Water cycling and fluid processes in subduction zones have been studied based on various observational, experimental and theoretical approaches. Stability fields of hydrous minerals in the subducting slabs, which have been constrained by the high-pressure experiments and thermodynamic estimates [1,2] are thought to primarily control the fluid supply at depth. Migration and re-distribution of the slabderived fluids (hereafter referred to as slab-fluid) has been predicted by numerical modeling [3-5]. Existence of such fluids has been examined by seismic studies along the subducting slabs and within the overlying mantle wedge [6,7], and by petrological studies of arc magmas that fingerprint the geochemical characteristics of slab-derived fluids [8,9]. Relatively well-constrained trace element behaviors upon slab dehydration [10-12], as well as heavy isotope tracers (such as Sr, Nd and Pb isotopic ratios) that are not fractionated through the processes of subduction and dehydration, have been utilized to successfully quantify the amount and composition of slab-fluids [13-15].

In addition to these constraints on the fluid processes in subduction zones from the deep side (i.e., dehydration of subducted slabs), several evidences have been accumulated from the surface side, including studies on fluid inclusions in volcanic phenocrysts and plutonmetamorphic rocks [16-18], and thermal waters and gases [19]. Within this context, the Arima-typebrine, which is a type of non-volcanic hot spring water with high chlorine content (~40000 ppm) and oxygenhydrogen isotopic ratios similar to magmatic/metamorphic thermal waters [20], has been argued to have originated from deep-seated brine [20-22]. The Arima-type brine may provide in valuable information concerning a slab-fluid in a fore-arc region where no magmatic information is available [23,24]. High ³He/⁴He ratios and large fault zones associated with the Arima-type brine suggest their deep origin, e.g., in the subducting slab and/or the mantle [25-27].

However, in order to discuss the origin of Arima-type brine, in particular its connection to the slab-fluid, better geochemical characterization is required, including Sr-Nd-Pb isotopic analyses which have not been done so far. This study aims mainly at developing an analytical method for the brine, including rare earth element abundances and ⁸⁷Sr/⁸⁶Sr, ¹⁴³Nd/¹⁴⁴Nd, ²⁰⁶Pb/²⁰⁴Pb, ²⁰⁷Pb/²⁰⁴Pb and ²⁰⁸Pb/²⁰⁴Pb ratios. High salinity and low abundances of the target elements and isotopes (except for Sr) in the brine require specific analytical procedures to be developed. Based on the newly obtained data, we then briefly discuss the relationship between the slab-fluid and the Arima-type brine.

The Arima hot springs, a typical locality of the Arima-type brine, are located in the southwest Japan arc (Figure 1a), where two oceanic plates, the Pacific plate and the Philippine Sea plate, subduct beneath the area from the east with a velocity of 9 cm/year for the former and from the southeast with a velocity of 4 cm/year for the latter. The depth of the slab surface beneath the Arima area is ~400 m for the Pacific slab, whereas 50 to 80 km with a large uncertainty for the Philippine Sea slab [28,29]. In spite of the active subduction, a Quaternary volcano is not formed here because the Pacific slab is too deep and the Philippine Sea slab is too shallow (Figure 1a) to fulfill the physiochemical conditions for arc magma generation [30].

Figure 1: Tectonic and geologic maps of the Arima area. (a) the map showing distribution of the Quaternary volcanoes (red circles) and geometry of the subducting Pacific and Philippine Sea slabs, with Itoigawa-Shizuoka Tectonic Line (ISTL) and Median Tectonic Line (MTL). The pinkish contour lines indicate the depth of the upper surface of Pacific slab (50 to 300 km depth with 50 km interval), while the purplish contour lines indicate that of Philippine Sea slab (10 to 200 km depth with 10 km interval). The aseismic parts are shown by the dotted line. (b) The geologic map around the Arima area after Maruyama et al. [31].

In southwest Japan, there are many faults striking in NW-SW or NW-SE slipping dextrally associated with the Median Tectonic Line [31]. The Arima-Takatsuki Tectonic Line (ATTL in Figures 1a and 1b) is one of these strike-slip faults that have formed in the late mid- Miocene [31]. The average slip rate for the eastern range-front segment of the tectonic line is estimated to be 0.5-1.5 mm/year dextrally and 0.1-0.8 mm/year vertically during the late Quaternary period [32]. As shown in the geological map of this area (Figure 1b), the basement around the Arima area is composed of late Cretaceous felsic volcanic rocks (rhyolite) of the Arima Group, granitic rocks (Rokko Granite), and late Eocene to early Oligocene non-marine sedimentary rocks with rhyolitic tuff layers of the Kobe Group [33]. The Arima Group directly covers the Rokko Granite in the south of ATTL and the sedimentary rhyolitic rocks in the north of ATTL in the Arima area [34].

The Arima-type brines occur typically in the Arima area, as well as the Osaka and Kii areas broadly in the Kinki region of southwest Japan. In the Arima area, the brines seem to upwell through ATTL and the subsidiary faults [35], with at least seven hot spring sources, i.e., Ariake, Gosha, Gokuraku, Kinsen, Tansan, Tenjin, and Uwanari. These hot spring waters exhibit a wide compositional range, which can be explained by mixing of two types of meteoric waters and a “deep brine” component [21,22]. The deep brine component is characterized by high solute concentrations (i.e., high Na, Cl, K, etc.), distinct from the meteoric waters. Based on the high salinity, we focus on the hot spring water from “Ginsuiso Kinsen” with ~40000 mg/L Cl from a 600 m depth pipe [23], which is composed primarily of the deep brine component, and analyze its composition for REEs and isotopic ratios as below.

Sample description

The “Kinsen” water sample, together with other spring waters in the Arima area, was collected in December 2010, directly from the well pipe before the water is pooled in an approximately 3×3×3 m tankto be oxidized and colored ‘gold (Kin-iro)’ for commercial use. Our sample water is visibly colorless, implying that it is less oxidized and has a lower propensity to precipitate the solutes. Kusuda et al. [23] have analyzed “Ginsuiso Kinsen” sample (which is the same brine with “Kinsen”, although the sampling date is different), 12 solute elements/ components and isotopic ratios of H, He, C, O: e.g., ~40000 mg/L Cl, ~20000 mg/L Na, 51.0 mg/L Li, δ18O=5.0‰, δD=-34.1‰, which is close to the estimated “deep brine” composition on the tritium-free basis [21-23].

Analytical method for rare earth element composition in high salinity brine

The high salinity and solute concentrations in the brine may disturb quantitative analyses of the rare earth elements due to the matrix effect. Moreover, in several water samples, there are some visible particles (or un dissolved materials). Accordingly, for determining the precise REEs composition, we have tested several preparation procedures as follows:

(1) raw and filtrated brine samples were first dried up, respectively. The condensed materials were chemically digested in the same way as the dissolution of silicate rocks after Yokoyama et al. [37] and then diluted with HNO3 suitable for ICP-MS analysis.

(2) raw and filtrated brine samples were centrifuged, and their supernatant liquids were pretreated straight for ICP-MS analysis; whereas, the precipitates were chemically digested in the same way as above, and then pretreated with HNO₃ for ICPMS analysis.

The ICP-MS analysis was conducted at the Department of Earth and Planetary Sciences, Tokyo Institute of Technology. In order to quantify the matrix effect due to high salinity, we synthesized thirty standard solution samples with the NaCl concentration from 0.0 to 5.0 wt.% with 0.5% increment, and the REEs concentration of 0, 0.1, 1 ppb by adding the standard solution XSTC-1 (SPEX CertiPrep Co. Ltd.). All the combination of NaCl and REEs abundances have been analyzed by ICP-MS under the same tuning condition. The results show linear correlations between the REEs concentration and the signal intensity (in count per second (cps) unit) for each NaCl concentration from 0 to 5% (Figure 2). The results also show that the intensity drastically decreases when the NaCl concentration exceeds 1% (Figure 2). Accordingly, we have diluted the unknown samples of 0.2 mL with 1.7 mL pure water, and have applied the standard addition method with addition of XSTC- 1 (0.1 ml) by 0, 0.1, 1 and 10 ppb to the diluted sample. As a result, the final diluted sample contained ~0.6 wt% NaCl.

Figure 2: Matrix effect of brine in ICP-MS measurement. Intensity (count per second) of ICP-MS measurement of La in synthetic standard solutions (0, 0.1 and 10 ppm) as a function of NaCl content (0.0 - 5.0% with 0.5% interval). The inset shows a magnified view near the origin.

In Figure 2, even for a specific NaCl content, a slight (less than ~5%) deviation from the perfect linear relation is recognized, indicating a complex matrix behavior that may also depend on the concentration of the target element. We therefore have performed ICP-MS analysis with the standard addition method using three combinations of XSTC- 1 concentrations ((I) 0-1 ppb, (II) 0.1-1 ppb and (III) 0.1-10 ppb). As will be shown, the results show good reproducibility, confirming a weak nonlinear matrix effect when NaCl content is relatively low. It should be noted that Eu was not quantitatively measured due to significant interference from Ba oxide, which unfortunately prevents us from discussing the Eu anomaly.

Heavy isotopic composition in high salinity brine

We have analyzed the isotopic compositions of Sr, Nd and Pb (⁸⁷Sr/⁸⁶Sr, ¹⁴³Nd/¹⁴⁴Nd, ²⁰⁶Pb/²⁰⁴Pb, ²⁰⁷Pb/²⁰⁴Pb and ²⁰⁸Pb/²⁰⁴Pb ratios) for the “Kinsen” brine. The target elements, i.e., Sr, Nd and Pb, have been extracted and separated from the brine sample by a co-precipitation method with iron (III) hydroxide (Figure 3). First, a sufficient quantity of highly purified and concentrated HNO3 is added to the sample to dissolve, any existing precipitates (Figure 3a). Then, highly concentrated aqueous ammonia is added to neutralize the sample. When pH is between 7 and 8, some visible but dispersed precipitates occur, after 24 hours which form a deposit that contains rare earth elements (Figures 3b and 3c). The recovery percentage of this method is more than 95% [36]. We have successfully obtained approximately 0.3 g precipitate from 500 ml brine. We dissolve the precipitate with HNO3 (7M) (Figure 3d) [37] from which Sr, REEs and Pb are extracted by column separation (Figure 4). The extracted Sr, Nd and Pb were finally analyzed by MCICP- MS (IsoProbe, GV instruments) installed at Earthquake Research Institute, The University of Tokyo.

Figure 3: Summary of pretreatment method for isotope analysis. (a) a brine sample added with concentrated original HNO₃ solution to dissolve all ingredients. (b) minor visible and dispersed precipitation occur after adding ammonia solution at pH 7-8. (c) the dried-up sample of the precipitate of (b). (d) the final solution.

Figure 4: Flow chart of separation method for isotope analysis.

REEs pattern of Arima brine

The results are listed in Table 1 and plotted in Figure 5a as compositional patterns normalized by depleted MORB mantle (DMM). The REEs abundances of the “Kinsen” brine (three lines with bluish colors) are approximately three times lower than those of DMM, and show a flat pattern with a slight increase to both the left for light REEs and the right for heavy REEs. The results are fairly consistent with those from the previous study, shown as a black line in Figure 5a [38], in terms of the overall level and a slight increase towards the left for light REEs, although the previous result shows a slight decrease from middle to heavy REEs.

Table 1: Rare earth element abundances of the Arima hot spring water “Kinsen” brine are listed in the “Analyzed abundance” row. Analytical precisions for all elements are below 2 % (2SD). The compositions of AOC-derived fluids and sediment-derived fluids are calculated forwardly starting from the compositions of raw materials before subduction for Pacific and Philippine Sea plates, respectively, as shown in the “Calculated abundance” row [39]. The estimated compositions of fluids derived from subducting slab at low temperatures for Philippine Sea plate, are shown in the “Estimated abundance” row. The compiled data used in this study are also listed in the “Compiled data” row. The optimized data to the analyzed abundance by a mixing among slab-derived fluid and river water are listed in the “Mixed fluid” row.

Figure 5: DMM (depleted MORB mantle)-408 normalized REE compositions. (a) Arima hot spring waters (blue lines, this study; black line from Tsuji et al. [38]. Blue lines correspond to three combinations of XSTC-1 in the standard addition method ((I) 0-1 ppb, (II) 0.1-1 ppb and (III) 0.1-10 ppb). (b) AOC-derived fluid beneath volcanic arc [39] and the estimated composition of fluid derived from eclogite at low temperature (lower five lines). The partition coefficients are obtained by extrapolating those of high-temperature experiments at 700-1000 ºC with 4 GPa [12,40]. (c) sediment-derived fluid beneath volcanic arc (top line) and the estimated composition of fluid derived from sediment at low temperatures (lower five lines). (d) river water near the studied area [41].

We now compare the results with the REEs abundances of possible “deep brine”, in particular the slab-fluids, and examine whether the slab-fluids could be a source of deep brine. Figures 5b and 5c show estimated REEs compositions of aqueous fluids dehydrated from the major constituent materials of a subducting slab: i.e., altered oceanic basaltic crust (AOC) and sediment. In Figure 5b, the REEs abundances of AOC-derived fluids (“PHS” derived from the Philippine Sea slab and “PAC” from the Pacific slab) beneath volcanic arc exhibit high concentrations [39], whereas those dehydrated at lower temperatures beneath the fore-arc region exhibit significantly lower abundances due to the large temperature dependence of partition coefficient between fluid and residual solid [12,15,40].

Likewise, the estimated REEs compositions of the sedimentderived fluids beneath the volcanic region exhibit high concentrations (Figure 5c) [39]. However, unlike the AOC system, the temperature dependence of partition coefficients between fluid and sediment is not well constrained. Therefore, we assume that, with a temperature decrease, the same degree of REEs concentration reduction occurs in the sediment system (Figure 5c) as in the basaltic system (Figure 5b). This is a crude assumption, and these results (shown by dotted lines in Figure 5c) should be regarded as preliminary estimates.

In any case, deep brine must mix with the near-surface water, as evidenced by δ18O- δD- tritium systematics [20-22]. In Figure 6, we estimate a best fit composition of fluid that mixes a slab-fluid (that consists of a mixture of AOC-derived fluid (Figure 5b) and sediment-derived fluid (Figure 5c) and the river waters in the Rokko Mountains (Figure 5d) [41]. The preliminary result suggests that the REEs abundances of the “Kinsen” sample (Figure 5a) may be broadly explained by the mixing of approximately 81-92% fluid derived from the Philippine Sea slab (at 400ºC) or 56% at 510ºC (Figure 6), with the addition of river water, being consistent with the previous studies based on the δ_{18O- δD- tritium systematics, although the misfit is rather large at this stage.}

Figure 6: DMM (depleted MORB mantle)-normalized REE compositions of calculated mixed fluid between slab-fluid derived from the Philippine Sea slab and the river water in Figure 5.

Heavy isotopic composition of Arima brine considered in arc systems

The results of Sr-Nd-Pb isotopic analyses are listed in Table 2. In addition to the “Kinsen” brine, “Tenjin” and “Tansan” have been also analyzed for examining heterogeneity among the hot spring sources in the Arima area. In order to test the slab-origin hypothesis, the isotopic compositions are compared with the isotopic compositions of the slabfluids (PHS-fluid and PAC-fluid), as well as the subducted materials of AOC and sediment (Figure 7). The “Kinsen” data are plotted in the range of slab-fluid derived from the Philippine Sea slab (PHS-fluid), being consistent with the REEs arguments in the last section. It is noted that the PHS-fluid is mainly derived from sediments [9] and therefore resembles the sediment in isotopic composition.

Table 2: Sr-Nd-Pb isotopic compositions of the Arima hot spring waters from the “Kinsen”, “Tenjin” and “Tansan” hot spring sources. The Nd isotopic ratio of “Tenjin” was not obtained. Other data are compiled from Nakamura et al, [14].

While the ²⁰⁷Pb/²⁰⁴Pb vs. ¹⁴³Nd/¹⁴⁴Nd diagram is sufficiently sensitive to resolve the type of slab-fluids (Figure 7), the Sr and Nd isotopic diagram is useful to discuss the contribution from the crustal rocks, especially the granitoids that form the basement of the Arima area. Figure 8 shows the Sr-Nd isotopic compositions of the basement granitoidsin the southwest Japan arc [42,43], as well as those of slabfluids (PHS-fluid and PAC-fluid) and the subducted AOC and sediment. The “Kinsen” data appear to be plotted near the PHS-fluid composition. At the same time, the data are included in the broad compositional range of the basement granitoids, although the basement granitoids in areas adjacent to Arima (box coded in dark purple in Figure 8) deviate from the “Kinsen” data, and we may not identify the unique source material of “Kinsen” in this case. However, considering the high ³He/⁴He ratio of the Arima brine [27], it is likely that the Arima brine is related to the PHS-fluid deep in the mantle and the subducted slab.

Figure 7: Pb-Nd isotopic composition of the Arima hot spring waters (red crosses) compared with the volcanic rocks in central Japan of Nakamura et al., [14]. Green and orange bands with black line represent the mixing of slabderived fluid with the mantle: green is for the Pacific slab (PAC), and orange for the Philippine Sea slab (PHS). The ellipsoid represents the compositional range for each end-component. Small circles represent the Quaternary volcanic rocks in central Japan for comparison, which can be explained by the mixing between slab-derived fluids (PAC- and PHS-fluids) and the mantle (DMM).

Figure 8: Sr-Nd isotopic compositions of the Arima hot spring waters compared with the volcanic rocks in central Japan of Nakamura and et al. [14]. Square boxes represent the compositional ranges of granitoid: light violet box represents that of granitoids broadly exposed in southwest Japan, whereas dark violet box represents those exposed near the Arima area. Other symbols are the same with those in Figure 7.

The new analytical procedures for REEs and Sr-Nd-Pb isotopic compositions have been presented in this study, based on which typical Arima-type brines have been analyzed. The most plausible interpretation for both the REEs abundance and the isotopic ratios is that the deep brine has been derived from a subducted Philippine Sea slab, being consistent with the oxygen, hydrogen and helium isotopic compositions. This may indicate that slab-fluids may ascend in the fore-arc region, in addition to the volcanic region where the slab-fluid triggers arc magmatism, possibly through a large fault zone along the tectonic line. Several Arima-type brines with high ³He/⁴He ratios in other areas along the Median Tectonic Line in southwest Japan (e.g., Kazahaya et al. this special issue) may also suggest a similar origin, which provides invaluable information on slabfluid in non-volcanic regions.

We would like to thank O. Ishizuka, K. Fujinaga, T. Ishikawa, M. Tanimizu and H. Sakuma for their help at various stages of this study, and M. Totani, the president of Arima Hot Springs Tourism Association, for permission and help with our field work and sampling. This work was partly supported by the Cooperative Research Program of Earthquake Research Institute, The University of Tokyo.