IGNEOUS ORIGIN OF CO 2 IN ANCIENT AND RECENT HOTSPRING WATERS AND TRAVERTINES FROM THE NORTHERN ARGENTINEAN ANDES

Thermal travertines are an archive of CO2 sources and sinks in hydrothermal systems. Two major regional factors control travertine precipitation: water availability and CO2 supply. Thus, travertines form a valuable archive of hydrodynamic variability and sources of main contributions to dissolved inorganic carbon (DIC). It is relevant to determine the main DIC sources of thermal waters (i.e., organic-matter degradation, recycled from older carbonates, emission of deep-seated magmatic CO2), as they are key inputs to calculate the lithosphere atmosphere CO2 budget. The Antuco travertine from the Central Andes represents one of such archives, with a total of 500 ky record of accretionary periods (dated as 425-320, 260 and 155 ky BP) related to high hydrothermal activity of hot springs. It consist of two travertine units: (1) a lower massive unit displaying large calcite pseudomorphs after aragonite that precipitated in proximal ponds with Origin of CO2 in hot spring waters 2 abundant water, and (2) an upper stratified unit showing more distal facies bearing siliciclastics and manganese and iron oxides. The replacement of aragonite by calcite in the lower unit was related to the decrease of water salinity in the thermal system through time. In the Antuco travertine, DIC δ C values of travertine parental waters of around -9‰ suggest that CO2 was related to igneous processes and volcanic activity, and released along deep-seated faults. Relative water/rock ratios derived from δO values and fluid-inclusion microthermometric data from travertine carbonates are consistent with an interpretation of greater water availability in the hydrothermal system during the Late Pleistocene than at present. The different petrographic features and isotopic signatures are interpreted to reflect increased water availability during more humid periods in the Altiplano, which triggered precipitation of travertine bodies. Travertine growth took place during interglacial-humid climate periods between Marine Isotope Stages (MIS) 3 and 9, which correspond to highstand events in large lakes of the Andean Altiplano. Results of this study illustrate that volcanic activity, furnishing rather constant CO2 and heat fluxes, were the key controls of thermalism in the Altiplane region during the Quaternary, whilst climatic changes (humid vs. arid periods in the Late Pleistocene) controlled mineralogy, fàcies, and architecture of the travertines. The combined use of δC and Sr/Sr signatures in carbonate precipitates has been proven to be of major relevance to evaluate the CO2 sources along fault zones in this study.

Two major regional factors control travertine precipitation: water availability and CO 2 supply.Thus, travertines form a valuable archive of hydrodynamic variability and sources of main contributions to dissolved inorganic carbon (DIC).It is relevant to determine the main DIC sources of thermal waters (i.e., organic-matter degradation, recycled from older carbonates, emission of deep-seated magmatic CO 2 ), as they are key inputs to calculate the lithosphere -atmosphere CO 2 budget.The Antuco travertine from the Central Andes represents one of such archives, with a total of 500 ky record of accretionary periods (dated as 425-320, 260 and 155 ky BP) related to high hydrothermal activity of hot springs.It consist of two travertine units: (1) a lower massive unit displaying large calcite pseudomorphs after aragonite that precipitated in proximal ponds with

INTRODUCTION
Thermal travertines are an archive of CO 2 sources and sinks in hydrothermal systems (Moore et al. 2005).The most abundant travertine deposits are carbonate bodies precipitated in the proximity of thermal springs due to CO 2 outgassing, although silica and minor amounts of Fe-Mn oxides are also common.The main factors controlling travertine precipitation in hydrothermal systems are CO 2 supply, meteoric water availability, heat source, and faults and fractures allowing deep-water (and solute) circulation (Frank et al. 2000).
In the Argentinean Altiplano of the Central Andes, hydrothermal circulation is driven by contemporary volcanism that provides rather constant heat sources and topographic gradients enabling deep-water circulation along major faults.Nevertheless, present-day hydrothermal activity in the region is limited to a few springs with low discharge, because water availability is constrained by present-day arid conditions (e.g., Klein et al. 1999;Godfrey et al. 2003).
Nonetheless, travertine deposits are a common feature of the Altiplano in northwest Argentina, having formed mainly during Pleistocene humid periods in discharge areas of the Altiplano hydrothermal systems (Blasco et al. 1996).Travertine deposits in this region, such as those located in the proximity of Antofagasta de la Sierra (Alonso and Viramonte 1985) and San Antonio de Los Cobres (the present study), are closely linked in space and time to volcanism and tectonism.
Although few studies provide data on the depositional architecture and geochemical aspects of travertine formation in the Altiplano (Rondeau 1990;Valero-Garcés et al. 2001), travertines formed in similar settings during the Pleistocene have been used to unravel climatic conditions (Rech et al. 2002) and to determine the timing of heat transfer in magmatic-thermal systems (Fournier 1989).
Additionally, the use of geochemical signatures and petrographic features offers the potential to unravel the details of hydrothermal systems through the Pleistocene.
This study documents isotope and fluid-inclusion data from a Pleistocene hydrothermal travertine deposit in the Andean Altiplano, with the general aim of identifying the sources of CO 2 into the thermal system and the factors that controlled travertine deposition.This region in the Altiplano is characterized by arid conditions with low water availability and low soil CO 2 production, and as a result, the vegetation is too sparse to initiate any soil formation (Messerli et al. 1993) or to produce significant amounts of soil-derived CO 2 , which by far is the most common carbon source for spring travertine formation (Pentecost 2005).
The results of this study illustrate how the petrographic features and the geochemical signatures of the Antuco travertine record changes in water availability in the Altiplano.The data reveal that the formation of the Antuco travertine deposits are the sink of deep-seated magmatic CO 2 sourced by hydrothermal waters flowing along a fault system related to a major lithospheric structure in a volcanically active area.As such, the travertine deposits of Antuco are an excellent archive of the influences of different CO 2 sources (endogenous vs. shallower sedimentary sources) and climatederived hydrological changes.This study includes modern hot springs and their recent deposits, as well as older travertine deposits dating back to the Pleistocene.The workflow and environmental proxies used can be broadly useful in the interpretation of paleothermal systems.

GEOLOGICAL SETTING
The Antuco old travertine deposit (24º 10' S, 66º 40' W; 4,180 m above sea level) is located in the Argentinean Altiplano (Fig. 1).The travertine deposits are found cropping out over an area of about 0.5 km 2 along old quarry fronts.The Antuco travertine and other numerous travertine bodies in the region have been related to thermal waters flowing along the NW-SE Calama-Olacapato-Toro lineament (COT), a left-lateral transcurrent fault system about 700 km long (Allmendinger et al. 1983).
A NW-SE-trending volcanic belt developed along the COT fault in the Miocene (Coira et al. 1993;Matteini et al. 2002), and some of the presently -active volcanoes have been active since the Pliocene.Amongst them the Tuzgle is the only active volcano near the Antuco thermal spring (Fig. 1).The first record of the Tuzgle volcanic activity in the region corresponds to an ignimbrite that formed between 1.2 and 0.7 Ma.The last eruption of the Tuzgle volcano corresponds to a 0.1 Ma andesitic lava (Schwab andLippolt 1974, Coira andKay 1993).N-S-elongated Cenozoic intermontane basins developed contemporaneous with the volcanic activity, on the Precambrian metamorphic-intrusive and Paleozoic-Mesozoic terrigenous basement rocks (Fig. 1).These basins were filled mostly by terrigenous sediments and interbedded lacustrine evaporites ("salars").

METHODS
Two field surveys (1998 and 2006) and sixty samples from the Antuco travertine deposits provided the basis for this study.In the samples, the mineral assemblage was determined using a Bruckner X-ray diffractometer (Cu Kα, 40 kV, 30 mA, and graphite monochromator).Fifty mirrorpolished thin sections were petrographically examined to establish the paragenetic evolution of carbonate cements using optical, catodoluminescence (CL), and scanning electron (SE) microscopy.
From the samples, five were selected for fluid-inclusion microthermometry, based on their representativeness of the different travertine facies and the presence of a sufficient number of fluid inclusions.The rock wafers were obtained using a low-speed wire-saw (Well) and, after a first lowpressure automatic polishing (Logitech) were hand-finished before being be analyzed for fluidinclusion microthermometry.Fluid-inclusion associations (FIA, sensu Goldstein and Reynolds, 1994) could not be confidently identified due to the limited number of fluid inclusions in the travertines.Nevertheless, an attempt has been made to group the fluid inclusions from equivalent crystallographic positions (e.g., crystal growth zones, hairline fractures following cleavage) that potentially record similar growth conditions and parental fluids.Fluid inclusions from prismatic aragonite crystals and calcite rimming conduits were analyzed following standard procedures (Goldstein and Reynolds, 1994).The vapor:liquid ratios were calculated from the optical measurement of fluid-inclusion sizes and bubble diameters.A total of 76 determinations of homogenization temperature and ice melting temperature were performed using a heating-freezing Linkam stage.Determination of the eutectic temperature was not feasible due to the small size and low salinity of the fluid inclusions.Aragonite fragments (0.15 g) were crushed under high vacuum (diffusion pump) and analyzed using an Extorr quadrupole to estimate the main evolved fluids.
Ten thick sections were used for δ 13 C and δ 18 O determinations.Fifty-seven micromilled samples (about 70 μg in weight) were drilled from selected precipitates and cements in the mirror-polished thick sections (150 μm -thick) with a Merchantek computer-monitored micromill.Use of a micromill provided pure samples of each carbonate cement phase, preventing isotopic contamination between phases.Once the mineralogy of each sample was determined and diagenetic alteration was excluded, δ 13 C and δ 18 O were measured in a Finnigan MAT 252 mass spectrometer equipped with a Kiel device.The results are expressed as per mil (‰), and the values are reported in relation to the Vienna Pee Dee Belemnite (V-PDB) international standard.The precision for both δ 13 C and δ 18 O for duplicate samples and standards is better than 0.1‰ (2σ).Four selected samples were also measured for 87 Sr/ 86 Sr in a Finnigan MAT 262 TIMS following a standard method (Pin and Bassin, 1992).The NBS-987 standard ( 87 Sr/ 86 Sr = 0.710274; σ = 0.000016) was used for accuracy and precision tests.
δ 13 C CO2 values of the parental waters in isotopic equilibrium with the Antuco carbonates were calculated from the δ 13 C of carbonates, the temperature of spring waters, and the homogenization temperature obtained from fluid inclusions.Water parameters in the present-day thermal spring (temperature, pH, and salinity) were measured during the field surveys.Water and air temperatures were measured with a thermometer.Waters were sampled separately for major and minor solutes and 1 vol% pure HNO 3 was added to water samples obtained for determination of minor solutes.
Trace and minor solute contents were determined by ion chromatography and ICP-OS and ICP-MS following standard procedures.Additional samples were obtained for δD and δ 18 O analyses and measured in a double-inlet mass spectrometer after equilibration with CO 2 (Epstein and Mayeda, 1953) and H 2 (Pt catalyst; Kirshenbaum, 1951;Horita et al., 1989) respectively.δD and δ 18 O values are reported in ‰ vs. V-SMOW, and reproducibility of international standards and sample duplicates were, respectively, better than σ = 1.5 and 0.2‰.
Four samples from the Lower Travertine were dated with high-resolution U-series dating methods.U and Th were measured on an ICP-MS (Finnigan Element) following techniques developed at the Minnesota Isotope Lab (MIL) (Shen et al. 2002).Two more samples from the Upper Travertine were analyzed by alpha spectrometry following chemical separation and isotope electrodeposition procedures (Talvitie, 1972;Hallstadius, 1984;Bischoff et al. 1988).Age calculations were based on the computer program by Rosenbauer (1991).

The Antuco Travertine
The Antuco travertine deposits appear closely related to the Antuco hot springs (Fig. 1).These deposits consist of two main carbonate bodies (Lower Travertine and Upper Travertine) separated by a major unconformity and a piedmont breccia deposit (Fig. 2).
The Lower Travertine (LT) is the main carbonate body in the area and is located some 20 m away from the present-day hot springs.It is dome-shaped and about 8 m thick and 30 m wide.This lower travertine body is made up of centimeter-to decimeter-thick carbonate beds (Fig. 2).
Travertine beds (Fig. 3A-C) are composed of calcite displaying pseudo-hexagonal twinned prismatic pseudomorphs after aragonite that reach up to 15 cm long and 4 cm wide.The shape and arrangement of the original aragonite crystals resemble those described in recent vent and proximal pond environments (Jones and Renaut 1996;Fouke et al. 2000).The originally aragonitic crystals have been extensively replaced by microsparite and sparry calcite.Porosity between the large aragonite crystals is partly filled by centimeter-to millimeter-size prismatic aragonite crystals, and millimiter-size aragonite needles and epitaxial overgrowths of calcite.Aragonite relics up to 5 mm across are still preserved with optical orientation following the original shape of the original crystals (Fig. 4).
Beds of the LT body are crossed by pipe-like conduits, rimmed by pink calcite crystals 1 to 15 cm-long (Fig. 4).These conduits are filled with micrometer-size iron and manganese oxide crystals.The conduit-rim calcite cements exhibit a columnar fabric (assemblage of roughly parallel, elongated calcite crystals, 0.1 mm to several centimeters in length) growing competitively inwards from the conduit walls.On the upper surface of the LT body, karstic cavities and cylindrical pits are partly filled with calcite cements with microcrystalline, dripstone, shrub, and columnar fabrics (with crystals up to 2 cm in length).
Piedmont alluvial deposits, 15 m-thick, overlie the Antuco LT (Fig. 2).They are composed of poorly cemented depositional breccias with large and angular volcanic clasts (up to 1 m-long) that cover the Lower Travertine body and the karst surface.Fragments of the LT have been identified within the breccias, illustrating the timing of their formation.
The Upper Travertine (UT) body is located about 100 m downstream from the present-day hot springs and overlaies the piedmont alluvial deposits.It consists of interbedded convex-concave sandstone layers and grey bedded travertine (Fig. 2) with variable amounts of volcanic clasts and carbonate fragments.Centimeter-to decimeter-thick layers of manganese and iron oxides are common near the basal part of the UT body.The main fabrics of this UT body consist of microsparite and shrub calcite crystals (sensu Chafetz and Guidry, 1999) up to 4 cm-long with relics of cyanobacterial filaments.Fabrics and sedimentary structures of this travertine body correspond to proximal and distal slope facies (sensu Fouke et al. 2000).

Age of Travertines
The terrigenous content of the dated samples is negligible since the 230 Th/ 232 Th ratio values are low (Table 1) and can be considered as not contaminated.Uranium-series disequilibrium dating indicates that the Antuco LT beds precipitated between 427 ± 60 and 324 ± 27 ky BP (aragonite crystal samples) and that conduit-rim calcite cements formed at 259 ± 30 ky BP.A sample located 15 cm below the top of the UT deposit yielded an age of 156 ± 11 ky BP (Table 2, Fig. 2).
The carbonate travertine of the UT body includes δ 13 C and δ 18 O values that are distinct from the isotopic compositions of those in the LT body.The δ 13 C V-PDB values range between +1.6 and +5‰, whereas the δ 18 O V-PDB between -4.6 to -11.7‰ (see Data Archive, Figs.3C and 5A).
The isotopic composition of present-day thermal waters (δ 18 O V-SMOW = -2.3‰) is enriched in 18 O with regard to the regional meteoric water (snow sample: δ 18 O = -10.3‰;Godfrey et al. 2003), whilst the δD (δD V-SMOW = -63.5‰) is similar (snow sample: δD V-SMOW = -60.9‰).The δ 13 C V-PDB values of the thin iron-oxide-coated calcite crusts that precipitate from thermal water near the spring oscillate between +2.9 and +3.1‰ and their δ 18 O V-PDB values between -1.3 and -1.1‰ (see Data Archive, Fig. 5A and B).These δ 18 O values suggest isotopic equilibrium between present-day thermal waters and recent calcite precipitates when applying the Friedman and O'Neil (1977) equilibrium equation. 87Sr/ 86 Sr values in two samples of aragonite from the LT carbonates are respectively 0.720274 and 0.720595.Two samples of calcite from the UT body display values of 0.718561 and 0.718768.

Temperature and Salinity of the Lower Travertine Parental Waters
Fluid inclusions occur in the primary prismatic aragonite crystals and in calcite crystals rimming the pipe-like conduits in the LT, but they are absent in the UT precipitates and in the karst-related cements.The following three main groups of fluid inclusions are evident in the LT: Group 1.This group corresponds to fluid inclusions in preserved aragonite crystals (Figs. 4 and   6).These primary fluid inclusions, elongated 12 to 20 μmlong and 2 -4 μmwide, lie parallel to the aragonite crystallographic c axis (Fig. 7).This group has two-phase inclusions with a vaporliquid (V-L) ratio of about 9 vol % on average.Some populations of fluid inclusions show inconsistent ratios (vapor bubbles from 7 to 60 vol %) and, in some crystal zones, include liquidonly inclusions.The originally elongated primary fluid inclusions (with consistent vapor:liquid ratio) appear to pass into a succession of aligned smaller fluid inclusions with variable vapor:liquid ratios, suggesting that stretching processes have partially affected this population.These inclusions with stretching patterns were rejected for microthermometric purposes following the recommendations of Goldstein and Reynolds (1994) and Bodnar (2003).The 15 measured primary liquid-vapor inclusions display homogenization temperatures ranging from 76 to 98ºC and ice melting temperatures between -1.5 and -2.7ºC (3.1 to 4.5 wt % NaCl equivalent; Bodnar, 1993).
The mean value of the homogenization temperature is 87.5ºC, and the mean salinity for the parental waters is 3.4 wt % NaCl equivalent (Table 3, Fig. 6).The main fluid released by crushing under vacuum of two aragonite crystalline samples, including both FI groups 1 and 2, was water steam (around 90 vol %) with minor amounts of N 2 and O 2 (around 8% of N 2 and O 2 ) and CO 2 (2%).The low value of atmospheric gases eliminates boiling as a process responsible for the inconsistent vapor:liquid ratios.No CO 2 bubbles are observed in the fluid inclusions, which is consistent with the very low CO 2 content measured in the extracted fluids Group 2. These fluid inclusions are also present in the aragonite crystals (Fig. 8).They intersect the Group 1 fluid-inclusion assemblages obliquely with respect to the aragonite c axis, following cleavage planes.These inclusions are small (diameter ranging from 5 to 20 μm; average about 8 μm) with vapor:liquid ratios between 6 and 12 vol % (Fig. 8).The 33 fluid inclusions measured from this population display homogenization temperatures from 95 to 110ºC and ice melting temperatures between -0.6 and -1.5ºC (1.1 to 2.6 wt % NaCl equivalent; Bodnar, 1993).The mean value of the homogenization temperature is 106.5ºC, and the mean salinity of the parental waters is 1.4 wt % NaCl equivalent (Table 3, Fig. 6).
Group 3.These fluid inclusions are located in the conduit-rim calcite cements (Fig. 9).They consist of two-phase inclusions with an average diameter of 9 μm (range from 5 to 20 μm).The inclusions are arranged along the growth zones of the columnar calcite crystals (Fig. 4) and commonly exhibit negative crystal shapes (Fig. 9).The 28 fluid inclusions measured from this population display homogenization temperature values from 103ºC to 132ºC and ice melting temperature values between -0.3ºC and -0.6ºC (0.5 to 1.1 wt % NaCl equivalent; Bodnar, 1993).
The mean value of the homogenization temperature is 115.2ºC, and the mean salinity value of the parental waters is 0.9 wt % NaCl equivalent (Table 3, Fig. 6).

Evolution of Thermal Water
Travertines are shaped by hydrothermal processes, and these commonly vary with time.The geochemical and petrographic evidence in this area clearly illustrates the nature and timing of this variability.In this study, δ 18 O V-SMOW values of parental waters were calculated from travertine δ 18 O values and homogenization-temperature data of primary inclusions using the well-established fractionation equations between water and aragonite (Zheng, 1999;Bahm et al. 2000) and between water and calcite (O'Neil et al. 1969;Friedman and O'Neil 1977).The results illustrate that LT aragonite crystals precipitated from low-salinity waters with δ 18 O V-SMOW values similar to those of present-day regional meteoric water (Fig. 5B) at temperatures of about 87ºC (boiling point of water at Antuco).Although the parental waters were hot and circulated through the thermal system, they maintained the original meteoric isotopic signature and their δ 18 O values did not show major changes due to water-rock interaction.The measured homogenization temperatures in the LT suggest that aragonite precipitated from boiling water under atmospheric pressure.
In contrast, calcite crystals that replace aragonite and karst-related precipitates have higher δ 13 C and δ 18 O values than the primary LT carbonates, and their isotopic values follow a positive trend (Fig. 5A).This trend is explained by the cooling, as suggested by microthermometrical data, and outgassing of thermal waters that circulated through the travertine body, triggering precipitation of replacive calcite and karst-related cements.These karst-related cements record the highest δ 13 C and δ 18 O values in the LT carbonates, suggesting that the karst system formed from cooled and outgassed thermal waters.
The homogenization temperature (115ºC) of the conduit-rim calcite cement (group 3) fluid inclusions and the δ 18 O values of calcite (-22.2‰)suggest that it precipitated from waters with δ 18 O V-SMOW values (about +2.5‰) higher than those precipitating the main travertine body (Fig. 5B), which provides evidence of water-rock interaction along thermal-water pathways.The higher homogenization temperatures suggest that these rimming calcite crystals formed under hydrostatic pressure equivalent to about 15 m of water column.
The homogenization temperatures recorded in the Group 2 fluid inclusions (106ºC) are considerably higher than those from the primary fluid inclusions of Group 1 (87ºC) and similar to fluid inclusions of Group 3.This observation suggests that the replacement of aragonite occurred as a result of the circulation of thermal fluids through the travertine body under hydrostatic pressure (Fig. 4).Fluid-inclusion data (from Group 1 to Groups 2 and 3) record a temporal trend towards hotter and less saline thermal waters (Fig. 6C).Less saline waters, similar to the group 3 fluid inclusions, were in disequilibrium with aragonite, causing their replacement by secondary calcite in a later stage of the evolution of the LT body.Evidence of this alteration process has been found in the aragonite relics (Fig. 4): besides the primary fluid inclusions (Group 1 of FI) that correspond to original aragonite-crystal parental waters, Group 2 of FI, located along cleavage plains in the aragonite relics, records the circulation of lower-salinity and higher-temperature late fluids.
δ 18 O values of LT parental waters and of present-day thermal waters were taken as the endmembers of the water evolution trend during the last 500 ka (Fig. 5B).The δ 18 O V-SMOW of parental waters precipitating the gray travertine (UT body) was calculated from the trend of parental water evolution in accordance with their age.In these, isotopic values range between -6.2 to -4.2 ‰ (Fig. 5B).The temperatures of precipitation (29 to 38ºC) of the UT body were calculated from the mean δ 18 O V-PDB of UT carbonate values and from parental-water δ 18 O V-SMOW values.
Water temperature in thermal systems depends on of the dynamics of the convection cells.The observation of present-day temperature of the spring waters (around 25ºC) is interesting in this regard.Low temperatures (or temperature changes) in thermal springs have been interpreted as due to mixing of thermal waters with shallower aquifers (López-Chicano et al. 2001;Dilsiz et al. 2004).
Nonetheless, the Antuco spring waters do not show clear evidence of mixing, for the following three reasons: (1) the thermal water is very rich in, B, Li, and As and shows Fe (and Mn) mineral precipitates, (2) water salinity values are also in the range of paleosalinities calculated from microthermometric measurements in the lower travertine (stage of deep and highly efficient convection cell), and (3) δ 18 O values of thermal waters are very high with regard to meteoric water.
Instead, the low temperature is interpreted to be related to the very high rock/water ratio.In this scenario, the heat transport is limited due to the scarcity of water, and thus the temperature of waters reaching the surface is low.
Values of salinity around 4 wt % NaCl, similar to those measured in the Group 1 fluid inclusions (in aragonite relics), have been reported as common in convective systems around magmatic intrusions (Cathles, 1977).In the case of Tuzgle volcano, the presence of free saline fluids at 8 km depth has been documented using magnetotelluric and gravity measurements (Sainato and Pomposiello, 1997).

Carbon Source
Soil-derived CO 2 is by far the most common source of carbon to spring travertine waters (Pentecost, 2005).The δ 13 C value of pedogenic carbonates is controlled by the predominant vegetation type in the region, but carbon can be fractionated differently as a function of the photosynthetic pathway of plants growing in the soil.δ 13 C values range between -30 and -16‰ for C3 plant soils and between -18 and -10‰ for C4 plant soils (Vogel, 1993), although C4 plants are scarce in high-elevation regions (Quade et al. 1989;Rech et al. 2002).Thus, values of δ 13 C around -6‰ for LT carbonates (around -9‰ for CO 2 , as derived from the equations in Bottinga, 1969;Ohmoto and Rye, 1979;and Romanek et al. 1992) in the Altiplano area are consistent with carbon dioxide being supplied mostly by igneous sources (Craig, 1963;White et al. 1990;Hoefs, 1997;Pentecost, 2005).
The δ 13 C difference (about 8‰) between both travertine bodies (Fig. 5A) might suggest a change in the carbon source.The increase in δ 13 C (and δ 18 O) in the UT could be interpreted as resulting from recycling of marine limestone units (Blasco et al. 1996) present at depth in the area.Isotopic mixing of δ 13 C from these older marine carbonates (with assumed values approximately between -2 and +5 ‰; Garrels and Lerman, 1981;Carpenter and Lohmann, 1999;Prokoph et al. 2008;Marquillas et al. 2007) would be irrelevant with respect to the igneous signature (accepted around -7 to -6 ‰; Hulston and McCabe, 1962;Craig, 1963;Ferrara et al. 1963;White et al. 1990) that clearly dominates along the LT deposition (Fig. 10).However, the contribution of CO 2 of carbonate origin could have increased during UT deposition that displays δ 13 C in the range of the marine carbonates, due to the lower water/rock ratios evident during its formation (sensu Taylor, 1987).
87 Sr/ 86 Sr values in the Antuco travertines show a strong radiogenic signature (Fig. 10) consistent with a source from leaching of old silicate basement rocks (Fig. 1; Precambrian and Paleozoic).The LT of Antuco records high radiogenic signatures (around 0.7204), similar to those reported in the Menderes Massif geothermal system (SW Turkey) for thermal waters that interact with metamorphic rocks (Vengosh et al. 2002(Vengosh et al. , 2003)).The same authors found much lower values (around 0.7078) for the Pamukkale (Turkey) thermal travertines, evidencing the recycling of marine Paleozoic carbonates present in the basement.The 87 Sr/ 86 Sr values in the Antuco UT, although strongly radiogenic and far from the marine values, as would correspond to a Pamukkale-like recycling, are slightly lower than the LT ones (Fig. 10).Nevertheless, these values do not follow a δ 13 C -87 Sr/ 86 Sr trend that would be expected in the presence of fluid mixing between the LT and marine carbonates.Thus, the observed δ 13 C trend cannot be explained as a result of contributions from recycling of older carbonates.Other sources of carbon, like the leakage from hydrocarbon reservoirs releasing CO 2 (e.g., Davies and Smith, 2006), have been discarded based on the following: (1) A hydrocarbon-related source commonly causes very low values (lower that -10‰; e.g., Zhang et al. 2008) or a large variability of values, when mixed with other contributions.The δ 13 C of calcite and aragonite precipitates are remarkably constant for the LT unit.More variability would be expected in case of mixtures of hydrocarbon-related and carbonate-recycled CO 2 .
(2) There is no geological evidence of hydrocarbons in the studied area.The Yacoraite Formation contains hydrocarbons at locations hundreds of kilometers eastwards, out from the Altiplano.The equivalents in the Antuco area consist of shallow-water calcarenites with no recorded presence of hydrocarbons (Alonso et al. 1984;Marquillas et al. 2007).
The δ 13 C values of the UT samples fit better with a fractionation trend of outgassing and cooling (Zheng, 1990).Isotopic fractionation between CO 2 and aragonite or calcite is well known in a wide range of temperatures (see equations in Bottinga 1969, Ohmoto and Rye 1979, and Romanek et al. 1992).The calculated δ 13 C values of the parental CO 2 do not vary considerably between the two travertine bodies, which suggests that the δ 13 C difference could result mainly from fractionation at different temperatures.Calculated parental δ 13 C CO2 values during precipitation of the Antuco LT are around -9‰, similar to present-day carbonates, whereas the UT carbonates, which precipitated distally from the spring at temperatures between 28 and 39ºC (calculated from the water evolution trend; see above), record parental δ 13 C CO2 values that range between -6.8 and -5.9‰.These slightly higher δ 13 C values for the UT can be attributed mostly to the outgassing effect that favors escape of light CO 2 from spring waters (White et al. 1990;Valero et al. 1999;Valero et al. 2001;Cartwright et al. 2002).
At present, the Tuzgle stratovolcano is the only active volcano in the region (De Silva and Francis 1991), and its magmatic chamber may constitute the heat and CO 2 source of the Antuco hydrothermal system.Nevertheless, there are also other (Pliocene-Pleistocene) volcanic centers in the area such as Tocomar (0.55 Ma), Negro de Chorrillos (0.45 Ma), and San Jerónimo (0.78 Ma) that have erupted along the COT during the Pleistocene (Coira et al. 1993;Petrinovic et al. 2006).
These could also be magmatic CO 2 sources in the region.Nonetheless, it seems that heat and CO 2 sources can be considered practically "constant" in the area over the last million years, in that (1) the available information on thermal gradients and fluxes around magmatic intrusions suggests cooling times in the order of 10 6 years (Ingebritsen et al. 2001, Deckart et al. 2005) and (2) the Tuzgle stratovolcano has been active during the last million years.

Paleoclimate and Water Availability
The primary source of precipitation in the Altiplano region in the Central Andes is related mainly to the arrival of Atlantic moisture via the Amazon Basin (Garreaud et al. 2003).The high lake levels recorded in the large lakes of the Altiplano (Titicaca, Salar de Atacama, and Uyuni Basin lakes) during the interglacial-warm stages (from MIS 3 to MIS 9) suggest humid climate conditions.
Their low lake level stages correspond to glacial stages (from MIS 4 to MIS 8) (Bobst et al. 2001;Fritz et al. 2004;Fritz et al. 2007;Placzek et al. 2006).Nevertheless, humid conditions prevailed during the Last Glacial Maximum in the Altiplano (Baker et al. 2001;Fritz et al. 2004;Sáez et al. 2007) (Fig. 11).This correspondence between maximum summer insolation on the Altiplano Plateau and increased moisture availability during the Late Pleistocene suggests an orbital control of the South American summer monsoon (Strecker et al. 2007).
The Antuco record spans from MIS 6 to MIS 9, but it is not continuous, because of stratigraphic hiatuses.The Antuco travertines were formed during three main growth periods, as constrained by the hydrothermal activity (Fig. 11), which correspond roughly to the interglacial-humid and glacialdry climate pattern of the Altiplano between ca.350 and 57 kyr BP.The 238 U/ 230 Th age dates illustrate that the first growth event, the lower travertine body (LT), was formed during the humid interglacial of MIS 9.The second event, evident in the stockwork of pipe-like conduits, which clearly crosscuts the LT beds, was formed around 60,000 years later than the LT body, during the dry MIS 8 and/or the humid MIS 7. The third travertine growth event corresponds to the Upper Travertine (UT) body, which was deposited during the dry MIS 6.
The similar δ 18 O V-SMOW values of LT parental waters and recent regional meteoric waters (Godfrey et al. 2003) (Fig. 5B) suggest that the LT travertine precipitated during periods of high availability of meteoric water in the thermal system.High water/rock ratios helped to maintain the original δ 18 O signature of meteoric waters.Homogenization temperatures, close to the water boiling point at the Altiplano altitude, and textural features such as the limpid aragonite crystals devoid of clastic impurities, indicate that the LT aragonite precipitated close to the spring, probably in pools.
Isotopic and fluid-inclusion data obtained from the conduit-rim calcite cement, with homogenization temperatures higher than the boiling point at the Altiplano surface, suggest that these carbonates formed in vents and precipitated from evolved thermal waters in conditions of lower water/rock ratio and higher δ 18 O V-SMOW of parental waters than the previously precipitated travertines (Fig. 5B).These conduits were formed around 260 ky BP (MIS 7-8) when the hydrological conditions were intermediate compared to the other periods of travertine growth (Fig. 11).From 260 ka to the present, the general increase of δ 18 O in carbonates and their lower formation temperatures (either interpreted or measured) reflect a decrease in the water/rock ratio, suggesting an increase in aridity towards the present-day conditions (Fig. 5B).
The high δ 18 O V-SMOW values of the UT parental waters and of the present-day calcites (Fig. 5B) suggest a decrease in the water/rock ratio.The UT was formed during more arid conditions and lower water availability, and it coincides with a glacial stage (ca.155 ky BP, MIS 6; Fig. 11).The presence of cyanobacterial filaments and clastic materials in these carbonates also suggest that they precipitated in downstream pools.These distal pools (or distal slope facies) commonly develop a rich variety of living organisms arranged along the thermal and chemical gradient (Renaut and Jones 2000;Fouke et al. 2000).Also, distal pools commonly contain clastic remains, either from outside the thermal system or produced by internal reworking.Isotopic trends of carbonate minerals and parental water reflect changes in the water supply that fed the Antuco thermal system during the last 500 ka.Late Pleistocene humid conditions assured a high water supply and a high water/rock ratio, favoring precipitation of travertines in pools from waters with low δ 18 O V-SMOW values (meteoric waters not modified through rock-water interaction).

CONCLUSIONS
From 260 ky BP to the present, water/rock ratios decreased as a consequence of increasing aridity, resulting in precipitation of carbonates enriched in 18 O.
Calculated parental δ 13 C CO2 values (around -9‰ V-PDB) of the Antuco travertines indicate a stable and dominantly igneous CO 2 source in thermal waters during the last 500 ka.Igneous CO 2 dissolved under pressure in thermally heated groundwater was fed through deep-seated faults.The outgassing of these CO 2 -rich rising waters at the surface drove travertine precipitation.Compositions of marine carbonates (lower Ordovician, lower Permian, and upper Cretaceous in age) are marked as rectangles for reference.Also indicated are the ranges of: (1) common δ 13 C values for deep-seated magmatic CO 2 (White et al. 1990;Zhang et al. 2008), and (2) common 87 Sr/ 86 Sr values corresponding to thermal recycling of old marine carbonates (Vengosh et al. 2002).

Figure and table captions
The values of the Antuco travertines suggest a magmatic provenance.
Figure 11.Sketch of the accretionary stages recorded in the Antuco travertines and their links to climatic records.A) Main growth phases in the Antuco travertine, illustrating the relative water availability for each growth phase.B) Marine isotope stages (MIS) and the Vostok CO 2 record (Petit et al. 1999).C) High and low lake level periods in the Altiplano lakes, reflecting humid and arid periods, are represented.Uyuni Lake changes are based mostly on the percent carbonate content (Fritz et al. 2004, Placzek et al. 2006).Atacama Lake changes are based in the lithofacies distribution (Bobst et al. 2001).Lake Titicaca changes are based on % of freshwater diatom Neogene volcanic activity and variations in the moisture content in the Central Andes influenced the formation of large travertine deposits in the Altiplano during the Late Pleistocene.Carbonates of the Antuco thermal travertine bodies precipitated during the last 500 ka, during which time changes in water availability are reflected in different petrographic features and isotopic compositions.In this area, three main periods of travertine growth are related to different phases of water availability that roughly agree with interglacial-humid periods in the Altiplano: (1) growth of the Lower Travertine body, about 425 and 320 ky BP, when water availability was high, (2) formation of the stockwork of vents crossing the LT about 260 ky BP, when water availability was intermediate, and (3) deposition of piedmont and the Upper Travertine body about 155 ky BP, when water availability was low.

Figure 1 .
Figure 1.Geological map of the study area in the Central Andes (modified from Matteini et al.

Figure 2 .
Figure 2. Schematic cross-section of the Antuco travertine deposits cropping out along an old

Figure 3 .
Figure 3. A) View of the Lower Travertine showing bedding (upper part) and vertical, palisade-like,

Figure 4 .
Figure 4. Interpretive sketch of the paragenetic sequence recorded in the Antuco Lower Travertine

Figure 5 .
Figure 5. A) Cross-plot of δ 13 C vs. δ 18 O values of carbonates from the travertine deposit.Note the

Figure 6 .
Figure 6.Attributes of fluid-inclusion groups identified in the Antuco travertine carbonates.A)

Figure 7 .
Figure 7. Photomicrographs of fluid-inclusion populations under plane-polarized light in a double-

Figure 8 .
Figure 8. Photomicrographs of fluid-inclusion populations under plane-polarized light in a double-

Figure 9 .
Figure 9. Photomicrographs of fluid-inclusion populations under plane-polarized light in a double-