, submarine lavas are rarely vesiculated. In contrast, flow top and bottom breccias and highly vesiculated flow tops in subaerial lavas provide laterally continuous zones of high porosity that results in anisotropic permeability that favors lateral flow and restricts vertical convection (Fornari and Embley, 1995; Fornari et al., 2004). One major compositional difference between subarial and submarine lavas is in the concentration of volatile elements, particularly sulfur. Sulfur concentrations of MORBs are typically about 1000 to 1500 ppm, in contrast to values of about 100 ppm in subarial lavas that have degassed during eruption (Wallace and Anderson, 2000). Because potentially acid-generating volatiles (e.g. SO 2, HCl, CO 2 ) are lost during eruption, subsequent hydrothermal leaching will not involve these species and the resultant hydrothermal fluids may not attain low phs that facilitate rock alteration. This is particularly true for hydrothermal systems developed in plateau basalts, such as at Hvalfjordur. Magmatic degassing during intrusive/eruptive events can result in short lived

12 228 H. Franzson et al. / Journal of Volcanology and Geothermal Research 173 (2008) compositional changes in geothermal systems due to acidification reactions (Armannsson et al., 1982). In hydrothermal systems developed in longer-lived central volcanoes (e.g. Hafnarfjall, Geitafell), late stage intrusive rocks can supply both heat and acid-generating volatile species that can change the pattern of alteration of the older, degassed lava flows. Degassing of sulfur from basalt can affect the mobility of other elements as well, particularly the base metals. For example, submarine basalts erupted at higher pressures due to the weight of the overlying water column and therefore do not degass sulfur as readily. Copper in submarine basalts is typically concentrated as chalcopyrite and is generally only mobilised by high temperature (N330 C) hydrothermal fluids due to the low solubility of chalcopyrite in sulfide-bearing fluids. In contrast, copper in subareally erupted basalts that have degassed sulfur, chalcopyrite may not be stable leaving copper in a form that is easily leached and transported by low temperature, oxidized fluids (Lincoln, 1981). The upper temperature limits in shallow geothermal fields are typically set by the boiling temperature curve of the fluid. In deeper, higher salinity systems rapid changes in the physical and chemical properties of the fluid at temperatures approaching the critical point provide similar temperature constraints. The higher pressures and salinities inherent in submarine geothermal systems therefore favour higher ultimate temperatures of water rock reaction in the deeper portions of these system. It is interesting that the general zonation of alteration minerals in Icelandic geothermal systems (Fig. 2) is similar to those observed in drill holes throughout the oceanic crust (Alt and Teagle, 2000 and references therein; Wilson et al., 2006) and ophiolites (Schiffman et al., 1991; Gillis and Banerjee, 2000). The major difference in subarial geothermal systems recharged by meteoric water and submarine systems recharged by seawater is the salinity of the fluid. Chloride complexing is significant for many cations and strongly enhances the solubility of most minerals, especially at elevated temperatures. Higher concentrations of major element cations in seawater-recharged hydrothermal fluids will also affect the stability of alteration minerals making it all the more surprising that the zonation of alteration minerals in seafloor hydrothermal systems is so similar to that in Iceland. Although the general pattern of alteration mineral zonation is similar, the extent of hydrothermal alteration and mass flux from high temperature seafloor systems is significantly different from that of the Icelandic geothermal systems. The geochemical data presented in this paper document that the major changes in the compositions of altered rocks from Icelandic geothermal systems result from the addition of elements during deposition of pore-filling alteration minerals. Isocon plots of altered samples, corrected for dilution, show relatively small mass fluxes for most elements, even at high degrees of alteration. This general observation is true across the range of alteration zones even in rocks that show epidote amphibole alteration. In contrast, basalt/seawater reaction results in significant metasomatism of the rocks (Seyfried and Ding, 1995). As seawater is heated by circulating in the oceanic crust on the recharge limb of hydrothermal convection cells, metasomatic addition of seawater-derived Mg and Ca into basalts releases hydrogen ions lowering the fluid ph (measured at 25 C) into the range that is typical of black smoker fluids. Calculated in situ ph at high temperature are about one unit lower than neutral (Ding and Seyfried, 1992). This low ph fluid has significant capacity for hydrogen metasomatism in the deep reaction zone and the rising limb of the convective circulation. Acid alteration of the basalts releases a significant flux of cations into the hydrothermal fluids resulting in highly cationleached alteration zones, enriched in aluminum and silica, in focussed upflow zones where integrated water rock ratios are high (Humphris et al., 1998; Zierenberg et al., 1988, 1995, 1998; Teagle and Alt, 2004). Silicified basalts in hydrothermal upflow zones can have silica contents in excess of 80 wt.% (e.g. Zierenberg et al., 1995). Alternatively, Fe and Mg metasomatism is hydrothermal upflow zones can result in pervasive chloritization of basalt resulting in bulk compositions approaching chlorite with SiO 2 as low as 35 wt.% (Humphris et al., 1998) and FeOT+MgON25 wt.% (Zierenberg et al., 1988). Thus the most significant difference between seawater and meteoric water recharged basalt-hosted hydrothermal systems is the mass flux from the system, not the mineral assemblages which seem to be controlled predominantly by the thermal stability of the alteration phases and secondarily by the composition of the hydrothermal fluids. 8. Conclusions This paper reports a method to compare hydrothermally altered basaltic rocks and their protoliths that facilitates correction for mass additions to the rock, during hydrothermal alteration. This method takes into account the amount of mineral deposition into the rock, as determined by petrography, and recalculates the mass addition using the immobile element concentration. The derived recalculated composition of the altered rocks can then be compared with the least altered protoliths from the same areas. The comparison showed that hydrothermal alteration causes pronounced Si-enrichment as well as a considerable Al, Fe, Mg and Mn enrichment. Elements such as Ca, Na, K, are also mobile and can be either depleted or enriched. Immobile elements include Zr, Y, Nb, Ce, and probably Ti. An important conclusion from this work is that the predominant process leading to chemical flux in these rocks is deposition of alteration minerals in primary pore space, but mass flux due to recrystallization and replacement reactions are of secondary importance. This contrasts with seawater-dominated hydrothermal systems where metasomatic reactions result in extensive mass flux. Another important conclusion is that similar chemical changes were experienced in the two Tertiary central volcanoes studied, suggesting that these may represent a general trend in Icelandic high-temperature systems. A study of a sample suite of flood basalts within a low-temperature zeolite alteration environment of Tertiary age also shows chemical changes akin to the former, but the changes are more subtle. It is also significant that the general zonation of alteration minerals in Icelandic alteration zones is similar to those in seawater-dominated hydrothermal systems in spite of the differences in the extent of metasomatism. This implies that temperature, not fluid composition, provides the first order control on mineral assemblages in basalt-hosted hydrothermal systems. Acknowledgements We thank Howard Day and Shuwen Liu for very helpful discussion on various aspects of this work, and the late Valgardur Stefansson for critically reading the manuscript and his vigorous support. Patrick R. L. Browne and an anonymous reviewer are thanked for critically reviewing the paper and suggesting a number of improvements. This paper was written during the sabbatical leave of the first author at the Geological Department, University of California, Davis in the USA. The financial support of Orkustofnun to complete this work is acknowledged. References Alt, J.C., Teagle, D.A.H., Hydrothermal alteration and fluid fluxes in ophiolites and oceanic crust. In: Dilek, Y., Moores, E., Elthon, D., Nicolas, A. (Eds.), Ophiolites and Oceanic Crust: New Insights from Field Studies and the Ocean Drilling Program. Geol. Soc. Am. Spec. Paper, vol. 349, pp Armannsson, H., Gislason, G., Hauksson, T., Magmatic gases in well fluids aid in the mapping of the flow pattern in a geothermal system. Geochim. Cosmochim. Acta 46, Arnorsson, S., Andresdottir, A., Processes controlling the distribution of boron and chlorine in natural waters in Iceland. Geochim. Cosmochim Acta 59, Arnorsson, S., Gunnlaugsson, E., Svavarsson, H., The chemistry of geothermal waters in Iceland. II. Mineral equilibria and independent variables controlling water compositions. Geochim. Cosmochim. Acta 47, Batiza, R., White, J.D.L., Submarine lavas and hyaloclastites. In: Sigurdsson, H., Houghton, B.F., McNutt, S.R., Rymer, H., Stix, J. (Eds.), Encyclopedia of Volcanoes. Academic Press, San Diego, CA, pp