Detachment-parallel recharge explains high discharge uxes at the TAG hydrothermal eld

13 Submarine massive sulfide deposits on slow-spreading ridges are larger and longer-lived than deposits at fast-spreading 14 ridges , likely due to more pronounced tectonic faulting creating stable preferential fluid pathways . The TAG 15 hydrothermal mound at 26◦N on the Mid-Atlantic Ridge (MAR) is a typical example located on the hanging wall of a 16 detachment fault. It has formed through distinct phases of high-temperature fluid discharge lasting 10s to 100s of 17 years throughout at least the last 50,000 years and is one of the largest sulfide accumulations on the MAR. Yet, the 18 mechanisms that control the episodic behavior, keep the fluid pathways intact, and sustain the observed high heat fluxes 19 of up to 1800 MW remain poorly understood. Previous concepts involved long-distance channelized high-temperature 20 fluid upflow along the detachment 10 but that circulation mode is thermodynamically unfavorable and incompatible 21 with TAG’s high discharge fluxes. Here, based on the joint interpretation of hydrothermal flow observations and 3-D 22 flow modeling, we show that the TAG system can be explained by episodic magmatic intrusions into the footwall of 23 a highly permeable detachment surface. These intrusions drive episodes of hydrothermal activity with sub-vertical 24 discharge and recharge along the detachment. This revised flow regime reconciles problematic aspects of previously 25 inferred circulation patterns and can be used as guidance to one critical combination of parameters that can generate 26 substantive mineral systems. 27


28
High temperature hydrothermal discharge at black smoker vent sites has been reported from mid-ocean ridge segments opening 29 at all spreading rates 12, 13 and is known to play a key role in global biogeochemical cycles 14-16 as well as in the formation of 30 massive sulfide ore deposits 1 . The style of venting, the composition of the discharged fluids, and the controls on vent field 31 locality all appear, however, to be affected by spreading rate-dependent processes 17 . At intermediate-to fast-spreading ridges, 32 where plate separation is compensated by magma emplacement, hydrothermal vent sites are located on-axis and hydrothermal 33 circulation is driven by heat released from a quasi-stable melt lens modulated by periodic dike emplacement events 3, 18, 19 . 34 cooling is likely highly episodic with vent fields along slow-spreading ridges only being active about 5% of the time 9 , possibly 48 paced by the frequency of magmatic intrusions. This episodic nature of hydrothermal cooling has been documented at the 49 TAG hydrothermal field, where drilling during ODP leg 158 probed the internal structure of the mound and fairly detailed 50 age constraints are available 29,30 . Mass-balancing the amount of sulfide at the TAG mound and the amount of fluid needed to 51 sustain the inferred total heat discharge revealed that TAG was probably only active < 2% of its approx. 50 kyrs life time 8 . 52 Interestingly, these two lines of argument that 1) deposits at slow-spreading ridges are larger due to long time spans of 53 activity and 2) high discharge fluxes require cooling to be episodic, are difficult to reconcile with each other unless each phase 54 of activity reuses the same plumbing system to form a long-lived deposit. But what critical combination of hydro-tectono-55 magmatic conditions is required for this to occur? Here, using the TAG hydrothermal field as an example, we identify a pattern 56 of circulation that can sustain transient high discharge fluxes at a fault-controlled vent-system -one that has the potential to 57 repeatedly focus hydrothermal discharge at the TAG mound over multiple cycles of activity.

58
The TAG hydrothermal field 59 The TAG hydrothermal field is located off-axis at 26 • N on the eastern flank of the Mid-Atlantic Ridge (MAR). The main site 60 of high-temperature venting is currently located at the TAG mound, where black smokers are discharging fluids at approx. 360 61 • C 31 . The system is highly productive with inferred energy discharge fluxes of 86 to 1,800 MW 9 ; with the spread reflecting 62 different types of measurements referring to localized discharge at the active mound or integrated total diffusive plus localized 63 heat discharge at the TAG segment. This hydrothermal activity has resulted in the accumulation of ∼ 2.7 Mt of massive 64 sulfides at the active mound and ∼ 20 Mt in the wider TAG area 32 . In addition to the focused high-temperature venting, 65 widespread diffuse venting is occurring as evidenced by the abundant anhydrite within the TAG mound that likely formed that the TAG mound is located on the hanging wall of the detachment directly at the intersection of two sets of normal faults, 75 one parallel to the spreading direction and one oblique oriented in SW-NE direction 32 (Fig. 1). 76 While these observations point to strong interrelations between tectonic faulting, magmatic activity, and hydrothermal 77 flow, identifying the driving heat source has been a challenge and with it the identification of circulation pathways. Slip on 78 the detachment, which progressively brings hotter footwall rocks closer to the surface, does not provide sufficient energy to 79 sustain the discharge fluxes at TAG, which most likely require a magmatic heat source 6 . Two main options appear plausible: 80 either the magmatic heat source is located beneath the neo-volcanic zone, or a magmatic intrusion in the footwall beneath 81 TAG is driving flow. Unfortunately, seismic surveys have struggled to resolve this question. While Kong et al. 37 found a 82 low velocity anomaly at 3-6 km depth beneath TAG, a later study by Canales et al. 6 could not identify an intrusion in the 83 TAG footwall. However, a 3-D tomography based on the data of the same seismic survey did reveal a low velocity anomaly 84 and a zone of inverted vertical velocity gradients beneath TAG 36 , possibly in support of a magmatic footwall intrusion (see 85 Extended Data Fig. 1d).

86
Based on the micro-seismicity data, deMartin et al. (2007) 5 proposed that a deep magmatic intrusion approx. 7 km 87 beneath the neovolcanic zone drives channelized high-temperature hydrothermal flow along the detachment to below the 88 active mound. This two-dimensional concept of channelized high temperature fluid flow along a detachment surface has been 89 highly influential and invoked to explain off-axis venting at Logatchev 11 on the MAR and Longqi 28 on the Southwest Indian 90 Ridge (SWIR). However, recent theoretical work showed that channelizing hot fluids over long distances along a low-angle 91 detachment is difficult. Hot fluids tend to rise vertically due to their high buoyancy, so that strong permeability contrasts are 92 necessary, which inevitably result in mixing processes and low vent temperatures incompatible with observations; except for 93 very special parameter combinations 11 .

94
An alternative flow solution is hinted at by the joint interpretation of the high-resolution bathymetry 32 and 3-D tomography 95 data 36 , which show that TAG is located at intersecting normal faults in the hanging wall and is centered above a slow seismic 96 anomaly in the footwall (Extended Data Fig. 1). It appears plausible that flow is driven by a series of footwall magmatic 97 intrusions with discharge being vertical in the direction of buoyancy along the cross-cutting faults in the hanging wall and 98 recharge occurring in the third-dimension along the detachment surface. Here, using a combined analytical and numerical ig. 1: The TAG hydrothermal field in models and data. a, High resolution (2m) AUV-based bathymetric data shows the location of the TAG and Mir sites, termination and corrugated surface of the detachment fault, extended detachment (black dashed line), and regions of axis-parallel (N-S) and oblique (NE-SW) faulting. The thin black box denotes lateral extent of Fig. 1c. In the sub-seafloor, dots represent location of microearthquakes 5 . The intrusion driving the current hydrothermal phase is sketched as gradient-color filled ellipse. Extended Data Fig. 1 provides further details on the sub-seafloor structure. b, Close-up of seafloor affected by cross-cutting normal faulting around the TAG mound and Mir Zone. The axis-parallel and oblique fault regions are bounded by green and blue lines, and their strike orientations are indicated in the inset rose diagram. c, Results of 3D hydrothermal flow modeling. The dark inclined plane inside the modeling domain represents the presumed detachment fault zone with incline angle 20 • and thickness 50 m, the blue lines with arrows denote pathways of numerical fluid tracers. Isotherms of 100, 200, 350 • C are shown as transparent surfaces. Recharge mass flux mainly occurs along the detachment surface. Discharge flow is vertical along a zone of enhanced permeability towards the active TAG mound. Note that only a part of the full modeling domain is shown for improved readability. The complete fluid velocity field is shown in Supplementary Fig.3. d and e show the temperature field on vertical profiles across the TAG vent for k d f = 10 −12 m 2 and 5 × 10 −15 m 2 , respectively. Energy discharge increases for higher detachment fault permeability due to a thinner thermal boundary layer.

3/18
approach, we show that this flow solution is robust and stable over a large parameter range and that its magmatic-tectonic 100 ingredients may represent a critical combination of parameters that make the TAG mineral system so prolific.

102
To explore the likely circulation pattern during phases of high temperature hydrothermal discharge, we use the three-dimensional 103 hydrothermal flow model HydrothermalFoam 38 , which resolves porous convection of pure water under single-phase condi-104 tions. Based on the high-resolution AUV-bathymetry 32 , micro-earthquake locations 5 , and tomographic 36 plus seismic reflec-105 tion 39 data, we implement the detachment surface as an inclined permeable plane dipping at 20 • . Here the assumption is 106 that the detachment surface is a zone of enhanced permeability with respect to the adjacent foot-and hanging walls 40 . The 107 cross-cutting faults at TAG are simplified as a pipe-or slot-shaped zone ( Supplementary Fig. 1) of enhanced permeability 108 that we assume intersects the detachment surface approx. 700 m below the seafloor. The presumed driving heat source in the 109 detachment footwall is implemented as a Gaussian-shaped fixed temperature boundary condition (see Methods). zone. In the case of a highly permeable detachment (k d f = 10 −12 m 2 ), the total conductive heat input is 219 MW (Fig. 1d).

127
If k d f is reduced to 5 × 10 −15 m 2 , the conductive boundary layer is thicker and the heat input is reduced to 15 MW (Fig. 1e).

128
Hence the total heat output scales with reaction zone permeability, which implies that having the heat source close to the 129 permeable flow zone is an effective way to increase the total heat output of a circulation system (see ref. 41  for a model run that defines the upflow zone as a permeable pipe in which the detachment fault is twice as permeable as the 134 pipe. About 85% of recharge mass flow occurs via the detachment and approx. 65% of the discharge occurs via the pipe, 135 which is mainly used for discharge flow. Vent temperature is high at approx. 405 • C. If the pipe permeability is increased by a 136 factor of 4 ( Fig. 2b), the pipe is used for both recharge and discharge flow, which results in a reduced vent temperature due to to 10 −12 m 2 range typically reported for shallow ocean crust 43 . However, as cautious note, it should be added that the sub-148 surface permeability structure of the highly faulted TAG segment is likely more complex and is likely to sustain more diffusive 149 low-temperature flow. Our simplified model setup was designed to capture the key flow characteristics of the focused high 150 temperature circulation system.

151
To evaluate the robustness of our findings, we have derived a semi-analytical solution for the power output of a hydrother-   Pie charts show the integrated mass flow rate of recharge (Q re : kg s −1 ) and discharge (Q dis : kg s −1 ), and hydrothermal power output (E dis : MW) at the seafloor. The number in each pie chart is the total value of the corresponding quantity. Wedges in each pie chart represent the proportion of flow through pipe/slot (green), detachment fault (orange), and background rock matrix (cyan). Comparing a and b on a like for like basis show that increasing k pipe , the permeability of the upflow zone, results in mixing and a decrease in vent temperature. Comparing b and c shows that increasing the detachment permeability k d f dramatically increases the discharge flow, which reduces mixing in the upflow zone so that the vent temperature is increased, also the power output is increased. Comparing c and d illustrates the effects of changing the upflow zone geometry from pipe-like to slot-like; additional recharge flow occurs and the total power output is increased by 40%.

157
The presented flow solutions illustrate the likely circulation pattern during phases of high temperature fluid discharge at TAG.

158
The fundamental difference to previous concepts 5, 11, 28 on fault-controlled circulation systems is that in our new model the  rock interaction was established. However, the conclusion was drawn that this interaction happened within a reaction zone at 174 depth, which was later exhumed by faulting.

175
While the presented numerical results are consistent with the available data on the current phase of hydrothermal activity, 176 they do not directly explain the episodic nature of the TAG hydrothermal system. As aforementioned, the TAG mound has been 177 episodically active since approx. 50,000 yrs with each phase lasting 10s -100s of years. It appears plausible that these phases to be re-activated. The current seafloor morphology suggests that cross-cutting normal faults act as conduits for hydrothermal 187 6/18 discharge 32 . However, for these pathways to be re-activated and not be replaced by other preferential pathways, the hanging 188 wall must not have experienced significant tectonic deformation throughout the life time of the TAG mound. One plausible 189 explanation is that extension is mainly accommodated by the detachment and possibly by magmatic accretion at the ridge-axis, 190 so that the hanging wall did not experience strong recent deformation.

191
An active highly permeable detachment that allows for efficient heat extraction from magmatic footwall intrusions, in com-192 bination with stable preferential pathways in the hanging wall that are re-activated throughout multiple phases of hydrother-193 mal discharge, may therefore be the ingredients facilitating the formation of large massive sulfide deposits at detachment- k denotes permeability, p total fluid pressure, ⃗ g gravitational acceleration, µ f and ρ f are the fluid's dynamic viscosity and 321 density, respectively. Considering a compressible fluid in a porous medium with given porosity structure, the mass balance is 322 expressed by where ε is the porosity of the rock. Note that we assume the matrix to be incompressible, so that the porosity is outside the 324 time derivative. The equation for pressure can be derived by substituting Darcy's law (Equation 1) into the continuity equation

325
(2) and treating the fluid's density as a function of temperature T and pressure p: with α f and β f being the fluid's thermal expansivity and compressibility, respectively. Again there is no rock compressibilty as 327 we consider the incompressible matrix case. Energy conservation of a single-phase fluid can be expressed using a temperature 328 formulation 19,38 , Thermodynamic properties of fluids, i.e. water, are calculated using the IAPWS-IF97 formulation 52, 53 , that provides the addition, the numerical model is based on the hypothesis that the TAG hydrothermal system is driven by shallow intrusion(s). 338 We therefore only consider the shallow part of the detachment 36 down to a depth of 6 km below sea level. The 3D model is where ⃗ S f ace is the surface vector of the f ace with magnitude of face area and pointing outside of the 3D model domain, N is 370 the number of faces. Based on the specific enthalpy (H f ) of the fluids, calculated from IAPWS-IF97, the total discharge heat 371 output can be calculated as with H 0 being specific enthalpy of seawater with temperature 2 • C.
where H hs represents the depth of the heat source centre below the seafloor, R x the semi-axis length of the heat source ellipse 393 along the x-axis. Therefore, the pressure difference driving fluid from recharge zone (detachment fault zone) into reaction 394 zone (above heat source) is approximately given by where ∆x denotes offset of heat source centre and pipe centre along x-axis (similar meaning with ∆y, ∆z).

11/18
This pressure difference operates over distance d(θ ) so that the magnitude of Darcy's velocity (or volume flux) from 397 recharge zone to reaction zone can be expressed as where µ U denotes dynamic viscosity of the upwelling hot fluid and k R is permeability of the reaction zone (i.e. detachment in 399 this model setup). d(θ ) is the distance between heat source boundary and pipe bottom centre (see the bluish dash-dotted line 400 in Extended Data Fig. 3b) i.e.
Combining Equation 11, 12 and 13, the total mass flux into the reaction zone is expressed by a surface integration over the 402 boundary of the reaction zone, The discharge zone, represented by a cylindric pipe with permeability k D , is much narrower than the recharge zone and 404 thus represents a stronger total resistance to a given fluid volume flux. The discharge flow is driven by a vertical pressure 405 gradient due to the density contrast of hot upwelling fluid and cold seawater. Similar to Equation 12, the discharge volume 406 flux can be written as Consequently, the discharge mass flux flow out of the reaction zone is where S pipe is the cross-sectional area of the pipe zone. Considering the structure of convection cells and reaction zone, we While H R can be expressed in terms of E cond and driving temperature T D by applying the energy conservation law (see Eq.  view of (c) with integrated geophysical data. Axis-parallel and oblique faults area are represented by green and white polygons. Yellow volumes below seafloor represent contour surface of -0.5 1/s of vertical gradient of P-wave velocity. Blue and orange volume represent contour surface -3% and -5% of P-wave velocity variation, respectively. Black incline surface underneath seafloor denotes detachment fault zone inferred from both 3D tomography data and micro-earthquake data. Extended Data Fig. 3: Model geometry of detachment fault controlled hydrothermal system. Assuming the geometry of heat source boundary patch is ellipse with semi-major axis R x and semi-minor R z , and semi-major axis is parallel with x axis of the coordinate system.  Fig. 4: Comparison of semi-analytical and 3-D numerical model predictions on hydrothermal power output. The dashed lines display the analytic relationship (see methods) between conductive heat input (E cond ), permeability of reaction zone (k R ), and driving temperature(T D ). The numerical models share the same parameters for geometry (see Extended Data Fig.3) and boundary conditions with the simplified analytic model but also include effects of variations in discharge zone permeability (k D ), shown as differing symbols. Power output mainly scales with reaction zone permeability and driving temperature, which both control thermal boundary layer thickness. Predictions of analytical and numerical models deviate at high permeability values, most likely because the analytical model assumes radial symmetry while the 3-D model evolves, without such constraints, to a nearly but not perfect symmetric upflow zone.