The WHATS-3 system consists of five components, including (1) a fluid inlet (so-called “pistol”), (2) four sets of dual-chambered sampling bottles, (3) a valve-manipulating system (Figure 1B), (4) an impeller pump unit, and (5) an electric control unit to operate the valve-manipulating digits and the impeller pump. The main body (including the sampling bottles and the manipulating system) is a titanium frame, semi-cylindrical in shape, with a dimension (H × W × D) of 700 × 375 × 395 (mm) (Figure 1A). A rectangular face 700 × 375 (mm) in size is required for tight attachment onto submersible vehicles; WHATS-2 had a larger attachment face of 595 × 560 (mm) (Saegusa et al., 2006). So far, the WHATS-3 has been attached on the payload space behind the right manipulator of the ROV Hyper-Dolphin, the rear payload space of ROV Kaiko (vehicle Mk-IV), and between the two manipulators of the Human Occupied Vehicle (HOV) Shinkai 6500 (Figure 2). Unlike most other fluid samplers including the WHATS-2, the pistol, the impeller pump unit, and the control unit are isolated from the main body. This is beneficial as it reduces the space required in the main payload space, usually in the front of submersible vehicles.

The pistol consists of a T-shaped titanium tube (O.D. 10 mm, I.D. 9 mm) and a thermometer (Pt 100Ω) blanketed with a steel shell. The thermometer, inserted from right side of the T tube through to the fluid inlet on the left and sealed with double silicon O-rings at the insertion end, is responsible for real-time monitoring of the fluid temperature during sampling, with a precision of ±1°C. Incoming fluid from the inlet on the left side flows to bottom end of the T tube, completely flushing any possible residues at the right side. For sampling of diffusive fluids, a funnel (>300 mL) is mounted on the inlet end as a reservoir of the geofluid to avoid mixing and contamination by ambient seawater. The impeller pump unit contains, in addition to the pump, an electronic motor contained within a pressure-resistance chamber sealed against a maximum pressure of 40 MPa. The pumping speed can be adjusted at three steps: ~400, ~200, and ~100 mL/min. In general, a slower pumping speed better avoids contamination of ambient seawater.

Fluid flow pathways of the pistol, the sampling bottles, and the impeller pump unit are connected in this order using polypropylene tubes (PP-PS-M, AS ONE, Osaka, Japan) and T-shaped connections (PP-PT-M, AS ONE, Osaka, Japan). These can also be made with Teflon, if necessary. The length of the plastic tube depends on attachment locations of each component, but is typically ~5 m. The inner volume of a 5 m-long tube is ~100 mL. The impeller pump unit is located at the most downstream part of the fluid pathway in order to avoid contamination from potential exfoliation of the tube in the pump.

Each sampling bottle consists of two ball valves (SS-4SKPS4, Swagelok, Ohio, US) mounted on the two ends of the bottle, two spherical fluid chambers, and two ball valves (SS-4SKPS4) located between the chambers in a series (see Figure 1C). The SS-4SKPS4 valve has a gas-tight capacity for pressure difference up to 41.3 MPa (guaranteed by Swagelok), the same valve has been used without problem in the former WHATS-series samplers. The bottle is of a stainless-steel construct, the surface is treated with a SilicoNert coating (Silcotek Corporation, Bellefonte, Pennsylvania, USA) to avoid corrosion and metal contamination (particularly Fe, Cr, and Ni). Blank of H2 from the stainless-steel bottle has been examined and confirmed to be as low as 0.03 μM (Kawagucci et al., 2010). As in the WHATS-2, specific gingko-leaf-shaped discs are mounted as handles of the outside valves (see Saegusa et al., 2006 for a detailed description; Figure 1C). The two spherical fluid chambers differ in volumes (20 and 50 mL) but they may be attached in either order. When sampling, the two fluid chambers are connected and function as a single fluid container. When the WHATS-3 is recovered on-board, the two chambers are manually isolated using inner valves. This dual-chamber system allows separation of a single fluid sample into two discrete subsamples, each to be used for chemical analyses of gas or liquid.

The construction of the valve manipulation system is newly designed for rapid and smooth operation, as well as increasing the robustness of the system (Figure 1B). As a common feature of the WHATS-series samplers, the handles mounted outside ball valves of the sampling bottles are operated by valve-manipulating digits. Specifically, the digits move horizontally at a right angle against the valves, rotating the valve handles for opening and/or closing them (see Saegusa et al., 2006 for detailed description). In the WHATS-2, the four sampling bottles were aligned on a plane, and the narrow and elongate valve manipulating arm (with one and two digits on each end) was mounted on a small block moving linearly by means of a rotating screw pole powered by an electronic motor. The delicate shape of the arm was vulnerable to external forces. Indeed, it was sometimes bent due to accidents during routine maintenance on-board, resulting in errors manipulating the valves in situ. The force transformation to linear movement was time-consuming with more than 1 min being required to manipulate the valves of a bottle. In addition, the sampling bottles were attached on the main body using stainless-steel tubes that also functioned as a fluid path. These tubes were sometimes bent or broken due to the sheer weight of the bottles when they were attached or detached. To resolve these issues, the WHATS-3 adopted a new semi-circular alignment with the sampling bottles set in 60° intervals (see Figure 1B). The valve-manipulating digits are mounted directly on blocks and rotate on a fixed axis via a roller-chain powered by an electronic motor. The sampling bottles are held on the built-in ports of the main frame, independent from tubes that make up the fluid path. The sturdy construction and the simplified geometry of valve manipulation in the WHATS-3 reduce potentials of inadvertent failures during operation, and the time required for manipulating the valves of a bottle is reduced to just a few seconds.

A power supply of DC24V is independently required for: (1) main electric system for valve manipulation, fluid pumping, and signal communication with the submersible vehicle (Figure 3) and (2) the thermometer. Cables and connectors used are made by SEACON (El Cajon, CA, US) and/or IMPULSE (San Diego, CA, US). The operating software for the main system and the thermometer was developed for Microsoft Windows operating systems and has been confirmed to function normally and successfully on Windows XP and Windows 7. The main functions of the software are: (1) to switch pumping ON/OFF, (2) to change the speed of pumping, (3) to operate the manipulator for opening/closing of the valves, and (4) to display and record the fluid temperature.

The WHATS-3 system is designed to resist a pressure of 35 MPa, corresponding to 3,500 m water depth, with a safety coefficient of 1.7. It has been successfully operated under 33 MPa in deep-sea field testing. The 35 MPa resistance covers ~50% of the global seafloor, including most of Mid-Ocean Ridges, Arc, and Back Arc where hydrothermal activity is expected (German and Seyfried, 2014; Beaulieu et al., 2015; Ishibashi et al., 2015). The three sealed chambers (the pump unit, valve-manipulating motor, and electronic unit) are designed to resist an external pressure of 40 MPa while the sampling bottles withstand 30 MPa. The actual external high-pressure resistance of the whole system was examined in a laboratory pressure bath and confirmed to match the designed levels. Successful operations of the valves and manipulator were confirmed under 32 MPa external pressure in the laboratory.

For the sampling bottles, in reality, resistance against external high pressure is not important because all sampling bottles are open to ambient seawater during the descent to the seafloor (see Saegusa et al., 2006 for detailed description). Rather, resistance against internal high pressure post-sampling is more crucial for successful gas-tight fluid sampling because the pressure sealed sampling bottles, containing sampled fluids with dissolved gases, must be recovered intact from the deep-sea. The actual internal-high pressure resistance was examined by connecting a hydraulic pump (TP-500, Syn Corporation, Kyo-tanabe, Japan) and a digital pressure gage (VSMC series, VALCOM, Osaka, Japan) to the two ends of a sampling bottle. When pure water was filled at 40 MPa and stored for 22 h at room temperature, the sampling bottle did not show any visible leakage. Furthermore, when the sampling bottle was filled with pure water at 35 MPa and 2°C (mimicking the deep-sea environment), closed, and heated up to 20°C, as expected at the sea surface and during the recovery, the internal pressure increased to 38 MPa but there was again no visible leakage. These examinations demonstrate that the newly designed sampling bottle is suitable for use under deep-sea condition at a depth of 3,500 m.

Sufficient substitution of the pre-filled fluid in the bottle (usually pure water) with the target geofluid was examined because unlike the simple rod-shaped sampling bottles in the WHATS-2, the relatively complex dual spherical chamber design of the sampling bottles in the WHATS-3 may make smooth and complete substitution difficult. In the WHATS-series samplers, the sampling bottles are positioned vertically (Figure 1) while the more dense seawater is introduced from the bottom for efficient substitution. During testing, pure water and seawater were used as the pre-filled fluid and the target fluid, respectively (Figure 4). Chloride concentrations of the fluid in the bottle were determined by ion chromatography (IC) (Dionex ICS-2100, Thermo Fisher Scientific) after 2,500 times dilution, within analytical errors of 2%. Substitution, represented by relative Cl concentration against seawater, was analyzed in relation to the amount of fluid pumping. The analyses were conducted in all three pumping speeds and both possible attachment orders of the two spherical chambers (large-small chamber and vice versa). The examination results proved that >98% substitution was achieved with 300 mL pumping under all settings of the pumping speed and both chamber attachment orders (Figure 4). Here, more than 98% substitution practically indicates complete substitution, when the 2% analytical errors is taken into consideration. In actual field usage, more than 600 mL of fluid is generally pumped to provide extra assurance of a complete substitution. The required time for pumping 600 mL was only 1.5, 3, and 6 min at high, mid, and slow speed pumping, respectively.

At the beginning of a cruise, the main body, the pump unit, and the control unit of the WHATS-3 are mounted on the submersible vehicle to confirm electronic power supply and communication. Just before the actual dive, the four sets of sampling bottles and plastic tubes are connected and filled by water, as any internal air bubbles may cause disastrous crushing of the bottles and tubes during descent. At the same time, tubes and connections are carefully checked for any fluid leakage. As the vehicle approaches the seafloor, the first sampling bottle is opened, and fluid pumping is started to sufficiently flush and substitute the pre-filled water in the fluid pathway with ambient seawater (Figure 5A). In the case of high-temperature hydrothermal fluid sampling, fluid temperature monitored in real-time gives an indication of the relative position of the pistol inlet with regards to fluid effluent. Fluid collection is undertaken at the highest possible temperature in order to collect as close to pure geofluid as possible. After successfully collecting geofluids into the last (fourth) sampling bottle, all external valves are closed to avoid fluid leakage from sampling bottles. When the vehicle is recovered on-board, all sample bottles are detached from the main body, and then all the internal valves are closed manually to separate the two chambers of a bottle (Figure 5B). The bottles are stored in a refrigerator prior to subsequent subsampling, as heating of the bottles increase the internal pressure and therefore risks leakage and explosion. After subsampling (see below), each of the fluid-path tubes (usually remains on the vehicle) and sampling bottles are sufficiently flushed with tap water to remove any remaining fluids and particles, and then rinsed with pure water prior to the subsequent dive.

The two chambers in a sampling bottle serves as subsamples for liquid and gas analyses. Subsampling procedure for the WHATS-3 is the same as the WHATS-2, as presented elsewhere (e.g., Kawagucci et al., 2016). Although the current sampling chamber volumes of 20 and 50 mL in the WHATS-3 is less than the previous versions of WHATS (150 mL), these are already sufficient for subsequent analyses, described as follows.

For liquid subsampling (Figure 5C), a ~5 mL aliquot of the fluid sample is used without filtration for on-board analyses of pH, alkalinity, H₂S, and NH₃ concentration. The remaining fluid in the bottle (either 15 or 45 mL) is filtered through a 0.2 μm pore sized filter and used for major cation and anion analyses with Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP-AES) and IC. Both ICP-AES and IC analyses generally consume ~0.1 mL of the sample because the sample is subjected to >1,000 times dilution with pure water prior to the analyses. The Cl concentration is sometimes quantified by the Mohr titration method without dilution (as it is more precise than IC) which consumes ~1 mL of the filtered subsample. For gas subsampling (Figure 5D), a chamber is connected to a vacuum line typically consisting of a glass-constructed gas extraction portion, subsampling bottles, pressure gauges, and a vacuum pump. Entire body of fluid in a chamber is introduced into the extraction portion and mixed with solid reagents, sulfamic acid (H₃NSO₃) and cadmium acetate [Cd(CH₃COO)₂, respectively, to extract inorganic carbon as CO₂ and precipitate H₂S as sulfide mineral. Subsampling bottles are optional, although 100 mL glass ampule, 50 mL stainless-steel bottle, and 2 L glass canister are usually used for general, He, and ¹⁴C analyses, respectively. The total gas content of a fluid sample is determined barometrically in the vacuum line. The volume of the vacuum line is adjusted to maintain an internal pressure not exceeding atmospheric pressure (0.1 MPa), because internal high pressure may cause gas leakage or explosion of the line. For example, a fluid sample collected in the larger chamber (containing 50 mL fluid) and expected to have a total gas content of 200 mM (likely for example in the hydrothermal fluids from the Okinawa Trough vents (Kawagucci, 2015)] requires a volume more than 250 mL in the vacuum line. For safety reasons, when processing the first sample from a new geofluid site where the total gas content is unknown, a 2 L glass canister is applied to the line to ensure sufficient volume is achieved. Although the small chamber of WHATS-3 is only one-tenth in volume compared to many other gas-tight samplers (Seewald et al., 2001; Saegusa et al., 2006), the amount of recovered major gases (CO₂, CH₄, H₂, etc.) are usually sufficient to quantify their concentrations and isotope compositions using commercially available detectors via gas chromatography (e.g., Helium Ionization Detector; VICI, Houston) and continuous-flow isotope-ratio mass spectrometry (e.g., Proskurowski et al., 2006; Reeves et al., 2011).

Pressure resistance and field operational feasibility of the WHATS-3 system were tested and confirmed at Kairei (2,450 m) and Edmond (3,300 m) hydrothermal fields in the Indian Ocean (Gamo et al., 2001; Van Dover et al., 2001). The whole system operated under a maximum external pressure of 33 MPa and experienced no operational or visible damage when inspected closely after recovery. Fluids collected from black smoker vents in each field were completely depleted in Mg (Table 2), suggesting successful samplings of “pure” hydrothermal fluids (Von Damm et al., 1985). Their fluid chemistry, including gas composition, showed good agreement with previously reported values (Kumagai et al., 2008; Kawagucci et al., 2016), despite limited sample numbers.

Furthermore, the WHTAS-3 was applied together with previous versions of the WHATS-series samplers [WHATS-2 and Cheap-WHATS (Kawagucci et al., 2016)], to determine the fluid chemistry of three recently discovered deep-sea hydrothermal sites in the mid-Okinawa Trough. These include Aki site in the Iheya North field (Kasaya et al., 2015) and Hitoshi and pseudo-Noho (pNoho) sites in the Sakai field (Nakamura et al., 2015).

The “Mg diagram” (Von Damm et al., 1985) for major ion species in each site demonstrated bimodal mixing trends between a hydrothermal fluid and seawater, regardless of the samplers used (Table 2). Gas concentrations show scattering on the Mg diagram, similar to observations from other Okinawa Trough hydrothermal sites reported in previous studies (Konno et al., 2006; Kawagucci et al., 2013a; Toki et al., 2016). This is because the Mg concentration is

an indicator for bimodal mixing between high-temperature fluid and seawater, but subcritical phase separation of CO₂-enriched fluids results in the formation of bubbles and CO₂ droplets (Sakai et al., 1990; Konno et al., 2006), serving as a third endmember in the samples collected. Erroneously high and variable gas concentrations, even in high-Mg samples, point to the incorporation of bubbles together with vent fluid and seawater because ambient seawater contains only ~2 mM of CO₂ and negligible amounts of CH₄ and H₂ (see Discussion in Kawagucci et al., 2013a). Iron concentrations also show scattering on the Mg diagram. Similarly high Fe level in samples from both WHATS samplers and Niskin bottles indicate possible Fe contamination occurring during on-board or on-shore sample processing, although the stainless steel used in the WHATS-3 and the deep-diving vehicles cannot be entirely excluded as potential sources of contamination.

The estimated endmember fluid composition of the Aki site is generally identical to those of the Original and Natsu sites, the other two sites located within the Iheya North field (Table 3). The identical major fluid chemistry implies that the three sites are connected through a subseafloor fluid reservoir, despite the three sites being separated at ~1 km horizontal intervals in the North-South direction. This possible connection among the three sites will be examined by future seafloor drilling operations; previous drilling operations around the Original site has already revealed a subseafloor high-temperature fluid reservoir extending ~500 m in the East-West direction (Takai et al., 2012; Kawagucci et al., 2013b). On the other hand, CH₄ concentrations and carbon isotope ratios are notably different among the three sites. This can be explained by a higher contribution of ¹³C-depleted biogenic CH₄ incorporation into the venting fluid at the Original site.

The fluid chemistry of the pNoho and Hitoshi sites are distinct from each other (Table 3). The uniformly higher major ion concentrations in the pNoho site, compared to Hitoshi and CLAM sites also in the Sakai field (Gamo, 1995), are likely caused by the greater extent of vapor-lost phase formed through subseafloor fluid boiling. Such intra-field variation of fluid chemistry has also been confirmed in the Izena Hole and Daiyon-Yonaguni fields, Okinawa Trough (Suzuki et al., 2008; Ishibashi et al., 2014) and is considered to be a result of preferable emanation of the more mobile vapor-rich phase. Both maximum fluid temperature and endmember Cl concentration were highest at the pNoho site, suggesting that this is the center of activity in the Sakai field. Possible subseafloor connections among the three sites within the Sakai field will also be examined by a future drilling project.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.