The self-potential (SP) method has a long history on land and plays an important role in the delineation and resource evaluation of metal sulfide ore bodies (Fox, 1830). It is generally believed that the negative anomaly of SP is related to the metal sulfide ore bodies (Sato and Mooney, 1960; Corry, 1985; Naudet and Revil, 2005; Komori et al., 2017). In view of the great difference between the marine and the land environments, geophysical methods are still not economical and efficient enough, which limits the exploration of marine resources. With the development of electronic instruments and better understanding of the mechanism of SP response, the application of SP exploration has been gradually developed from land prospecting to shallow water and deep-sea exploration of polymetallic sulfide (Corwin, 1973; Corwin, 1976; Tao et al., 2013). In addition to mineral exploration, the underground water flow driven by seafloor thermal activity and thermal gradients will also produce detectable SP anomalies. The SP measurement method has also been applied to the study of geothermal and hydrothermal activity of the deep seafloor (Heinson, 1999). Eppelbaum (2019) introduced a new parameter into SP interpretation, i.e., the “self-potential moment”, and this method has been effectively applied at several ore deposits in the southern Caucasus.

In 1973, Corwin (1973) began to develop the marine SP detection system, and in 1976 an SP anomaly around 300 mV was discovered that was related to the offshore extensions of sulfide deposits (Corwin, 1976). In 2000, Sudarikov and Roumiantsev conducted SP and Eh (electric potential half cells, oxidation reduction potential) survey at the Logatchev hydrothermal vent in the Mid-Atlantic Ridge and inferred the spatial distribution characteristics of the hydrothermal plume near the vent (Sudarikov and Roumiantsev, 2000). Cherkashev et al. (2013) used a deep-sea towed SP instrument to locate seafloor sulfide deposits associated with hydrothermal vents near the Mid-Atlantic Ridge. Kawada and Kasaya (2017, 2018) also observed negative SP anomalies and associated hydrothermal sulfide deposits in the Izena hydrothermal field of the Okinawa Trough in Japan by using the deep-sea towed SP array. Safipour et al. (2017) mounted SP electrodes on transient electromagnetic equipment and detected negative SP anomaly over inactive sulfide in the Tyrrhenian Sea.

The configuration described above connects the electrode array via a cable or insulated elastic rod and is easy to operate in a marine environment. However, there are also several inconveniences relating to soft connected towed SP arrays. The towed array is susceptible to being distorted by ocean currents, and the distance between the soft connected electrodes changes greatly. Both of these factors will affect the precision of the measured SP amplitude, especially in mid-ocean ridges, where the seafloor topography varies greatly (Constable et al., 2018).

To improve the stability and efficiency of marine SP configuration used in the deep-sea environment, combined with the ideas of Constable et al. (2018), the autonomous underwater vehicle (AUV) was modified with four channel electrical field sensors in its tail to detect the marine SP responses (hereafter referred to as an AUV-SP), which was expected to be helpful for seafloor sulfide exploration. To test the stability and determine the influencing factors, two AUV experiments were conducted during July 2019 at Qiandao Lake in eastern China. The water depth was about 40 m, and there was little artificial interference around the test site, making it an ideal place to test the new system and analyze the electrical field noise. The main purposes of the AUV-SP system test were (i) to investigate the optimized position of SP electrode on the AUV to minimize the interference of noises from the AUV and (ii) to verify the developed system of the developed system by using the known artificial current source. In addition, measured multicomponent data were inverted using self-potential tomography (Jardani et al., 2008; Revil et al., 2008; Rittgers et al., 2013) to verify the capability of multicomponent self-potential detection.

The SP measurement system mainly consists of two parts, i.e., electrical field sensors (Ag/AgCl non-polarizing electrodes) and a data logger chamber. The electronic circuit of the data logger was encapsulated in the pressure chamber, which was mounted on the back of an AUV with a diving capacity of 4500 m. The electrical field sensors and two orthogonal extension rods of 2 m were arranged horizontally and vertically at the tail of the AUV (Fig. 1). Electrodes 1 and 2 were fixed at the end of the horizontal extension rod along the moving direction of the AUV to measure the inline component of the electrical field ER. Electrodes 5 and 6 were attached on the end of vertical extension rod to measure the vertical component EV. Electrodes 3 and 4 were mounted on the left and right sides of the abdomen of the AUV to measure the horizontal component orthogonal to the AUV ET. The ET data for the electrodes on the surface of the AUV were affected by the propulsion motor and low electrode space (less than 30 cm), leading a larger noise level than the rear electrodes. The component ET was greatly distorted; thus, the following data processing and analysis will focus on ER and EV.

The installation of the SP receivers did not affect the data flow or function of other sensors mounted on the AUV, including the magnetic field sensor, sonar, underwater cameras and other plume-detecting sensors. Navigation system consists ultra-short baseline, depth sensor and attitude orientation (Wu et al., 2019).

The forward problem of SP can be expressed by a Poisson equation of potential (Jardani et al., 2008): 3∇⋅σ∇V=∇⋅Js=ℑE=-∇V, where ℑ is the volumetric current density (A m-3), Js is the current density vector (A m-2), σ is the conductivity (S m-1), E is the electrical field vector and V represents the SP field.

By changing the source term of Eq. (3), the SP response to various effects can be calculated, such as the flow potential related to groundwater flow (Ahmed et al., 2013) or the geobattery model related to polymetallic deposits (Rittgers et al., 2013).

In order to investigate the reliability and interpretability of electrical field data measured by the AUV-SP system, the regularized least-squares method was applied for the inversion of the artificial SP source. Assuming that m is the bulk current density model vector and m0 is the reference model, the inversion can be expressed as the minimization of the following objective function: 4ϕ(m)=WdΔd2+αW(m-m0)2, where Wd is the data-weighting matrix and Δd is the difference between the data d obtained by forward modeling and the observed data dobs, also known as data residual. Here, the measured electrical field component is selected as the observation data, W is the model-weighted diagonal matrix, and α is the regularization factor. The regularization factor α was used to balance the model constraints and data-fitting function in the process of solving the model. It can be determined by the L-curve criterion (or generalized cross-validation method) or by experience. Because the electrical field decays rapidly with distance, the inversion of the seafloor SP data only gives a shallow current density distribution, which cannot accurately reflect the situation of the space current dipole source under the seafloor. Therefore, it is necessary to introduce a depth-weighting function to enhance the sensitivity of the grid element far from the observation point. The depth-weighting function Wz should be placed in the model constraints as follows (Biswas and Sharma, 2017; Portniaguine and Zhdanov, 1999): 5Wj=∑i=1NJij2p/4,j=1,…,M, where Wj is the jth element on the main diagonal of matrix Wz and J is the sensitivity matrix (or Jacobian matrix). M is the number of model vector elements, and N is the number of the observation point. The p value is constant and generally equal to 1. The extent of depth weighting can be adjusted by the value of p. For multicomponent inversion, the full three-component electrical field were used and the total amount of observation data became 3×N.

In application, the SP data can be inverted in many ways to determine the flow direction of the fluid or the location of the self-potential source. When the source location is to be solved, it was usually assumed that the conductivity distribution in the space is known or determined by other electrical exploration methods (DC conductivity or induction-based electromagnetic methods), and thus the SP source is equivalent to the external current source. Therefore, this study used the inverted volumetric current density distribution to locate the simulated artificial source. On land, the SP method usually directly measures the electric potential between the roll electrode and a fixed electrode; therefore, the potential data V are directly chosen for inversion. However, in the marine environment, it is impossible to fix a reference electrode on the seafloor when the ship or AUV is moving. Therefore, the SP method conducted in the marine environment usually measures the gradient of electric potential, i.e., the electrical field, and the potential is calculated indirectly by numerical integration. In the case of this study, the measured electrical field components were directly inverted.

Owing to the limited observation data collected by the AUV-SP system near the artificial source, the inversion result is nonunique. Therefore, the constraints of known information, such as the lake water conductivity (1 S m-1, measured with the environment sensor from the AUV) and lake bottom sediment conductivity (assumed to be 0.1 S m-1), were added to the inversion. To investigate the advantages of multicomponent inversion, single-component and two-component observation data were used for the inversion and the inversion results were compared. Figure 7 shows the results obtained by inversion of electrical field data of a single component (horizontal component and vertical component, Fig. 7a and b, respectively) and two components (horizontal component and vertical component, Fig. 7c) under the same constraint conditions. Negative charge density was recovered surrounding the negative pole of the artificial source in all results, which preliminarily revealed the reliability of the system and the effectiveness of the inversion method. Indeed, because the AUV-SP system's navigation path in the lake is mainly near the negative pole of the artificial source, the negative pole position of the artificial source can be precisely located, while positive pole is not easily located due to the lack of observation data of the surrounding area, although some positive charge density is retrieved along the survey line. By comparing the inversion results between a single component (Fig. 7a and b) and multiple components (Fig. 7c), it was observed that the inversion results of the two horizontal components of the electrical field were more focused, and the negative charge intensity was concentrated, while the negative charge density obtained from the inversion results of the single vertical component of the electrical field was accompanied by positive charge anomaly, which dispersed the negative charge energy and deviated from the actual situation. Combinations of vertical and horizontal components of electrical field data restrict the horizontal distribution of charge density and make the inversion more consistent with the actual artificial source. In addition, the current flow direction (Fig. 7b, d, f) also shows that the inversion of only a single component will produce the wrong current direction and that additional constraints (e.g., boundary limits, heterogeneous distribution of conductivity) were required to obtain reasonable inversion results. Therefore, multicomponent detection of electrical field is necessary when addressing complex geological structure, especially involving flow of fluids.

A seafloor SP measurement system that consisted of three pairs of perpendicular electrodes attached to the AUV (Qianlong 2) was introduced. By deploying an artificial current source, the multicomponent electrical field responses in the water were measured by the AUV in Qiandao Lake. The observed results were consistent with the numerical simulation, which verifies the feasibility of the AUV-SP system for multicomponent SP exploration. At present, the AUV-SP system can work in water for about 40 h and complete the SP measurement of a high-resolution area of about 20 km2.

The test results of two surveys showed that the AUV-SP system has good consistency and repeatability and that the overall noise level of the measured data is relatively low, which meets the requirements of near-seafloor SP exploration at a large scale. The inversion of the test data of the AUV-SP system demonstrated that the position and intensity of the inverted current source were basically consistent with the actual position and intensity of the artificial current source. Therefore, the SP response measured by the AUV-SP system can be used to locate the seafloor SP source to achieve an accurate exploration and evaluation of seafloor sulfide deposits. In the case of simple geoelectric structure, a single vertical component measurement would be sufficient to locate the SP source, but in solving complex problems such as hydrothermal or fluid flow, a three-component SP measurement is crucial.

In view of the vibration noise of the connecting rod presented in this test, our next step will be to select a more suitable material and design a bracket with better rigidity to reduce the vibration noise. The experience and data accumulated in this experiment also provide a reference for the future research of active-source electromagnetic detectors on AUVs.