High-temperature (>300 °C) off-axis hydrothermal systems found along the slow-spreading Mid-Atlantic Ridge are apparently consistently located at outcropping fault zones. While preferential flow of hot fluids along highly permeable, fractured rocks seems intuitive, such efficient flow inevitably leads to the entrainment of cold ambient seawater. The temperature drop this should cause is difficult to reconcile with the observed high-temperature black smoker activity and formation of associated massive sulfide ore deposits. Here we combine newly acquired seismological data from the high-temperature, off-axis Logatchev 1 hydrothermal field (LHF1) with numerical modeling of hydrothermal flow to solve this apparent contradiction. The data show intense off-axis seismicity with focal mechanisms suggesting a fault zone dipping from LHF1 toward the ridge axis. Our simulations predict high-temperature venting at LHF1 only for a limited range of fault widths and permeability contrasts, expressed as the fault’s relative transmissibility (the product of the two parameters). The relative transmissibility must be sufficient to “capture” a rising hydrothermal plume and redirect it toward LHF1 but low enough to prevent extensive mixing with ambient cold fluids. Furthermore, the temperature drop associated with any high permeability zone in heterogeneous crust may explain why a significant part of hydrothermal discharge along slow-spreading ridges occurs at low temperatures.
High-temperature hydrothermal systems at slow-spreading ridges are commonly linked to fault zones that cut the seafloor several kilometers off axis (McCaig et al., 2007). Extensional deformation increases permeability and makes fault zones the preferred pathways for hydrothermal fluids. The Mid-Atlantic Ridge (MAR) black smoker vent fields TAG (Trans-Atlantic Geotraverse; deMartin et al., 2007), Logatchev (Petersen et al., 2009), and Petersburg (Shilov et al., 2012) are all located at such faults that outcrop as much as 16 km away from the ridge axis. Fault-related vent fields are active over longer periods of time than ridge-centered hydrothermal systems at fast-spreading ridges (50–100 k.y. versus 100–1000 yr; Jamieson et al., 2014). As a consequence of this longevity, fault-controlled submarine massive sulfide deposits at slow-spreading ridges tend to be larger than their on-axis counterparts at fast-spreading ridges (Fouquet, 1997). Fluids feeding these systems must be hot enough (>300 °C) to transport metals from the crust to the seafloor (Hannington et al., 2011). If fault zones were isolated conduits, hot hydrothermal fluids could easily flow through them without significant cooling. However, fault zones are open systems, and, depending on their geometry and fluid-dynamic properties, significant mixing between hot hydrothermal fluids and colder ambient fluids occurs. Therefore, the fundamental dilemma in high-temperature fault-controlled hydrothermal systems is that a certain permeability contrast is required to provide a preferred fluid pathway, but that mixing with colder ambient fluids should result in a significant reduction in temperature.
These processes can be excellently studied at the fault-controlled Logatchev 1 hydrothermal field (LHF1), which is located on a segment of the MAR undergoing tectonic extension and core-complex formation. Here we combine new seismological data with numerical modeling of hydrothermal flow to elucidate the hydrological regime beneath LHF1 and to show under which conditions fluid flow along fault zones can be reconciled with the observed high vent temperatures.
THE LOGATCHEV 1 HYDROTHERMAL FIELD
The LHF1 is located 8 km off axis at 14.75°Ν on the MAR (Fig. 1A). High-temperature discharge has been stable at ∼300–350 °C for the past decade (Schmidt et al., 2007). Evidence for detachment faulting and core-complex formation is manifested in outcropping serpentinized ultramafics and gabbroic rocks, in surface structures (Petersen et al., 2009) as well as in off-axis seismic activity. New seismological data were acquired during two deployments in 2009 (Fig. 1; Grevemeyer et al., 2013) and indicate that deformation is concentrated along the eastern rift mountains in the vicinity of LHF1. Hypocenters of normal-faulting earthquakes roughly line up with the outcrop of a fault zone near LHF1 (Fig. 1B) and likely mark the subsurface continuation of the fault zone. This fractured, more permeable region could be the preferred pathway for hydrothermal fluids feeding the LHF1 vent system, and the diffuse seismicity pattern surrounding the normal-faulting earthquakes may reflect thermal stresses caused by this hydrothermal circulation. The absence of earthquakes at depths >5 km below seafloor (i.e., >9 km below sea level; Fig. 1B) suggests that temperatures there are too high for brittle failure (>600 °C), indicating a high geothermal gradient and possibly the presence of an intrusion providing heat to drive the hydrothermal convection.
MODELING FRAMEWORK AND SETUP
Based on the geophysical data discussed above, we constructed a two-dimensional numerical model for hydrothermal flow along the 8-km-long transect shown in Figure 1. The numerical mesh follows the observed bathymetry and includes the seismically imaged fault zone. Because the precise width of the fault zone cannot be inferred from the seismic data, we have conducted calculations with width, d, of the fault zone varying between 1 m and 2 km to cover the range from a narrow permeable fault zone to a much wider deformation zone containing multiple faults. In addition, we have varied the fault zone’s permeability, kf, to obtain different permeability contrasts (c = kf/kb = 3, 10, 30, and 100) with respect to a background permeability kb = 0.5 × 10−15 m2. This constant background permeability was chosen such that a homogenous model without fault zone predicts both high-temperature venting as well as magmatic temperatures at depth for a reasonable heat input. The heat source is simulated by a Gaussian-shaped profile and an integrated heat input of 12.5 kW per meter ridge axis, an average value for the MAR (Sinha et al., 2004). The heat source is assumed to lie beneath a cluster of micro-earthquakes at transect kilometer 6.5, which is ∼2 km to the west of LHF1 (Fig. 1B). See Table 1 for rock properties and other model parameters.
The governing equations for convection of supercritical pure water, taking into account its thermodynamic properties, are solved on an unstructured triangular mesh using the finite element method (see the GSA Data Repository1 and Hasenclever et al.  for details). All simulations were run until steady-state temperature fields evolved and hydrothermal heat flow at the seafloor was equal to magmatic heat input.
Figure 2 shows the three different regimes of hydrothermal convection that we identified in our 66 model calculations. In Figure 2, the fault has a 10× higher permeability compared to the surrounding crust, and only its width is varied. When the fault width is small (d = 10 m; Fig. 2A), the rising hydrothermal plume is not captured, so that high-temperature venting occurs above the heat source rather than at the location of LHF1. When the fault width is increased to 30 m (Fig. 2B), the plume is deflected into the fault zone, and high-temperature venting (337 °C) is predicted at the location of LHF1. Increasing the fault width even further (150 m; Fig. 2C) leads to strongly enhanced upward mass flux within the fault combined with a significant decrease in vent temperature to ∼285 °C, in disagreement with measured vent fluid exit temperatures at LHF1 (300–350 °C; Schmidt et al., 2007).
We have systematically varied fault width and permeability contrast in our model calculations and find similar trends for each permeability contrast (Fig. 3). When the fault is very narrow, vent temperatures are as high as observed but venting occurs above the heat source rather than at LHF1 (open symbols in Figs. 3A and 3B). For wider fault zones, the hydrothermal plume is redirected toward LHF1 but vent temperatures are gradually reduced (filled symbols in Figs. 3A and 3B). We have also tested varying background permeabilities (see additional 32 simulations in the Data Repository) and find that only the overall maximum vent temperature decreases if background permeability is higher, consistent with the thermodynamic processes described by Driesner (2010). Nonetheless, for each permeability contrast there is an optimal fault width that maximizes the vent temperature at the fault termination.
The fluid-dynamic and thermodynamic mechanisms responsible for the trends in Figure 3A can be explained by the fluid mass fluxes (Figs. 2 and 3C). The fluid-dynamic influence of a more permeable fault zone is twofold: (1) the inclined geometry of the fault zone induces a horizontal pore-pressure gradient that deflects the mainly buoyancy-driven vertical fluid flow, and (2) the higher permeability within the fault zone results in higher fluid velocities and thereby a higher mass flux compared to upflow outside the fault zone. The example runs shown in Figure 2 illustrate this. In Figure 2A, the mass flux near the fault zone is slightly enhanced but not sufficient to deflect the plume. With a wider fault (Fig. 2B), the horizontal pore-pressure gradient induced by the permeability contrast is sufficient to redirect the entire plume. The mass flux inside an even wider fault zone (Fig. 2C) is further enhanced and causes entrainment of ambient cold seawater, thereby reducing the temperature of the venting fluid.
Qin denotes total magmatic heat input in W/m, is fluid mass flux (both recharging and discharging in steady state) in kg/s/m, and h is the temperature-dependent specific fluid enthalpy in J/kg evaluated at a constant pressure of 30 MPa. Employing the equation of state for pure water, Equation 1 can be solved for the venting temperature Tvent, providing a theoretical relationship between discharge mass flux and discharge temperature. This theoretical relation is shown as three curves in Figure 3C, corresponding to cases in which 0%, 10%, and 20% of the basal energy input is conductively transported through the seafloor. In our simulations, we quantify the conductive heat loss through the seafloor to be 10%–20% of the basal heat input, which explains why most of our predicted venting temperatures plot between these theoretical temperatures.
The striking agreement between modeled vent temperature and the theoretical estimate highlights the strong relationship between discharge mass flux and vent temperature. If we define a fault’s relative transmissibility as the product of its width and its permeability contrast to the surroundings, we can examine how relative transmissibility, discharge mass flux, and vent temperature are related. From Figure 3D, we see that the higher the relative transmissibility, the higher the total mass flux and the lower the vent temperature—and only a narrow range of relative fault transmissibilities exists for which venting occurs at the fault zone termination with the observed high temperatures.
The repeated observation of high-temperature off-axis venting on fault traces could have two possible explanations. (1) The driving heat source is located vertically beneath the fault outcrop and the fluids rise more or less vertically through less-fractured low-permeability rocks, exiting by chance on a fault outcrop. While this could be an explanation for a single vent site, we see no reason to expect it to be generally applicable for the many cases of fault-related high-temperature venting (e.g., TAG, LHF1, etc.). (2) As suggested by our numerical results, venting fault zones must have a relative transmissibility high enough to redirect a vertically rising plume but low enough to minimize mixing processes between hot hydrothermal and cold ambient fluids. This then raises the question of whether this condition is an innate characteristic of venting faults or whether processes associated with hydrothermal flow somehow modify fault width and/or permeability to favor high-temperature flow.
Potentially important processes for facilitating fault-controlled high-temperature venting are precipitation reactions within the fault zone that clog pore space in an initially too-wide or too-permeable fault zone. Anhydrite, for example, precipitates from ambient seawater when it is heated over a temperature interval of 150–200 °C (Seyfried and Bischoff, 1981). This would lead to a progressive sealing of porosity in the fault zone, reducing its relative transmissibility. Silica precipitation may have a similar effect. Such precipitation reactions would progressively reduce mass flux within the fault zone until mixing processes between hot hydrothermal and cold ambient fluids are minimized. At this point, hydrothermal upflow temperatures would reach their maximum. Testing such a self-adjusting mechanism for various initial fault geometries is, however, beyond the scope of this study.
The here-studied processes may well be the reason why diffuse low-temperature venting is more common than high-temperature venting (Baker et al., 1993). Especially slow-spreading ridges such as the MAR, whose tectonically dominated accretion mechanisms form more heterogeneous crust than fast-spreading ridges, may have a much larger fraction of low- than high-temperature discharge. Supporting evidence comes from global mass-balance budgets for hydrothermal metals in the ocean: If hydrothermal cooling of the young ocean floor would predominantly result in high-temperature venting, the amount of copper carried by the fluids and deposited at the seafloor in the neovolcanic zone should be much larger than observed (Cathles, 2011; Hannington, 2013).
SUMMARY AND CONCLUSIONS
High-temperature black smoker systems along slow-spreading ridges such as the MAR are almost always related to tectonic fault zones and therefore are commonly found off axis. Here we have shown that the intuitive assumption of more efficient high-temperature upflow along high-permeability fault zones only can work under specific conditions. To do so, fault zones need to be just permeable and wide enough to capture and redirect hydrothermal plumes rising from depth but, because they are not isolated conduits, must not be too wide or permeable in order to prevent cooling through mixing with ambient colder fluids. The common occurrence of fault-linked high-temperature vent fields strongly points at a not-yet-quantified self-adjusting permeability that depends on pore space–clogging reactions between hydrothermal and ambient cold fluids.
Based on these findings and the seismological data, a consistent picture for the LHF1 vent field may be drawn. A magmatic intrusion below a cluster of earthquakes 2 km west of LHF1 drives high-temperature hydrothermal convection. A fault zone, indicated by normal-faulting earthquakes, has the optimal relative transmissibility (either by coincidence but more likely by some self-adjusting precipitation reactions) to redirect the rising hydrothermal plume toward the location of the LHF1, resulting in the observed high-temperature venting at the tip of the fault zone.
Our findings further show that an intrinsic relationship exists between permeability, mass flux, and upflow temperature. The higher the permeability, the higher the mass flux, and the lower the vent temperature. This simple relationship may well explain the sparse high-temperature vent fields along the MAR and why the heterogeneous crust of the Atlantic, with its strong permeability contrasts, is predominantly cooled by lower-temperature fluid flow.
Constructive reviews of Philipp Weis improved the paper considerably. This paper benefited from valuable feedback of anonymous reviewers and helpful remarks of Colin Devey. We thank Editor Ellen Thomas for the editorial handling. This project could be realized due to the funding of the German Research Foundation (DFG) project RU1469/2-1.