The gas samples are characterized by high concentrations of O₂, N₂ and Ar with respect to air values (Table 1), indicating a significant contamination by atmospheric-derived gases. In the CO₂-N₂-O₂ ternary diagram shown in Figure 3A, the data are on a mixing line between air and a deeper gas that is rich in CO₂. This mixing can occur during sample collection, or may represent air trapped in the fumarolic system. We quantified this contamination and corrected the chemical composition using Eqs 1, 2, and by assuming that O₂ was only derived from the atmosphere (Rizzo et al., 2019). i c o r r e c t e d = i s a m p l e − i a i r * F 1 − F (1) F = O 2 s a m p l e O 2 a i r (2)

These equations enabled calculation of the concentration of a species denoted by i , before air addition in the system (i.e., i c o r r e c t e d ). i s a m p l e , i a i r , O 2 s a m p l e and O 2 a i r represent the concentration of this species i and the O₂ content, in the sample and in air, respectively. F is the fraction of atmospheric-derived gases (expressed in %). Corrected chemical compositions are presented in Table 2. The results show that CO₂, N₂, H₂ and CH₄ were the main components before addition of air in the system. An uncertainty in the correction must be considered because some O₂ consumption may have occurred before gas emission at the surface.

Noble gas samples are also affected by atmospheric contamination, as shown by the low values of the ⁴He/²⁰Ne ratio (Table 1; Porcelli et al., 2002). We corrected the He content and isotopic ratio (i.e., R/R_a) by using Eqs 1–3, respectively (Craig et al., 1978). R c R a = R R a s a m p l e * H e 4 N e 20 s a m p l e − H e 4 N e 20 a i r H e 4 N e 20 s a m p l e − H e 4 N e 20 a i r (3)

The He isotopic ratio corrected for contamination by atmospheric-derived component i . e . , R c R a is obtained by using the ratio of the sample i . e . , R R a s a m p l e and the H e 4 N e 20 ratios in the sample and in air (i.e., H e 4 N e 20 s a m p l e ; H e 4 N e 20 a i r = 0.318 ).

The low concentration of CO₂ in air (Table 1), compared to the deep source identified in Figure 3A implies that atmospheric contamination has a small or negligible impact on the measured C-isotope composition (δ¹³C-CO₂), even if the fraction of atmospheric-derived gases is important in the sample. This isotopic ratio has therefore not been corrected.

Figures 3B, C show the temporal evolution of F, the fraction of air in the reactive gas samples, and the ⁴He/²⁰Ne used to study the contamination by atmosphere in the noble gas samples. Figure 3B shows that the proportion of air in the gas mixture has tended to increase continuously since the end of the 2011–2012 unrest. This is also evident in Figure 3C in which the ⁴He/²⁰Ne decreases over the same period, indicating an input of atmospheric-derived ²⁰Ne compared to the deep-radiogenic ⁴He. As most of the data from the literature and the 2015–2022 samples from the INVG come from the same fumaroles (NK), we can assume that this increase in air contamination is unrelated to the sampling. In addition, the samples were taken at different periods each year. The possibility of meteorological control seems unlikely, as it should result in random variations rather than the general trend observed over several years. We can therefore infer that air-derived gas is present in the subsurface. As the samples are admixtures between atmosphere and volcanic/hydrothermal sources, the contribution of the latter is becoming decreasingly important. The self-sealing process of the channels that carry the deep gas upwards or the opening of air circulation conduits could be the cause of these temporal anomalies. Nevertheless, the possibility of a lowered flux of magmatic/hydrothermal gases seems the most likely explanation.

In volcanic systems with a well-developed hydrothermal system, periods of quiescence and unrest are tracked by CO₂/CH₄ (Chiodini, 2009 and references therein). Increases of CO₂/CH₄ are usually attributed to CO₂ exsolution from a magmatic body and it generally reflects periods of unrest during which the melt is depressurized. Figure 4A shows the temporal evolution of the CO₂/CH₄ ratio for the fumaroles of Nea Kameni. From July 2010 to March 2012, a significant increase of this ratio was interpreted as due to the upwelling of new and poorly degassed magma (Rizzo et al., 2015). After this event, in mid-2012, the volcano regained stability, with CO₂/CH₄ attaining a value that is similar to the pre-unrest period. Since then no significant anomalies were recorded and our chemical data from 2015 to 2022 follows the same trend than the one measured after the volcanic unrest.

N₂ concentrations corrected for atmospheric contamination are plotted over time in Figure 4B. Similarly, when compared to the CO₂/CH₄ ratio evolution from July 2010 to March 2012, the N₂ concentration has dramatically increased, with the highest values reached in July 2010 (up to 55%). The N₂ is derived either from subducted sediment where it is fixed as NH 4 + or from the mantle (Sano et al., 2001; Barry and Hilton, 2016). The similarity between the high N₂ content and the 2011–2012 unrest suggests that the magmatic body was enriched in N₂. This anomaly stopped around March 2012. The deep N₂ concentration has slightly increased over time although an uncertainty in the corrected values must be considered as the fraction of air increased after 2012 (Figure 3B).

The temporal evolution of H₂ concentration for different fumaroles of Nea Kameni are shown in Figure 4C. From March 2011 to February 2012 a sharp increase of H₂ has been detected in a fumarole of Nea Kameni (Tassi et al., 2013). Based on studies of geothermal gas equilibrium (Giggenbach, 1980) and the H₂/H₂O, Tassi et al. (2013) attributed this change to: i) a primary H₂ production by thermal dissociation of H₂O from a heat pulse and ii) a passive enrichment caused by steam condensation at shallow depth. The origin of the heat pulse has been related to a fresh magma injection or increase in rock permeability due to the seismic swarm. The H₂ concentration has significantly decreased after February 2012.

Figure 4D shows the number of earthquakes (M > 1; depth <30 km) inside the caldera (cf. Figure 2B for location of the area), represented as cumulative per year since 2008. The only period of intense seismic activity was between October 2011 and February 2012 with few earthquakes recorded since. Moreover, several sources of deflation were reported in the caldera between 2012 and 2017 (Papageorgiou et al., 2019).

Our geochemical data together with the seismic and ground deformation data show that no new magma upwelling has occurred since 2012. Indeed, the low CO₂/CH₄ ratios and non-atmospheric N₂ concentrations record an absence of input of deep gas (i.e., CO₂, N₂) in the system while the low H₂ concentrations suggest the lack of new significant thermal perturbations. Thus, the geochemical signature records the evolution of the magma emplaced during the last unrest.

Figure 5A shows the He isotopic composition of the fumarole gases plotted versus the ⁴He/²⁰Ne ratio. Most data plot on a mixing curve representing a binary admixture between air (or air saturated water) and a magmatic gas having He isotopic ratios between 3.3 and 3.9 R_a. This overlaps the He isotope ratios measured in phenocryst olivines from mafic enclaves (3.0–3.6 R_a) include in the 1,570–1,573 and 1925–1928 dacitic lavas of Nea Kameni (Rizzo et al., 2015) and implies that the fumaroles are directly tapping magmatic volatiles. This contrasts with the higher ³He/⁴He measured in submarine thermal springs from the northern part of the Santorini caldera (6.6 R_a) and at the Kolumbo seamounts (7.1 R_a) (Carey et al., 2013; Rizzo et al., 2016; Moreira et al., 2019) (Figure 5A). If the Kolumbo fluids truly represent the magmatic ³He/⁴He, the lower values of the Nea Kameni fluids imply there is a contribution from either ⁴He-rich fluid in the hydrothermal plumbing system or assimilation of metamorphic basement by the magma that feeds the Nea Kameni hydrothermal system (Rizzo et al., 2015).

CO₂/³He and δ¹³C-CO₂ are used to constraint the source of CO₂ (Marty and Jambon, 1987; Sano and Marty, 1995). CO₂/³He and δ¹³C-CO₂ seems to track a binary mixing between a limestone (CO₂/³He = 1x10¹³; δ¹³C-CO₂ = 0‰) and a mantle (CO₂/³He = 2x10⁹; δ¹³C-CO₂ = −4‰) endmember (Figure 5B). It could imply that the CO₂ emitted at Nea Kameni is largely derived from limestones either from the subduction of carbonate-rich sediments and/or assimilation of the carbonate basement (Rizzo et al., 2015). If we consider that the gases from Kolumbo and Santorini have a common mantle origin in terms of CO₂ and He, then the Kolumbo gases shows that the CO₂ prior to its dissolution in water has likely a more pronounced mantle signature (CO₂/³He = ∼1.26x10¹⁰; δ¹³C-CO₂ = ∼-0.4‰) than CO₂ discharged at Nea Kameni fumaroles (CO₂/³He > ∼1.84x10¹⁰), as with He, although with some chemical and δ¹³C-CO₂ fractionation (CO₂/³He = 9.71x10⁹-3.34x10¹⁰; δ¹³C-CO₂ = −0.87‰–0.80‰) that we assume to be due to gas-water interaction before emission (Rizzo et al., 2019) (Figure 5B).

Figure 6A shows the temporal evolution of He isotopic ratio corrected for atmospheric contamination (R_c/R_a) for the fumarolic gases of Nea Kameni. The He-isotope composition of volcanic gases records the ascent and degassing of new volatile-rich magmas (Caracausi et al., 2003; Rizzo et al., 2006; 2009; 2015; 2019; Nuccio et al., 2008; Boudoire et al., 2020; Torres-González et al., 2020; Sandoval-Velasquez et al., 2023). This phenomenon was observed at Santorini during the last unrest, when ³He/⁴He increased from 3.62 R_a in October 2007 to 3.94 R_a in January 2012 (Rizzo et al., 2015). After this event the He isotopic ratio has decreased albeit showing fluctuations (3.3–3.9 R_a). The origin of the variation is not clear but given the low and uniform CO₂/CH₄ ratios, H₂ concentrations and the lack of significant seismic activity inside the caldera (Figures 4A–C), we exclude the upwelling of new batches of magma.

The long-term variation of He concentration of the fumarolic gases of Nea Kameni is shown in Figure 6B. It was relatively low between 1988 and 2007, but increased from September 2009 to January 2012. In May 2012, He concentrations dropped back to lower values and since then gradually increased, until July 2022 when a slight drop occurred again. Given the low He content in the air (5.24 ppm), the increase in atmospheric contamination cannot explain the increase in He concentrations observed. This trend is anti-correlated with the CO₂/³He ratio (Figures 6B, C). For instance, during 2011–2012 the high He content and the low CO₂/³He were also accompanied by higher ³He/⁴He and CO₂/CH₄ values that appear to be consistent with the arrival of magmatic fluids (Figures 4A, 6A–C). As the increase of He concentration after 2011–2012 unrest is not linked to magma recharge, we suggest that the decoupling between He and CO₂ is due to a preferential dissolution of CO₂ as the magmatic gases rise to the surface through the aquifer (Rizzo et al., 2019). This can strongly impact the chemical composition of fluids (e.g., He concentration, CO₂/CH₄, CO₂/³He values) (Rizzo et al., 2019). In addition, the CO₂ dissolution may also fractionate the carbon isotope composition (δ¹³C-CO₂) at some pH and temperature conditions (Clark and Fritz, 1997).

Selective gas dissolution in water has been modeled at variable temperatures (25°C–374°C) using a similar approach as the one presented in Rizzo et al. (2019). We use this model to calculate the theorical composition curve of the residual CO₂ (CO₂/³He vs. δ¹³C-CO₂), once this gas preferentially dissolves in water. We have varied the boundary conditions, such as temperature, pH and initial bulk composition to fit the dissolution curves on our data and discuss the reasons of the temporal decrease of the CO₂/³He ratio (cf. section “Constraints on the hydrothermal temperature and variations of the gas-water equilibrium conditions”). We made this model by representing a condensation process under equilibrium condition. We have considered an open-system in which the gas-saturated parcels of water were continuously removed. The CO₂/³He ratio has been corrected by using Eq. 4 (Rayleigh, 1896): R v R v 0 = f α − 1 (4) where R v is the ratio of interest (i.e., He/CO₂), in the residual gas phase, R v 0 is the initial bulk ratio which has to be hypothesized (i.e., residual gas fraction = 1), f is the residual gas fraction that we vary from 0 to 1 and α is the fractionation factor determined by the solubility ratio between the two species considered (i.e., k H − C O 2 / k H − H e ). The Henry solubility constants were determined at high temperature by using the equations presented by Fernandez-Prini et al. (2003) and Harvey (1996). These equations require the calculation of the saturated vapor pressure which has been calculated by using the equation presented in Wagner and Pruss (1993). To transform the instantaneous, He/CO₂ in the residual gas phase to the CO₂/³He, we used the average R_c/R_a value recorded over our period of study (∼3.6 R_a).

As explained above, the CO₂ dissolution also affects the C-isotope composition in the residual gas phase. To correct for this effect, we used Eq. 5, presented in Clark and Fritz (1997): δ 13 C − C O 2 = δ 13 C − C O 2 0 + ε ⁡ ln f (5) where δ 13 C − C O 2 is the C-isotope composition in the residual gas phase, δ 13 C − C O 2 0 is the initial C isotopic ratio, ε is the fractionation factor between DIC and gaseous CO₂ and f is the same fraction of residual gas as the one presented previously. To calculate the fractionation factor ε we used Eq. 6, presented in Zhang and Quay, (1995) and Allègre, (2008): ε D I C − C O 2 g = H 2 C O 3 ε H 2 C O 3 − C O 2 g + H C O 3 − ε H C O 3 − − C O 2 g + C O 3 2 − ε C O 3 2 − − C O 2 g H 2 C O 3 + H C O 3 − + C O 3 2 − (6)

The isotopic fractionation between CO₂ (g) and CO₂ (aq) depends on the water temperature and on the molar fraction of the different dissolved organic carbon species (i.e., H₂CO₃, HCO₃ ⁻ and CO₃ ^2-), which is directly correlated to the pH of the solvent that we varied. Each single fractionation factor was determined at high temperature by using the equation presented in Zhang and Quay, (1995).

We used the geothermometric functions based on the CO/CO₂ and H₂/N₂ ratios (Tarchini et al., 2019 and references therein) to estimate the hydrothermal temperature over our survey period (Table 2). As CO was considered unreliable in our samples, we assumed a liquid phase and adopted the same FeO-FeO_1.5 hydrothermal gas buffer than previously used at Nea Kameni by Tarchini et al., 2019. The H₂/N₂ geothermometer track the ammonia dissociation and we therefore based our calculations on the concentrations obtained after atmospheric correction (Table 2). The results presented in Table 2 suggest a decrease of the hydrothermal temperature between 2015 and 2022 (i.e., from ∼260°C to ∼196°C).

We attempt to observe this temperature variations through the C-He isotope compositions. As discussed earlier, the decrease of the CO₂/³He ratio over time is related to the process of preferential CO₂ dissolution in water compared to He, rather than CO₂ and He source changes or increase of air contamination. We thus assume that the initial/pristine CO₂/³He and δ¹³C-CO₂ values do not change significantly over time, as present values reflect a low level of magmatic activity comparable to pre-2011 unrest. To construct our models of CO₂ dissolution presented in Figures 7A, B, we assume that this pristine CO₂ composition lies on a Limestone-MORB binary mixing curve (Sano and Marty, 1995). We use the hydrothermal temperature estimate (∼260°C in a liquid phase) determined by the CO/CO₂ and H₂/N₂ geothermometric functions in October 2012 to approximate the pH-dependent initial compositions (Figures 7A, B; Tarchini et al., 2019).

In the first model presented in Figure 7A, we use a constant pH of 5 (measured and assumed in the gas-water interaction model at Kolumbo; Rizzo et al., 2019) and vary the gas/water equilibrium temperature to fit the C-He isotope compositions measured in April 2016 and July 2019. The theoretical CO₂ dissolution curves are calculated for gas/water equilibrium temperatures of 260°C, 195°C, 100°C and 25°C. We highlight that at this pH value, the variability in the CO₂/³He and δ¹³C-CO₂ values could be explained by a temperature decrease through time. This temperature decrease favors the CO₂ dissolution, as shown by the decrease of the residual gas fraction over time. However, temperature estimates based on this model are unrealistic knowing that surface fumarolic gases are at least 90°C.

In the plot of Figure 7B we use a second model that considers a fixed temperature range between 260°C and 195°C, as estimated in 2012 and 2019 from the H₂/N₂ geothermometer. Assuming the different pH-dependent initial compositions, we calculate the corresponding CO₂ dissolution curves for pH values of 8, 6 and 4. The results show that for all the pH values in this range, the C-He isotope data fit with a possible decrease of the hydrothermal temperature. These temporal variations in gas/water equilibrium temperature could also be accompanied by a drop in the pH of the water but it seems unlikely over this period of few years. The plot in Figure 7B also highlights that there is no way of fitting the gas/water equilibrium temperatures in 2012 and 2019 with the estimates given by the H₂/N₂ geothermometer if we assume a pristine composition on the Limestone-MORB binary mixing curve (having a curvature coefficient R of 1). This may imply that: i) the H₂/N₂ geothermometer and the CO₂ dissolution model do not trace the same temperatures within the hydrothermal system, due, e.g., to a multistep vertically-elongated system; ii) the temporal increase in air contamination compromises the temperature estimations given by the H₂/N₂ geothermometer; iii) the composition of CO₂ before dissolution is characterized by a lower δ¹³C-CO₂ and a higher CO₂/³He than the values hypothesized in our models (cf. Figure 7B). The latter hypothesis could more easily explain some of the literature data, which seem to define a trend toward the organic sediment endmember (cf. Figure 5B). These C-He isotope compositions (obtained in 2013) could then be explained by slight variations in equilibrium conditions, rather than a source change.

The results obtained from the H₂/N₂ geothermometer and the CO₂ dissolution models predict decreases of the hydrothermal temperature over the period 2015–2022, which is consistent with the absence of recharge of the magmatic reservoir. It is nevertheless important to note that other parameters such as the equilibrium pressure, the salinity of the water or the gas-water volumetric ratio are not considered in the models. They may have an impact on the chemical and isotope compositions. However, there are no thermodynamic approaches that include these parameters for calculating the evolution of the δ¹³C-CO₂ during CO₂ dissolution at high temperature and it is thus impossible to quantify their effects.