Evolution of the Thermal and Dehydration State of Sediments Entering the North Sumatra Subduction Zone

Shallow slip on the plate‐boundary fault during the 2004 Mw 9.2 Aceh‐Andaman Earthquake, offshore North Sumatra, has been linked to thick incoming sediments on the oceanic plate with advanced diagenetic and sediment property changes at the depths of plate boundary fault development. We couple age control, physical, and thermal property measurements from International Ocean Discovery Program (IODP) drilling with multichannel seismic reflection data to reconstruct the thermal structure and evolution of the incoming sediment column, up to the point of accretion/subduction. Lithospheric thermal rejuvenation around 58 Ma is required to explain anomalously high heat flux at IODP Site U1480, and heat flux within the trench is suppressed by very high sediment accumulation rates during the development of a thick trench wedge. Accumulation of up to ∼4.5 km of thick Nicobar Fan and trench wedge sediments produces temperatures >150°C within the basal sediments where the décollement develops, resulting in total pre‐subduction diagenetic dehydration of basal sediments. The smectite‐illite transformation within these basal sediments produces sufficient fluid to explain a polarity reversal on a pre‐décollement reflector. We suggest that the boundary between basal‐pelagic and siliciclastic‐fan sediments has trapped fluid over the last ∼1 Myr, as a result of primary lithological properties, diagenetic fluid release, and cementation, controlling décollement formation at this weakened level. Pre‐subduction dehydration of large portions of the accreted sediment column strengthens the décollement beneath the prism, extending co‐seismic velocity‐weakening behavior close to the trench, which may occur at other subduction zones with similar sediment compositions, thicknesses, and/or temperatures.

The frictional stability of the décollement at a subduction zone depends on dynamic geologic factors such as fluid pressure (e.g., Ellis et al., 2015;Ranero et al., 2008;Spinelli & Saffer, 2004), consolidation state of the surrounding material, and the mineralogy of fault zone materials (Fagereng & Toy, 2011). Scholz (1998) defined a frictional stability parameter from a rate-and state-variable friction law that depends on material properties and effective normal stress. The friction law is derived from experimental laboratory studies of artificial faults in largely homogeneous materials (e.g., Dieterich, 1972;Lockner & Byerlee, 1986;Rabinowicz, 1958). Seismogenesis is limited to the décollement that exhibits velocity-weakening behavior (a negative frictional stability parameter  ) (Scholz, 1998(Scholz, , 2019. The limits (up-and down-dip) to seismic behavior are the transitions from velocity-weakening (seismic slip) to velocity-strengthening (stable-sliding). Determining real frictional properties for the material of and around the décollement at depth in the dynamic environment of a subduction zone remains extremely difficult since the seismogenic décollement is typically 5-40 km below seafloor.
Various studies have attempted to estimate the width and position of the seismogenic zone at subduction zones, where earthquakes can initiate, often by modeling the thermal regime and comparing the results to observed seismicity/earthquake rupture zones were available, with similar methods used for forecasting where no historic ruptures have occurred (e.g., Hyndman & Wang, 1993 [Cascadia]; Hyndman et al., 1995 [Nankai]; Hippchen &Hyndman, 2008 andKlingelhoefer et al., 2010 [Sumatra]; Smith et al., 2013 [Makran]; Gutscher et al., 2016 [Ryukyu]). Seismic imaging techniques are used to define the geometry of the subducting slab and sediment thickness (at the deformation front), and parameters such as convergence rate and plate age, are employed to match the results to seafloor heat flux measurements. Seafloor heat flux values are estimated by shallow heat probe measurements, borehole measurements, and/or derived from bottom-simulating-reflectors of gas hydrates. The seismogenic zone is proposed to lie between the 150°C and 350°C isotherms (Hyndman et al., 1997), where 350°C is the downdip limit at the onset of the brittle-ductile transition (Scholz, 2019).
A key driver to link seismogenesis and thermal state is that thermally controlled diagenetic and low-grade metamorphic reactions occurring during burial change the frictional properties of sediments (e.g., Vrolijk, 1990). Potential reactions include the smectite to illite transition (e.g., Saffer et al., 2008); the opal to quartz transition (Spinelli & Underwood, 2004); carbonate and zeolite cementation; and hydrocarbon maturation (Moore & Saffer, 2001). These processes consolidate sediments, release fluids and alter sediment frictional properties; they are estimated from laboratory studies to take place over intervals between ∼50°C and 175°C (e.g., Ernst, 1990;Mizutani, 1970;Pytte & Reynolds, 1989). In most subduction zones, such temperatures are reached some distance landward of the deformation front, below the forearc. Total completion of these reactions may release volumes of fluid up to 29% of the original mineral volume for opal and 40% for smectite (Hüpers et al., 2017).
The temperature of the décollement sediments (or the sediment-basement interface; sometimes the position of the décollement) depends on sediment burial history. Where the sediment column is sufficiently thick on the incoming plate, basal sediments have been proposed to reach the temperatures of these diagenetic and low-grade-metamorphic reactions, before sediments enter the subduction zone (e.g., Hüpers et al., 2017;Ike et al., 2008;Smith et al., 2013;Spinelli & Underwood, 2004). This has led to the hypothesis that thick input sediments can bring the change from velocity-strengthening (stable-sliding) to velocity-weakening (seismic slip) (the updip limit of seismogenesis) further updip, by drawing the onset of mineral transformations and associated cementation seaward and increasing the potential seismogenic width and thus earthquake magnitude (e.g., Dean et al., 2010;Moore & Saffer, 2001;Saffer et al., 2008;Smith et al., 2012). Such a "shallow" rupture also increases the size of a potential tsunami as rupture extends into deeper water above the outer forearc.

Tectonic Setting
The Sunda subduction zone extends from Bali to the Andaman Islands, and forms by normal to oblique subduction of the Indo-Australian plate beneath the Sunda plate. In the north (offshore North Sumatra and the Nicobar Islands), the incoming plate sedimentary section reaches thicknesses >5 km at the trench, thicker than most other accretionary margins, due to the presence of the Bengal-Nicobar Fan system. The succession was sampled on the oceanic plate, ∼230 km seaward of the North Sumatra subduction zone by International Ocean Discovery Program (IODP) Expedition 362   (Figures 1 and 2). It is composed of thick Nicobar Fan deposits that range from clay to finegrained sand, overlying a basal layer of pelagic and tuffaceous sediments. The sediments are mostly poorly consolidated and unlithified, except for the deepest intervals. IODP drilling also encountered intrusive and extrusive igneous intervals within the deepest units overlying oceanic basement at Site U1480; these units are ∼10 Myr younger than the 68-Myrold oceanic basement (Figure 3; Backman et al., 2019) and are probably associated with volcanic activity on the flanks of the Ninety East Ridge (NER). This has important implications for the thermal evolution of the sedimentary succession, which we explore in this paper.
The 2004 M w 9.2 Aceh-Andaman earthquake ruptured ∼1,600 km of the northern segment of the Sunda subduction zone. Coseismic slip during the earthquake reached far seaward beneath the accretionary wedge, and potentially to the trench (Bletery et al., 2016), and the extensive rupture and large coseismic slip resulted in a tsunami that claimed over 250,000 lives (Lay et al., 2005). Toward the southern end of the rupture zone, Dean et al. (2010) identified a high-amplitude, negative polarity (HANP) seismic reflector at depth within the incoming plate sediments, which is interpreted to develop into the plate boundary fault. We hereafter refer to the HANP as polarity reversal on reflector R11 (Figure 2). Dean et al. (2010), Geersen et al. (2013), and Gulick et al. (2011) have argued that this seismic character indicates diagenetic fluid generation at depths of 2-3 km, occurring in the thick input sediments before subduction. The along-subduction zone extent of this reflector broadly corresponds to a region of unusual prism structure and morphology, interpreted to also result from unusual material and basal properties (e.g., . Beneath the accretionary prism, the reflector becomes less reflective, suggesting fluids have been released along prism faults (Dean et al., 2010).
IODP Expedition 362 identified reflector R11 as the downward transition from predominantly fan to pelagic sediments, and found geochemical evidence for the recent release of fresh water at depth in the boreholes, STEVENS ET AL.
10.1029/2020GC009306 3 of 21 Figure 1. Regional tectonics, plate age, and seafloor heat flux. (a) Eastern Indian Ocean setting. Toothed line is Sumatra-Andaman subduction zone deformation front; bathymetry (2,000 m contours) with 90 East Ridge (NER); red box, location of (b). (b) Age contours of oceanic crust (Jacob et al., 2014); red-white-blue points, seafloor heat flux measurements from Jessop (2012) in mWm −2 ; brown squares IODP sites, orange diamond and associated value show average surface heat flow at Site U1480 calculated by IODP in mWm −2 ; black lines (1 and 2) are seismic profiles. SSZ -Sumatra subduction zone. (c) Predicted basal heat flux from Stein and Stein (1992) model for cooling of oceanic crust (without any sedimentation efftects), with surface heat flux points from (b). Heat flux measurement colors indicate measured value relative to the predicted value for plate age at the site (of Stein and Stein, 1992); red = high; blue = low; white = close to the predicted value. interpreted as diagenetic fluid release having already started even >200 km from the subduction zone (Hüpers et al., 2017;McNeill et al., 2017). Hüpers et al. (2017) used thermal hydrogeologic modeling to show that these fluids were likely produced through dehydration of amorphous silica, where this reaction would be complete before the basal sediments reach the trench. They also predict that the smectite to illite clay dehydration reaction would be almost entirely complete before the basal sediments entered the subduction zone. This differs from other accretionary margins, for example, Nankai and the Lesser Antilles/Barbados, where hydrogeologic models show that smectite dehydration at the level of the décollement begins and ceases below the forearc (Moore & Saffer, 2001;Saffer & Tobin, 2011).
In this study, we build on the work of Hüpers et al. (2017) by investigating the thermal structure and evolution of the entire sediment section on the North Sumatra incoming plate between the NER and the subduction zone deformation front. We combine multichannel seismic (MCS) data with ground-truthed material properties data from IODP sites in the input section of the subduction zone ( Figure 1) and use a time-varying thermal model (Hutnak & Fisher, 2007) to determine the temperature on the seismic profiles through time offshore the rupture area of the 2004 earthquake, testing a range of parameters to examine uncertainties. Using this thermal history model we map the progression of diagenetic dehydration reactions through the sediment column during its deposition and estimate the amount of fluid liberated from the basal sediments that underlie reflector R11. We subsequently test the likelihood of the unusually thick sedimentary section generating a shallow seismogenic updip limit and increased earthquake rupture area. STEVENS ET AL.

Datasets
We use MCS data from the northern Wharton Basin (Figure 1; Geersen et al., 2013;McNeill et al., 2017) and core and log data from IODP Expedition 362 boreholes (Figure 1; Dugan et al., 2017;McNeill et al., 2017). The seismic data are from Gaedicke (2007; cruise SO186-2 of the SEACAUSE campaign on FS Sonne, with data jointly owned by German and Indonesian institutions) and porosity, biostratigraphic, and gamma ray data are from IODP Expedition 362 Sites U1480 and U1481. We also incorporate regional seafloor heat flux data, derived from the World Heat Flow Data Collection (Jessop, 2012).

Sedimentary Parameters
Borehole sediment properties were correlated onto and along the seismic profiles, that is stratigraphic age and boundaries, lithology, velocity, porosity, natural gamma ray/radiogenic element composition, thermal properties, and direct comparison of seismic with core and log properties.
The subparallel, laterally-continuous seismic reflectors can be confidently traced along most of each seismic section ( Figure 2). We interpret 14, roughly evenly spaced reflectors R1-R14, in accordance with the interpretation of Pickering et al. (2020), and correlate these from the drill sites to the subduction zone deformation front and between the two profiles (using a third crossing seismic profile, not presented here). Reflector R11 is the horizon that develops into the HANP pre-décollement reflector of Dean et al. (2010). We interpolate the depth-converted positions of these reflectors at Site U1480 (see below for velocity information) and apply ages from the borehole stratigraphy to provide a chronology for the seismic stratigraphy, assuming that the reflectors are isochronous along the profile. Reflector R2 onlaps onto reflector R3 within the trench wedge, and is therefore not intercepted at the IODP drill sites further seaward, we estimate an age for this reflector of ca. 1.0 Ma based on the average sediment accumulation rate through the trench and our depth-conversion of the reflectors. Reflector R13 is not resolvable in the seismic data at Site U1480, so is also excluded because its age cannot be reliably defined.
Three major units A, B, and C are interpreted (note these differ from the lithostratigraphic units of IODP Expedition 362). Unit A (Figures 2 and 3), the wedge filling the subduction trench (the trench wedge), is ∼2,500 m thick at the deformation front, but only 140 m thick at Site U1480. It is composed of calcareous clays, fine-grained sand, silty sand and clay, minor calcareous ooze, and ash at the borehole location but these lithologies may differ in the trench, and it is bound by the seafloor (R1) and reflector R4. At the location of U1480, our Unit A corresponds to all of IODP Expedition 362 lithostratigraphic Unit I and uppermost Subunit IIA. Unit B consists of Nicobar Fan thin-to medium-bedded predominantly siliciclastic silts, clays, and fine-grained sands, bound by reflectors R4 and R11, which is ∼1,100 m thick at Site U1480 (our unit B corresponds to most of IODP 362 Subunit IIA, and all of subunits IIB and IIC) and ∼1,800 m thick at the deformation front. Unit C is the deepest, between reflector R11 and the top of the oceanic basement (R14); at Site U1480 this is an ∼150 m thick interval that consists of mostly pelagic claystone, minor chalk, and tuffaceous material, includes semi-lithified and lithified materials and has a relative lack of terrigenous siliciclastic sediments (our Unit C corresponds to IODP 362 Units III-V in McNeill et al., 2017). The sediments of Unit C slowly accumulated over a period of ∼50 Myr, whereas the overlying 1250 m of Units A and B (at Site U1480) accumulated rapidly over ∼9 Myr (Backman et al., 2019;McNeill et al., 2017). The total sediment thickness changes from ∼1,400 m at Site U1480 to a maximum of ∼5,100 m at the deformation front on seismic profile 1 (Figure 3), primarily due to thickening of Unit A. Shallow sediment accumulation rates, therefore, increase dramatically within Unit A as the plate approaches the subduction zone.

Regional Heat Flow
The oceanic crust on the profiles formed at ∼70-60 Ma (Figure 1; Jacob et al., 2014). Inset C in Figure 1 shows predicted surface heat flow based on the Stein and Stein (1992) oceanic plate cooling model, and available seafloor heat flux probe measurements from Jessop (2012) plotted against the age of the oceanic crust at their location (Müller et al., 2008), including from temperature measurements made in the IODP Expedition 362 boreholes (McNeill et al., 2017). There is a significant scatter in the data, although there is some clustering around the predicted heat flux values. Plate cooling models (e.g., Stein & Stein, 1992) assume that all heat transfer is conductive, and do not account for heat transfer via convective water circulation. Seafloor probe heat flux measurements are also subject to factors that may bias measurements, for example the tilt of the instrument as it enters the sediment and near surface effects such as local fluid flow at faults (Stein & Abbott, 1991). This is particularly important where oceanic basement rocks are exposed, or sediment cover is thin, and convection is strong (Stein & Abbott, 1991;Hutnak &Fisher, 2007), for example at the NER, and may account for anomalously high heat flow values there ( Figure 1). Near Site U1480, the surface heat flux probe measurement from Jessop (2012) is 96.3 mWm −2 . In contrast, advanced piston corer temperature tool (APCT-3) formation temperature measurements in Hole U1480 E to a depth of ∼200 mbsf (meters below seafloor) generate an estimated surface heat flux of 72.6 ± 5.1 and 75.0 ± 6.6 mWm −2 , using a Bullard plot, and the temperature gradient and thermal conductivity, respectively, with errors given at the 95% confidence interval. There are numerous faults within the sedimentary column in the Expedition 362 borehole area (e.g., Geersen et al., 2015;Stevens et al., 2020) that could generate anomalous heat flow values, but Site U1480 is >500 m away from the nearest seismically resolvable fault. We, therefore, argue that the heat flux estimates generated from the borehole APCT-3 measurements are more reliable than the seafloor measurements. We test our thermal models at Site U1480 against these different estimates, however in the absence of a series of drilling-controlled heat flux measurements toward the deformation front, we also compare our model results with the seafloor probe measurements from Jessop (2012).

Thermal Modeling and Input Parameters: Time-Depth Conversion, Thermal Rejuvenation and Radiogenic Heat Production
Thermal models of the sedimentary section were constructed using the SlugSed one-dimensional fluid and heat transport modeling software coded in MATLAB (Hutnak & Fisher, 2007). This model uses accumulation rates of sediment layers, where the sediment-basement interface subsides and creates accommodation space in which sediment accumulates and compacts by a user-specified porosity-depth function. We define the sedimentary layers between the seafloor (reflector R1) and reflector R14 (excluding reflectors R2 and R13, see above), and use a porosity-depth function derived from the measurements made at Site U1480. We construct one-dimensional models at broadly evenly-spaced localities along each seismic profile (approximately 1,000 common depth points [CDPs] apart, where CDP spacing is 6.25 m). At each model locality, we depth-convert interpreted horizons to return sedimentary layer thicknesses in meters. Depth conversion was performed on horizon picks (as opposed to applied to the entire MCS data -we only required horizon depths at discrete locations) using interval velocities derived from the original seismic processing (at approximately 1,000 CDP intervals). These velocities are not derived from tomographic studies, however, they show gradual lateral changes, are consistent with the core-log-seismic integration following drilling at Sites U1480 and U1481 , and are similar to velocities determined nearer the trench (Qin and Singh, 2017). We clip the maximum seismic velocity within the sedimentary section to 4.5 kms −1 , consistent with Qin and Singh (2017) and Ghosal et al. (2014). This produces a basement depth at the trench of ∼10 km below sea level, consistent with Singh et al. (2012). Nevertheless, uncertainties in seismic velocity could be a source of error in our models, and therefore we test the effect of varying velocities by ±15% which we consider exceeds the true error bounds (results for ±15% velocities are presented in the supporting information).
Each one-dimensional model is stressed as the simulated sedimentary layers accumulate between the seafloor and a subsiding sediment-basement-interface where evolving basal heat flux is applied. We apply a heat flux that decays with a 1/√time relationship, based on the Stein and Stein (1992) thermal model for the oceanic lithosphere. In this study, our focus is on the thermal state of the sediments (as opposed to the shallow oceanic crust). So although there are other lithospheric cooling models available (e.g., McKenzie et al., 2005), the good fit between the Stein and Stein (1992) model and global seafloor heat flux measurements suggests it is appropriate for our study.
Extrusive igneous intervals at Site U1480 (Unit IV; Figure 3c) that are ∼10 Myr younger than the 68 Ma oceanic basement (Backman et al., 2019;McNeill et al., 2017) indicate active magmatism at the NER ∼58 Ma. We suggest that such magmatism is related to a melting event that occurred after the formation of the oceanic lithosphere. We have tested the effect of a thermal rejuvenation of the lithosphere at 58 Ma, that is, an increase in/reset of the basal heat flux to the level expected at the formation of the oceanic basement that affects the thermal evolution of the sedimentary section. We assume a melting event at the NER influences locations within ∼100 km of the NER flank (∼275 km from the NER center), based on typical plate thicknesses.
The Expedition 362 results provide the ground truth of sediment material properties that we incorporate into the model. We use core porosity data and thermal conductivity measurements from Site U1480 to define bulk thermal conductivity for the sediment column. The thermal conductivity measurements for most of the core samples (particularly mud and calcareous samples) can be fitted either to a porosity-controlled geometric mean model (e.g., Lovell, 1985; Figure S4) using a sediment grain thermal conductivity of 3.0 W m −1 K −1 and a bulk conductivity defined by the proportion of solid sediment grains to pore fluid or with no clear depth trend or correlation with porosity. We used the geometric mean for the models presented here but also tested how surficial heat flux is affected by using an average of the core measurements as a single bulk conductivity throughout the sediment column (see supporting information Text S1.5 and Figure S5).
The modeled sediments are thermally equilibrated between the sediment-basement interface and the seafloor, which is held at a constant temperature of 1.25 °C from IODP seafloor measurements . Basal sediments are heated by the geothermal gradient as the sediment column thickens. We also include radiogenic heat production from the sediments estimated from natural gamma downhole log measurements at Site U1481 ( Figure 4) using an existing method to convert natural gamma ray (NGR) API (American Petroleum Institute) counts (recorded by downhole NGR instruments) to heat production in STEVENS ET AL.  . Natural Gamma Ray logging data for uncased intervals of sites U1480 and U1481 and conversion to radiogenic heat production based on the method of Bücker and Rybach (1996). μWm −3 (Bücker & Rybach, 1996). Core sample NGR measurements made at both Sites U1480 and U1481, support that radiogenic properties are similar between the sites. In our calculations of heat production, we exclude samples where the caliper is at maximum reach, that is, where the borehole wall is washed out or collapsed. This risks excluding the NGR values of lithologies prone to being washed out, for example, sandy sections, however, Site U1480 NGR core measurements suggest that NGR does not vary significantly with lithology or depth ( Figure S6). Our analysis of the NGR logs for Site U1481 indicates average radiogenic heat production values of 1.37 and 1.54 μWm −3 for Units B and C, respectively. Unit A was not logged, but core NGR measurements on IODP lithostratigraphic Units I and II are similar to each other (varying between 20 to 80 counts/s for Unit I and 30 to 90 counts/s for Unit II; McNeill et al., 2017), suggesting Unit A probably produces a similar radiogenic heat to Unit B.
For the thermal capacity of sediment grains and seawater (not available from borehole data), we use data from an analogous environment where these measurements were made on similar lithologies (Goto & Matsubayashi, 2009).

Dehydration Calculations
We run our 1D models to generate temperatures for the entire height of the sediment column at each model locality at 10,000-years time steps. For reflectors R1 -R14 (where R1 is present-day seafloor, and R14 is top of the basement; R13 is excluded) we extract time-depth and time-temperature history at each model location by tracing the depositional history of these horizons through the model domain. We account for compaction by finding the "bulk-sediment-velocity" from the harmonic mean of sediment-grain-and pore-water-velocity (see Hutnak & Fisher, 2007). We then assess the progress of diagenetic, thermally-controlled mineral dehydration, by applying reaction kinetics to the temperature histories of each reflector. For the two-step process of amorphous silica dehydration, where opal-A is transformed to opal-CT, and opal-CT to quartz we use the kinetic expressions of Mizutani (1970), in which the rate of change in the molar fractions of Opal-A (OA), and Opal-CT (OCT) abundance is given by Equations 1a and 1b, respectively: Where OA and OCT are the mole fractions of Opal-A and Opal-CT from the previous time step (1 and 0, respectively at the time of deposition), A is the frequency factor (A OA = 7.51 × 10 −4 s −1 and A OCT = 2.30 × 10 −4 s −1 ), E is the activation energy (E = 6.69 × 10 4 J mol −1 ), R is the gas constant (J mol −1 K −1 ) and T is the temperature (K). We then calculate the rate at which fluid volume is generated (V silica (unit volume s −1 )) by these reactions using Equation 2 (Kimura et al., 2012): Where H is water content of the three silica phases (2.1-12.1 wt% for Opal-A, average 7.3 wt%; 1.0-8.9 wt% for Opal-CT, average 5.5 wt% (Day & Jones, 2008) and 0.4 wt% for Quartz (QTZ) (Graetsch, 1994)), OA i is the initial content of opal-A in the bulk sediment (we use a value of 18 wt%, which Hüpers et al. (2017) showed is required to explain pore water freshening at Site U1480), n is porosity, h is the thickness of the dehydrating sediment column; ρ sed is the density of sediment grains, and ρ iw is the density of interlayer water (we assume 2,700 and 1,030 kg m −3 , respectively). For the smectite to illite transformation, separately we apply both the reaction kinetic derived by Huang et al. (1993) from laboratory experiments at high temperatures, and a range of potassium ion concentrations (T = 250-325 °C, [K + ] = 100-3,000 mM) (Equation 3a); and the reaction kinetic derived by Pytte and Reynolds (1989), which incorporates a synthesis of field data, and fitted to observations of the reaction progress from a wide range of thermal histories in different geological settings, with peak temperatures from 70°C to 250°C and sediment ages ranging from ca. 10 years to 300 Myr (Equation 3b): Where S is the molar fraction of smectite in the Illite-Smectite (I/S) mixed layer from the previous time step (we assume S = 0.8 at the time of deposition), and A S is the frequency factor (A S-Huang = 8.08 × 10 4 s −1 , A S-Pytte = 5.2 × 10 7 s −1 ). In our application of Equation 3a Here H S is the water content of smectite (we assume H S = 20 wt%, where the smectite water interlayer basal d-spacing is 15 Å; Bird [1984]), and S i is the initial mass fraction of smectite in the bulk sediment. We separately tested rate parameters from Equations 3a and 3b to assess the impact of using the Huang et al. (1993) and Pytte and Reynolds (1989) reaction kinetics, respectively. The IODP Expedition 362 results indicate the bulk sediment is composed of 65 wt% clays , where we assume 80% is smectite (from S = 0.8 in I/S mixed layer). This is consistent with the measurements of Rosenberger et al. (2020) at Site U1480. However, we acknowledge that their measurements at Site U1481 showed a lower proportion of smectite, which would reduce the fluid expulsion rates by a factor of 2. It has been suggested that oven drying of samples may produce artificially high porosity results (Brown et al., 2001); Brown and Ransom (1996) defined a porosity correction which we choose not to use. The Brown and Ransom (1996) correction would increase the grain-to-pore-space ratio, that is to increase the value of (1 − n) in Equation 4, so in that respect, our calculations likely underestimate the released fluid volume.
In addition to assessing the progression of dehydration reactions at individual reflectors (e.g., at R11/the décollement level), by running the depositional history of each reflector through these equations, we reconstruct the dehydration state (in terms of rate of fluid expulsion per unit volume for silica/smectite) of the entire incoming sediment section at specific time steps, for example, before deposition of the Nicobar Fan, immediately before the development of the trench wedge, and at the present day (see Section 3.3).

Initial Thermal Model Tests
The primary variables in our model are: sediment accumulation rate; radiogenic heat production; and timing of the imposed basal heat flux. Sediment accumulation rates are controlled by our seismic interpretation and the age model ( Figure 3). The effect of variation in sediment accumulation rates is tested through the application of ±15% picked velocities within the sediment column in time-depth conversion (see Section 2.4). To test the effects of radiogenic heat production within the sediment column and rejuvenation of the basal heat flux, we ran repeat one-dimensional models at the location of Site U1480. We compare our temperature and surface heat flux results from repeat model runs with IODP measurements, to assess their validity and derive the best estimate. The IODP measurements indicate that the geothermal gradient within the shallowest ∼200 m of sediments at Site U1480 is ∼44.4 °C km −1 with a surface heat flux of 72.6 ± 5.1 mW.m −2 (lower estimate) . Extrapolation of this thermal gradient down to basement depths suggests a temperature at the sediment-basement interface at Site U1480 of ∼60 °C.
An initial model stressed over 11 periods of sediment accumulation defined by the seismic reflectors, applying no radiogenic heat production within the sediment column and using a basal heat flux decaying from the basement age of 68 Ma predicts a present-day seafloor heat flux of 62 mWm −2 . This is consistent with the predictions of the Stein and Stein (1992) model, but is lower than the borehole-derived estimation, indicating additional processes are required to generate higher heat flux. We, therefore, tested the effects of radiogenic heat production and lithospheric thermal rejuvenation on surface heat flux.
We applied a sediment radiogenic heat production value of 1.3 μW.m −3 throughout the sediment column, based on NGR log data (average derived value, see above, Figure 4). This results in a present day seafloor heat flux of 63 mWm −2 . Due to the remaining significant discrepancy with the lower borehole-derived estimate of 72.6 ± 5.1 mW.m −2 , we assessed what level of radiogenic heat production within the sediment column would be required to match our model output to the borehole-derived estimate. Only when ∼9 μW. m −3 radiogenic heat production is applied throughout the sediment column do we attain a surface heat flux value of 72.6 mW.m −2 (all other parameters unchanged). This far exceeds heat production values typical of any sediment type (e.g., siltstones 1.8 μWm −3 ) (Rybach, 1986) including those encountered at Site U1480, and equates to an NGR API reading of 570 (Bücker & Rybach, 1996) whereas measured values at Site U1481 are <150 API (Figure 4). We conclude that radiogenic heat production within the sediment column makes only a minor contribution to the resulting thermal structure and seafloor heat flux.
We assume that 1.3 μW.m −3 is a good approximation of radiogenic heat produced within the sediment column, and subsequently tested the effect of a perturbation to the imposed basal heat flux. During the course of the model (which starts at a basement age of 68 Ma) we reset the basal heat flux to the initial value (510 mW.m −2 ; Stein & Stein, 1992) at 58 Ma -representing the effect of lithospheric rejuvenation coeval with volcanic lava flows in Unit IV at Site U1480 (Backman et al., 2019). This "thermal rejuvenation" generates a final surface heat flux of 68 mW m −2 at the location of Site U1480, which is now within the 95% confidence level of the lower borehole-derived heat flux estimate (72.6 ± 5.1 mW.m −2 ). The combination of thermal rejuvenation and sediment radiogenic heat production results in a basement temperature of 56.6°C (average thermal gradient ∼42 °C km −1 ). These results are generally consistent with Hüpers et al. (2017) (∼60°C at the base of IODP lithostratigraphic Unit III and a geothermal gradient of 44.4°C.km −1 ). Therefore, we suggest lithospheric thermal rejuvenation played a key role in the thermal history of the sediments currently at and near the location of Site U1480, in a similar mechanism to that recognized for Hawaii by Detrick and Crough (1978).

Burial History
Following the model tests described above, we apply the best estimate of parameters to the entire incoming sediment section and model its time evolution as it approaches the subduction zone. Sediment accumulation rates increase toward the deformation front controlled by the increasing thickness of seismic packages between reflectors. Radiogenic heat production through the sediment column is based on the measured NGR log values. A lithospheric thermal rejuvenation event is applied to all model localities within 100 km of the NER.
We highlight the results from 3 localities on seismic profile 1: at 68 km (Site U1480), 195 km (Pseudo site 1), and 290 km (Pseudo site 2) along-profile (232, 100, and 5 km from the deformation front, respectively) (Figure 5, locations on Figure 2). In Figure 5 we present the burial and temperature history, and dehydration reaction progression at the R11 or pre-décollement level since the deposition of R11 at 9.3 Ma. Pseudo site 1 is the position at which R11's reflection polarity becomes reversed, and Pseudo site 2 is our closest model to the deformation front. The lateral variation in the sediment thickness between seismic reflectors results in different sediment accumulation rates for some time periods (i.e., the gradients of the burial depth curves in Figures 5a-5c) between the three pseudo site localities. Seismic velocities within the deepest sediments close to the deformation front are the least certain, so potential errors in the time-depth conversion are highest here.

Temperature Structure and Progression of Mineral Dehydration Through the Sediment Column
From the sediment column temperatures generated by our models at each model locality, we reconstruct a two-dimensional (2D) grid of temperatures for the entire incoming sedimentary section at specific time steps (Figure 6), specifically: 9.3 Ma (R11), onset of Nicobar Fan deposition; 2.2 Ma (R4), the onset of main trench wedge deposition; 1.0 Ma (R2), the approximate time of maximum fluid generation beneath R11 at Pseudo site 2 due to clay mineral dehydration (based on the Pytte and Reynolds (1989) kinetic); and the present day (R1). Accumulation of the >2 km thick trench wedge is largely responsible for the high basal temperatures close to the deformation front, reaching >100 °C. However, from Pseudo site 1 to the deformation front, temperatures of ∼100 °C are already reached within basal sediments that underlie R11, before the start of trench wedge deposition (Figure 6c). For each time step, we also extract the surface heat flux across seismic profile 1 (Figure 6a). As expected, surface heat flux is much higher where we apply thermal rejuvenation. The surface heat flux is generally consistent with predictions of the Stein and Stein (1992) model, with some variation related to basement topography. However, during the development of the trench wedge, where trench wedge sediments are thickest (∼250 km along-profile toward the deformation front), there is a significant decrease in the surface heat flow and deviation from the global values due to the blanketing effect of the young sediments (also see Figure s5).

STEVENS ET AL.
10.1029/2020GC009306 11 of 21 Figures 5d and 5e show the molar fraction change for the modeled dehydration reactions for Site U1480 and for Pseudo sites 1 and 2. The associated rate of fluid expulsion due to these mineral reactions is presented in Figure 5 G-I. At Site U1480 our results suggest that the opal to quartz reaction is still ongoing, with a maximum fluid production rate due to silica dehydration of ∼0.2 × 10 −15 s −1 (per unit volume of bulk sediment) ca. 1.7 Ma, consistent with Hüpers et al. (2017), since when this process becomes the largest fluid source. Our calculations also imply that the smectite-illite reaction has yet to begin at Site U1480, consistent with Hüpers et al. (2017). At Pseudo site 1, the modeled opal-quartz reaction is complete by the present-day. Compaction remains the largest source of fluid release throughout burial history at this location, however, the smectite-illite reaction starts around the onset of trench-sediment deposition (2.2 Ma). Higher sediment accumulation rates earlier in the burial history of Pseudo site 2 (and associated early more rapid increase in temperature) lead to earlier onset and completion (prior to trench wedge development) of the opal-quartz STEVENS ET AL.

10.1029/2020GC009306
12 of 21 reaction (Figure 5f). The smectite-illite reaction begins just before the onset of trench wedge development, and accelerates rapidly at 2.2 Ma, becoming by far the dominant fluid source over the most recent time periods. The Pytte and Reynolds (1989) kinetic suggests that fluid expulsion from the smectite to illite transformation peaks ca. 1.0 Ma (the approximate age of reflector R2), whereas the Huang et al. (1993) kinetic suggests it peaks within the last 200 kyr (Figure 5i).
Our reconstructions of the dehydration state of the incoming sediment section are presented in Figure 7 (for smectite dehydration, Pytte and Reynolds (1989) kinetic (see Figure S8 for results using the Huang et al. (1993) kinetic)) and S7 (for silica dehydration). Some fluid is produced through amorphous silica dehydration within the basal sediments in the present day around the location of Site U1480 (Hüpers et al., 2017). However, toward the deformation front, and particularly near where R11 becomes reversed STEVENS ET AL.

10.1029/2020GC009306
13 of 21 polarity, silica dehydration predominantly occurs shallower than R11, ∼1-2 km below the seafloor (even when ±15% velocities are used in the seismic time-depth conversion). Our results suggest diachronous completion of the silica dehydration reaction within the basal sediments that underlie R11, where burial activated the reaction progressively further seaward, with the reaction complete in the furthest seaward of these sediments (∼250 km) shortly after 2.2 Ma.
In contrast, our reconstructions for smectite dehydration suggest that this reaction was only just beginning within basal sediments at 2.2 Ma. We calculate that, approximately halfway through the deposition of the current trench wedge sediments at 1.0 Ma, fluid expulsion from smectite dehydration was focused at the R11 (or pre-décollement) level, (dashed line in Figures 7c-7e [at zero depth in B]), in the region where we observe reversed polarity on the reflector in the present day. Our reconstruction for the present day indicates that, by this time, beneath the trench wedge, smectite dehydration has progressed to the sediments immediately above R11 (Figure 7e). This implies that the basal sediments currently at the deformation front are fully dehydrated, before entering the subduction zone. Figure 7a shows the corresponding thickness of fluid produced by the basal sediments (underlying R11, dashed line) along seismic profile 1 (a cumulative plot of the total fluid from the reaction, showing the contribution from each time period is shown in Figure 8d). We calculate that, where R11 polarity is reversed, similar thicknesses of fluid are generated in the period from 1.0 Ma to the present compared to the much longer 9.3-2.2 Ma period, and that ca. 50% more fluid is generated in the 1.0 Ma to present day period than the 2.2-1.0 Ma period. These results suggest a very strong correlation between where fluid is being generated by smectite dehydration in the last 1 Myr and the region of reversed polarity on reflector R11 (the pre-décollement reflector). This correlation is even more pronounced when we apply the Huang et al. (1993) reaction kinetic ( Figure 8c) and remains robust when ±15% velocities are used in the seismic time-depth conversion ( Figures S8-S12).

Heat Flux Across the Incoming Section
We compared our model results to regional measurements of seafloor heat flux, by extracting heat flux across the shallowest 2 m of sediment from each 1D model used to build the 2D thermal structure of the incoming sediment section (Figure 6a). The thermal rejuvenation event applied within 100 km of the flank of the NER increases heat flux values above those expected for the plate age (Stein & Stein, 1992). From ∼125-250 km along-profile, our results (60-63 mWm −2 ) match the expected values (61.8 mWm −2 ) and the closest probe measurement (61 mWm −1 ; Jessop [2012]; Figure 1b). However, within 40 km of the deformation front, seafloor heat flux from our models decreases slightly within the last 2.2 Ma. The high sediment accumulation rates (SAR), especially within the trench (≥1000 m Myr −1 ; Figures 3 and S5i), suppress the surficial heat flux relative to the basal input ( Figure S5). At Site U1480, maximum SARs ∼225 m Myr −1 between 2.6 and 1.7 Ma suppress the seafloor heat flux by only around 1.3%, and after the SAR slows to <20 m Myr −1 at 1.7 Ma, the heat flux rebounds. At Pseudo Site 1, higher SARs since 2.2 Ma (up to 400 m Myr −1 ) result in ∼2.5% suppression, by the present day. At Pseudo Site 2, where SARs are much greater since 2.2 Ma (∼1,000-1,700 m Myr −1 ), the present-day model seafloor heat flux is suppressed by ∼15% ( Figure S5c) (or 17.5% using a constant bulk sediment conductivity). Hutnak and Fisher (2007) modeled that sediment accumulation rates of 500 m Myr −1 may suppress the surface heat flux by as much as 32%, and Hutchison (1985) suggests that a SAR of 1,000 m Myr −1 would suppress the heat flux by as much as 55%. However, these studies used a constant SAR over 10-20 Myr. Harris et al. (2020) suggested that, at the Cascadia margin, a SAR of ∼13,00 m Myr −1 over ∼3 Myr suppresses surface heat flux by >50%, but 3 Myrs is still significantly longer than any period of such high sediment accumulation in our study. For example, at Pseudo Site 2, our highest SAR (1,740 m Myr −1 ) persists for only 0.5 Myr between 2.2 and 1.7 Ma, after which the SAR slows significantly allowing seafloor heat flux to rebound, but not return to equilibrium with the basal input, and then the heat flux continues to decrease. The variable SAR throughout the deposition of the Nicobar Fan and accumulation of the trench wedge produces multiple instances of seafloor heat flux suppression and rebound ( Figure S5). Should the high SARs within the trench continue into the future (e.g., on the timescales used by Hutchison [1985] and Hutnak and Fisher [2007]), we would expect much greater suppression of the seafloor heat flux. Heat probe measurements at/near the deformation front elsewhere on the Sumatran margin indicate even lower values (Figure 1; Jessop, 2012) than implied by our models. We suggest that such low values are from measurements taken in areas where: the SAR has been high and more consistent over a longer period; there is anomalously low basal heat flux; or there has been recent slope failure. Slope failure is common close to the steep outer forearc of the Sumatra margin (Tappin et al., 2007), and may have deposited sediment thicker than the probe penetration depth (e.g., 5-10 m), effectively blanketing the seafloor at the point of measurement, the instantaneous SAR would be extremely high compared to the longer term average, disproportionally suppressing heat flux to the seafloor. We also note that radiogenic heat production within the sediment column offsets the suppressing effect of sedimentation for example, by ∼6% at the deformation front (within the last 2 Myr), consistent with the study by Hutchison (1985).

Mechanisms for Trapping of Diagenetic Fluids and Explanation of Seismic Properties
Full waveform inversion analysis of seismic data within our area (Qin and Singh, 2017) indicates high P wave velocities (>4 km s −1 ) through sediment Unit C (underlying reflector R11) close to the deformation front. To reconcile these high background velocities with the polarity reversal and high amplitude of reflector R11 in seismic reflection data, these authors suggest that the reflector represents a thin (70-80 m) layer anomalously high porosity (∼30%, >20% higher than the background level) containing overpressured fluid. This agrees with the ∼80 m thick freshening anomaly and increased porosity interval detected at R11 depth at Site U1480 by Hüpers et al. (2017). We suggest that a mobilized fluid source of "thickness" 20-25 m (30% of 70-80 m) could be trapped at the R11 level and explain the polarity reversal.
Based on our results (Figure 8), we estimate that, since 9.3 Ma, fluid well in excess of 20-25 m thickness has been generated by compaction and dehydration of the sediments below R11 along most of the length of seismic profile 1 (Figure 8e), where compaction and the dehydration of smectite as it transforms to illite are the key fluid sources. Our calculations may even underestimate expelled fluid volumes from mineral dehydration since we do not consider the expansion of mineral-bound water upon its release where its density decreases by ∼5% (e.g., Bethke, 1986). However, the reflection amplitude and polarity of R11 are only anomalously high and reversed within ∼100 km of the deformation front (landward of Pseudo Site 1). If our calculations are accurate and any of the fluid generated seaward of Pseudo Site 1 was focused into a layer at R11, much of it must have since been drained. Alternatively, Pseudo Site 1 marks the seaward limit of a process that causes trapping of fluid at R11. We propose that this process is cementation within the turbidite-rich sediments of the Nicobar Fan (Unit B) immediately overlying reflector R11. Silica and calcite ions are released as a by-product of the smectite-illite transformation and transferred from shales to sandstones by migration of pore waters, where they then produce quartz overgrowths and calcite cement at temperatures >60°C (Boles & Franks, 1979). This process has frequently been invoked in the context of frictional properties at subduction zone décollements (e.g., Ikari et al., 2007;Marone & Scholz, 1988;Moore et al., 2007;Saffer et al., 2012). Our model of the smectite-illite transformation suggests that in the present day it is ongoing within the sediments immediately overlying R11, where the reflector's polarity is reversed (Figure 7e), and essentially complete in the underlying sediments. We suggest that fluid is expelled from the pelagic claystone sediments of Unit C by compaction and the smectite-illite transformation since ca. 1.0 Ma are focused (possibly facilitated by pervasive faulting throughout the incoming sediment section (related to the Indian Ocean diffuse plate boundary, e.g., Geersen et al. [2013]; Stevens et al. [2020]) where the faults are un-cemented) into a thin high-porosity layer at the R11 level. Above this level, concomitant cementation turbidite sands within the trench wedge and fan stratigraphy (and locally of cross-cutting fault planes) could create an impermeable seal preventing further upward migration. We also suggest that this supports the argument made by Geersen et al. (2013) that fault planes at R11 depth act as impermeable barriers.
We suggest that in shallower parts of the sediment column above the present-day zone of smectite dehydration (<2,000 mbsf; Figure 7e) there is efficient draining of fluid to the seafloor: we observe no seismic evidence of fluid trapping in this interval, yet we predict recent fluid generation through silica dehydration ( Figure S7). Draining of fluid from the sediment column above R11 is likely facilitated by a combination of relatively porous and permeable (uncemented) fan and trench wedge sediments and pervasive faulting (e.g., Geersen et al., 2015;Stevens et al., 2020).

Thermal Structure Influence on Décollement Development and Seismic Behavior
The position of the décollement in the outer forearc ranges in different locations from relatively shallow within the sediment section (e.g., the Mediterranean Ridge/Hellenic subduction zone; Westbrook & Reston, 2002), to the sediment-basement interface (e.g., central Sumatra; Cook et al., 2014). Some margins also exhibit significant along-strike variation in the depth of the décollement e.g. along the Cascadia margin (e.g., Han et al., 2017), and south-central Chile margin (Olsen et al., 2020). The position of décollement formation is thought to be dictated by preferential propagation through frictionally weak intervals in the incoming sediment column, and this control may be lithological, physical, hydrological, or a combination (Underwood, 2007).
The North Sumatran example studied here shows apparent combined lithological, diagenetic, and hydrological properties controlling the position of décollement formation. Other examples worldwide include the frontal décollement at the Barbados accretionary prism which, from ocean drilling, forms along a frictionally weak radiolarian ooze layer, marking the lithological boundary between pelagic clays below and turbidites above (Housen et al., 1996;Moore, 1982). We note this stratigraphic position is equivalent to that in North Sumatra (the boundary between pelagic and turbidite-rich fan sediments). The frontal Cascadia décollement offshore Washington forms in the deep sediment boundary between carbonate clays below and sandy turbidites above interpreted as a weak frictional layer and proposed to drive the observed landward vergence here (Adam et al., 2004;Gutscher et al., 2001). In general, in Cascadia, the position and properties of the décollement have been linked to forearc structure (Han et al., 2017;MacKay, 1995;MacKay et al., 1992).
For North Sumatra, we suggest that the depth of the décollement at the deformation front is specifically controlled by the enhancement of a pre-existing lithological boundary deep (∼4,000 mbsf close to the deformation front) within the incoming sediments. We propose that rapid accumulation of ∼2 km of sediment within the trench produces high temperatures in the basal sediments generating and expelling mineral-bound fluid through the smectite-illite transformation and leading to quartz overgrowth and calcite cementation in overlying sands. Liberated fluids get focused within a thin high porosity layer close to the boundary between the pelagic sediments (Unit C) below and Nicobar Fan sediments (Unit B) above, where cemented sand intervals act as an impermeable cap. This could generate high pore fluid pressure within the layer (e.g., Spinelli & Underwood, 2004), which would reduce the effective stress and enhance frictional weakness (Scholz, 1998), which could then be exploited by the initiating décollement. This subsequently promotes low basal shear stress along a stratigraphically deep frontal décollement, hypothesized to lead to the formation of landward-vergent structures in the outermost prism (Dean et al., 2010; Globally, décollement seismic reflectivity and polarity are interpreted as showing the presence/absence of change in physical properties across the décollement, either in terms of relative consolidation or pore pressure. For example, in Costa Rica and Nankai-Muroto, the décollement beneath the outer prism is marked by high amplitude, reversed polarity reflections, interpreted as high pore fluid pressures generated by diagenetic dehydration (Bangs et al., 2004;Ranero et al., 2008) representing the outer aseismic section of the plate boundary fault. The seismic properties are similar to those found in the input section offshore North Sumatra outboard of the subduction zone. In the case of North Sumatra, we show that the deepest 1-1.5 km of the sediment column has probably completed the smectite-illite transformation (Figures 5 and 7) and most diagenetic fluids have been generated and escaped before subduction. Therefore, once these sediments are accreted or subducted they do not provide a source of diagenetic fluid to resupply the décollement and are largely already quartz-cemented. In the North Sumatra outer prism, the décollement quickly loses its reflection amplitude and reverse polarity landward of the deformation front suggesting fluid rapidly escapes, for example along active prism thrust faults (Dean et al., 2010). This would remove or reduce fluid overpressure at the décollement, increasing plate coupling and promoting strain accumulation; and pervasive quartz-cementation in sands close to the décollement level would act to promote velocity-weakening behavior (e.g., Ikari et al., 2007). We interpret the décollement to have seismogenic potential at this point (Ellis et al., 2015;Ranero et al., 2008;Saffer et al., 2008;Spinelli and Saffer, 2004), within the outer part of the subduction zone (as also proposed by Dean et al., 2010;Gulick et al., 2011;Hüpers et al., 2017). The accretion of a thick, mostly dehydrated sediment column thus generates a strong and cohesive prism and décollement fault capable of transmitting elastic stress far seaward during an earthquake, widening the seismogenic zone and increasing the potential earthquake and tsunami magnitude (e.g., Spinelli & Saffer, 2004).
Our findings support the hypothesis that the thickness of the incoming sediment column has significant implications for its thermal structure and the fluid budget both landward and seaward of the subduction zone deformation front, and the frictional behavior of the décollement (e.g., Saffer et al., 2008;Spinelli & Underwood, 2004). Other subduction zone margins with thick sediment inputs and/or increased temperatures/ heat flow such as the Makran (Smith et al., 2013), and parts of the Cascadia Salmi et al., 2017), Hikurangi (Ghisetti et al., 2016), Lesser Antilles and Hellenic margins (Underwood, 2007), may experience similar pre-subduction sediment dehydration and resulting décollement properties and seismogenic behavior.

Conclusions
Ground-truthed sediment properties from IODP drilling, combined with multichannel seismic reflection data and seafloor heat flux measurements of the incoming sediments offshore North Sumatra, have allowed us to build a detailed model of the thermal structure and dehydration state of the incoming sediments. Our results indicate: (1) High temperatures within the deeply buried (∼5.5 km) basal sediments reach 125-160 °C at the ∼4.5 km depth of the décollement at the deformation front. This is predominantly due to gradual (but high rate) accumulation of the Nicobar Fan and trench wedge sediments (2) Anomalously high heat flux at Site U1480, ∼230 km west of the subduction zone can be explained by thermal rejuvenation of the lithosphere ∼10 Myr after its formation by magmatism associated with the NER (3) Anomalously low surficial heat flux within the trench is due to the suppressive effect of high rates of sediment accumulation (4) Sediment radiogenic heat production has a largely negligible effect on the thermal state of the sediments, but may offset the effects of sedimentation on heat flux through the sediment column by up to 6% for the thickest sediments (5) Dehydration over the last ∼1 Myr of the basal sediments beneath the horizon that develops into the décollement generates a sufficient volume of fluid to explain the observed reflection polarity reversal of this horizon close to the deformation front if these fluids are focused and trapped into a 70-80 m thick high porosity layer (6) Silica cementation coincident with the smectite-illite transformation that occurs through the deepest (>2,000 mbsf) sediments prior to their accretion/subduction. This acts to enhance prism cohesive strength and promote velocity-weakening behavior at the décollement in the outer prism (7) Offshore North Sumatra, combined lithological, diagenetic, and hydrological properties appear to control the position of décollement formation deep in the sediment section and at a prominent stratigraphic boundary. Significant pre-subduction dehydration of the incoming sediments excludes these sediments as potential sources of diagenetic fluid within the prism. This impacts the fluid budget of the subduction zone as a whole, and results in the décollement beneath the prism being well-drained and positioned within frictionally strong materials, which may extend the seismogenic zone close to the trench.