Thermochronological insights into reactivation of a continental shear zone in response to Equatorial Atlantic rifting (northern Ghana)

West Africa was subjected to deformation and exhumation in response to Gondwana break-up. The timing and extent of these events are recorded in the thermal history of the margin. This study reports new apatite fission track (AFT) data from Palaeoproterozoic basement along the primary NE-SW structural trend of the Bole-Nangodi shear zone in northwestern Ghana. The results display bimodality in AFT age (populations of ~210-180 Ma and ~115-105 Ma) and length distributions (populations of 12.2 ± 1.6 and 13.1 ± 1.4 µm), supported by differences in apatite chemistry (U concentrations). The bimodal AFT results and associated QTQt thermal history models provide evidence for multiple cooling phases. Late Triassic – Early Jurassic cooling is interpreted to be related with thermal relaxation after the emplacement of the Central Atlantic Magmatic Province (CAMP). Early to middle Cretaceous cooling is thought to be associated with exhumation during the Cretaceous onset of rifting between West Africa and Brazil. Late Cretaceous – Cenozoic cooling can be related with exhumation of the Ivory Coast – Ghana margin and NNW-SSE shortening through western Africa. Furthermore, our data record differential exhumation of the crust with respect to the Bole-Nangodi shear zone, preserving older (CAMP) cooling ages to the south and younger (rifting) cooling ages to the north of the shear zone, respectively. This suggests that the Palaeoproterozoic BN shear zone was reactivated during the Cretaceous as a result of deformation in the Equatorial Atlantic region of Africa.

The NE-SW striking Bole-Nangodi (BN) shear zone in northwestern Ghana represents a Palaeoproterozoic crustal-scale shear zone within the West African Craton (WAC) 1,2 ( Fig. 1) that preserves evidence for a deformation overprint that coincides with the Palaeoproterozoic Eburnean Orogeny 3 . The BN shear zone also contains structural evidence for subsequent distinct deformation phases 3 , however, the absolute timings of these brittle deformation phases are unknown. This study aims to characterise the Phanerozoic reactivation history of the shear zone and its role in the exhumation history of NW Ghana.
South of, and parallel to, the BN shear zone, the Ivory Coast-Ghana (ICG) continental margin forms a NE-SW trending basement ridge, which represents the eastern prolongation of the Romanche Transform Zone in the equatorial Atlantic Ocean (Fig. 1a) 4 . Geochronological and thermochronological studies have revealed that the ICG marginal ridge formed during the Cretaceous in response to rifting between the WAC and Brazil during the opening of the equatorial Atlantic Ocean [5][6][7][8] . Recent studies 9 proposed that NE-SW oriented onshore faults in Ghana and the Ivory Coast are linked to an oceanic fracture zone 90 km South of the St Pauls transform zone in the Atlantic Ocean and that the eastern limit of the St Pauls Transform Zone can be linked with onshore faults in Ghana 10 . It is further suggested that the pre-existing shear zone may have controlled the orientation of the present-day oceanic transform zone 9 , however the theory remains inconclusive. Hence, this study hypothesises that the BN shear zone shares a common low-temperature thermo-tectonic history with the ICG marginal ridge. We present Apatite Fission Track (AFT) thermochronological data and associated thermal history models to constrain the low-temperature thermal and exhumation history of the BN shear zone, aiming to provide temporal    (Fig. 2a). The plot shows a broad distribution of single grain ages and a trend of younger grains associated with higher uranium concentrations and vice versa. which acts as a weighted distribution for the most likely or 'expected' tT path. Note that the confidence interval of the probability distribution for several models is rather large and, therefore, for those models the magnitude and exact timing of cooling is not always well constrained. The modelled samples display three distinct cooling periods (Fig. 3), which are defined by the three groups listed above. Models for samples in group 'SE of BN shear zone' exhibit rather slow cooling from the Triassic until present day temperatures. Models for samples in group 'within the shear zone' show fast cooling since the Late Triassic -Early Jurassic (ca. 200 Ma), followed by thermal quiescence until the end of the Cretaceous. The long residence of the samples in the APAZ during most of the Jurassic and Cretaceous likely explains the rather large single grain age dispersion observed for samples from within the BN shear zone (Fig. 2a). Since the start of the Cenozoic, the modelled tT history suggests increased cooling rates, bringing the samples above the APAZ during the Palaeogene (Fig. 3). Models for samples in group 'NW of BN-shear zone' exhibit fast cooling since the Late Jurassic -Early Cretaceous (ca. 150 Ma). Similar as for the samples from within the shear zone, thermal quiescence was observed, before the samples were brought to near-surface temperatures during the Palaeogene. The two samples that record the youngest age peaks for the study area (BN-127, P1 = ca. 56 Ma, BN-132 P1 = ca. 39 Ma) were not modelled due to insufficient available AFT length data. However, the P1 ages obtained for those samples coincides with the timing of the final cooling step in the thermal history models (ca. 60-35 Ma). Individual models, modelled track length distributions and modelling parameters can be found in Supplementary Fig. S5 and Supplementary Table 2.
In summary, a Late Triassic-Jurassic cooling signal and a middle Cretaceous cooling signal were modelled, consistent with the two distinct age populations. In addition, late Cretaceous -Palaeogene cooling is recorded for several thermal history models, confirmed by the youngest AFT age peaks in the study area. However, this modelled Palaeogene cooling signal could represent a modelling artefact, induced by constraining the model to present-day ambient temperatures, and therefore caution is required during interpretation in further discussion.
The role of uranium on the recorded AFT data. The thermal history models display a rather complex thermal history record ( Fig. 3) with evidence for multiple cooling phases. The identified AFT age-populations (i.e. P1 and P2) preserve different parts of the thermal history record ( Table 2), suggesting that apatite grains associated with each population record cooling at different temperatures (i.e. upper and lower APAZ temperatures). In order to strengthen this observation, fission track length distributions were compared to AFT ages. If the different age-populations indeed preserved different parts of the thermal history, they could likely record different fission track length distributions.  AFT results plotted as radial plots. Central age values were calculated using RadialPlotter and for dispersions >25% and P(χ 2 ) less than 0.05 age peak discrimination was performed using the RadialPlotter software 51   For each of those samples, confined track lengths related to AFT age populations P1 are displayed in a blue frequency plot and those related to AFT age populations P2 are displayed in a red frequency plot (Fig. 4a). The MTL for confined lengths recorded in apatites related with population P1 (based on 223 length measurements) is 12.21 ± 1.56 µm. The MTL for apatites associated with population P2 (based on 140 length measurements) is slightly higher at 13.10 ± 1.36 µm (Fig. 4a), suggesting that P1 grains experienced more extensive AFT annealing than P2 grains.
In addition, the relation between AFT age, confined track length and 238 U concentration was evaluated (Fig. 4b). As shown, there is a positive correlation of P2 ages associated with longer confined track lengths and lower 238 U concentrations, while younger (P1) ages correlate with shorter track lengths and higher 238 U concentrations. There are no clear correlations observed between AFT ages and Cl concentrations 30 (Fig. 2b), nor between AFT ages and Dpar measurements 31 (Supplementary Fig. S6). We, therefore, suggest that the concentration of uranium controls AFT annealing in our study area. The relation between uranium concentration and AFT age has been observed for other, unrelated, study areas as well [32][33][34] . 35 Suggests a model where elevated concentrations of Uranium lead to radiation enhanced annealing. More recently, 35 observed, using transmission electron microscopy measurements, that alpha particle induced annealing of radiation damage indeed occurs in apatite. This study illustrates that the effect of radiation-induced annealing needs to be evaluated in fission track studies.

Discussion
Late Triassic to Early Jurassic cooling. We propose that the Late Triassic -Early Jurassic cooling of the Ghanese upper crust is related to thermal relaxation after emplacement of the Central Atlantic Magmatic Province (CAMP) (Fig. 5a). Geochronological and 40 Ar/ 39 Ar studies that measure the emplacement ages for CAMP typically reveal a short duration for the most intense period of magmatism between 202 Ma and 189 Ma in the central and northern WAC, while in western Sierra Leone and towards the east, in Nigeria, manifestations of CAMP related volcanism span between ca. 234 Ma and ca. 140 Ma 15,18,19,21 . Although outcrop evidence for CAMP related volcanism is limited within the study area, the large extent of CAMP emplacement over four continents including Europe, South America, North America and Africa (Fig. 5) indicates that CAMP related heat flow likely also affected the study area in Ghana 14,15,20,36 . In addition, the Jurassic is the time when the extensive Karoo Large Igneous Province (LIP) and Basin in southern Africa formed 36 , which initiated the development of the so-called African erosion surface 37 . We suggest that the emplacement of LIPs such as CAMP caused a temporary increase in the geothermal gradient and the AFT ages in Ghana record the subsequent cooling (Fig. 3). Similar interpretations were made for other thermochronological studies conducted in NE Brazil 38,39 , juxtaposed to the Ghana margin and within the extent of CAMP. The study by 38 reports AFT and Zircon Fission Track (ZFT) ages of approximately 200 Ma and interpret slow cooling related to the decline of magmatic activity in the region. The study by 39 reports AFT ages of approximately 200 Ma and relate the data to crustal heating, induced by CAMP. Alternatively, 40 describe substantial Permian -Jurassic plateau exhumation in NE Brazil to be related with erosion of topography that was generated during Gondwana amalgamation. As a result of continuous rifting along the African/South American margins prior to the final stages of rifting in the Equatorial Atlantic, this study assumes a passive rift model throughout the study area allowing for above average geothermal gradients 42 . The thermal history models for samples NW of the BN shear zone reveal associated fast cooling during the Early and middle Cretaceous (Fig. 3), coinciding with the earliest stages of intra-continental break up, which resulted in rifting at the equatorial Atlantic during the middle Cretaceous 4,[7][8][9][10]22,41,43 . The Early to middle Cretaceous cooling event is therefore interpreted in terms of exhumation in response to the two initial geodynamic stages of break-up.  Central age values are calculated by RadialPlotter and for dispersions >25% age peak discrimination was performed using the RadialPlotter software 51 version 8.3 http://www.ucl.ac.uk/~ucfbpve/radialplotter/. Percentage of data associated with each peak is bracketed adjacent to the age of each peak. The X-axis represents the amount of 238 U in ppm. The left Y-axis represents 2 standard deviations from the central age (Ma). The right 'curved' Y-axis shows increasing age in Ma. 238  Late Cretaceous -Cenozoic Cooling. Our thermochronological data and models provide minor evidence for a third cooling phase that affected the study area during the late Cretaceous -early Cenozoic. Late Cretaceous to early Cenozoic AFT ages ( Table 2) were measured for samples located in the central BN shear zone and in vicinity to late brittle faults (D6 and D7 deformation phases) as recorded by 3 . Published off-shore AFT data from 7,9 reveal cooling ages ranging between 65 Ma and 92 Ma, which are interpreted to constrain the final stages of break-up and exhumation of the ICG marginal ridge (Fig. 5b). We interpret the late Cretaceous -early Cenozoic cooling along the BN shear zone to be related with the exhuming ICG margin and/or with folding and stike-slip faulting as response to NNW-SSE shortening within western Africa 26,27 . Differential exhumation with respect to the Bole-Nangodi Shear Zone. The BN shear zone acts as the most significant structure within the study area and is oriented parallel to the modern expression of transform zones in the equatorial Atlantic Ocean (Fig. 1). A gridded interpolation map (Fig. 6) was constructed to assess the role of the shear zone in relation to the cooling history of the study area. This map reveals distinctively different cooling ages for samples northwest of the BN shear zone compared to samples southeast of the BN shear zone, which is interpreted as a record of differential exhumation with respect to the shear zone. More specifically, our data suggests that the BN shear zone was reactivated during middle Cretaceous extension, exhuming the north-western block (footwall) with respect to the subsiding south-eastern block (hanging wall, comprising the Volta Basin). Therefore the north-western block exposes a deeper section of the northern Ghanese thermal history, compared to the south-eastern block. This record of differential exhumation supports the interpretation by 10 that the BN shear zone acted as a control on the NE-SW orientation of the Atlantic transform fault that formed during Gondwana break-up (Fig. 6). Structural observations within the study area describe deformation phase D5 as localised E-W to NE-SW striking brittle deformation associated with dextral strike-slip shearing under a transcurrent regime 3 . Given the excellent correlation of the D5 fault trend with the significant step in AFT ages on the gridded AFT map (Fig. 6), this work dates the D5 brittle deformation phase to the early-middle Cretaceous. The D5 fault trend can be extrapolated to the St-Paul fracture zone, suggesting that strain may have partitioned towards the Bole-Nangodi shear zone during rifting and transform fault development (Fig. 6c). The youngest AFT central ages measured in this study are located in the central section of the BN shear zone (Fig. 6) where there is a significant wedge-shaped geometry. This region is bound by faults of either D5 or 'Late Fault' relative ages. Therefore, the 'Late Faults' are interpreted to be related with the youngest AFT ages for this study (Late Cretaceous -Cenozoic). Given the similarities in AFT age data and structural setting, this work further relates the tectonic interpretation for the study area with those obtained for the ICG margin and the Gombe Fault in the Upper Benue Trough 4,23,24,27 . According to field studies, stratigraphic correlations and (for the case of the ICG margin) low temperature thermochronology, both the Gombe Fault and ICG margin were active and exhumed during the Late Cretaceous (towards the end of the Equatorial Atlantic rifting period) as the equatorial margin became passive 7,9,26,27 . Previous studies conducted by 45 have noted similar styles of reactivation along pre-existing shear zones in the South Tien Shan (Central Asia). In light of these relationships, we suggest that the youngest region of the study area (Fig. 6), located in the central BN shear zone, is interpreted as having undergone minor Late Cretaceous to early Cenozoic exhumation in response to a compressional stress regime, as discussed above 26 . We, therefore, interpret the 'Late Fault' stage brittle faults located at the central BN shear zone near samples BN-127 and BN-016 to represent step-over faults or restraining bends. These structures are typically associated with strike-slip faults of this nature 46 . Further work would be required to confirm this interpretation, as structural field observations around brittle deformation zones within the study area are limited.

Conclusions
Apatite fission track thermochronological results from northern Ghana characterise the thermal and deformation history of the NE-SW striking Bole-Nangodi shear zone and the surrounding area in context with continental rifting between West Africa and Brazil. Our thermal history models suggest that the Ghanese crust cooled during the Late Triassic -Early Jurassic, which post-dates the emplacement of CAMP and coincides with the development of the Karoo Basin and LIP. The cooling signal further corresponds to the time of the development of the African erosion surface. Samples to the northwest of the Bole-Nangodi shear zone record early to middle Cretaceous cooling, contemporaneous with the initiation of rifting in the equatorial Atlantic. Our data suggest that the shear zone was reactivated at that time, exhuming northwestern Ghana with respect to the southeast. Late Cretaceous -Cenozoic cooling was revealed for samples in the vicinity of late brittle faults and coincides with the timing of kilometre-scale exhumation in the Ivory Coast-Ghana margin and NNW-SSE shortening in western Africa. Our study furthermore suggests that strain associated with the St-Paul transform fault may have propagated into the continental interior to reactivate the Bole-Nangodi shear zone (Fig. 6c), demonstrating the role of inherited on-shore structures during Equatorial Atlantic rifting.

Methods
Apatite Fission Track thermochronology. Apatite Fission Track thermochronology is based on the spontaneous fission decay of 238 Uranium and is used to constrain the low-temperature (~60-120 °C) thermal history through the apatite fission track partial annealing zone (APAZ) 47 . For this study, apatite grains were picked, mounted in epoxy resin, ground and polished to expose internal apatite sections. The apatite mounts were etched in 5 M HNO3 for 20.0 ± 0.5 s at 20.0° ± 0.5 °C to expose the spontaneous fission tracks and subsequently imaged on a Zeiss AXIO Imager M2m Autoscan System at The University of Adelaide. Fission track densities, Dpar and lengths were measured using the FastTracks software. Where possible, a minimum of 20 grains were analysed per sample. The objective was to measure at least 100 confined fission track lengths, a target that was not always achieved. The 238 U and 35 Cl concentration of each counted grain were measured using spot analyses on a New Wave UP213 laser connected to an Agilent-7500cs ICPMS 48 . U concentrations were determined by measuring 238 U/ 43 Ca ratios, calibrated against NIST610, NIST612 and Madagascar apatite standards (U-Pb age of 473.5 ± 0.7 Ma 49 ). Age calculations were carried out following 50 using a zeta approach 51 with Durango as secondary age standard ( 40 Ar/ 39 Ar age of 31.44 ± 0.18 Ma 52 ). A total of 102 measurements on Durango apatite were carried out for this study, spread over 3 different (well-characterised and homogeneous in U) crystals with mean U concentrations of 15.54 ± 0.08 ppm, 12.67 ± 0.07 ppm and 9.80 ± 0.10 ppm. The resulting mean AFT age of 31.65 ± 0.7 Ma for Durango, obtained in this study, is in excellent agreement with the published age. Accuracy of the applied methodology against the traditional external detector method is demonstrated in 53 35 Cl measurements were carried out following 54 , who were able to demonstrate that electron probe and LA-ICP-MS data are the same when analysing a number of different apatites with different 35 Cl compositions. Durango was used as accuracy check, producing a mean Cl concentration of 4.4 ± 0.1 ppm, which is in good agreement with published values. Data reduction was performed using the Trace_Element_IS data reduction scheme in the Iolite software 55 . More details on the methodology, can be found in 33,50 . Data Presentation and Modelling. AFT central ages were calculated with RadialPlotter 56 and represent the apparent AFT cooling age of the analysed samples. This central age is not necessarily a good estimate of a cooling event and therefore subsequent thermal history modelling is needed to derive the tT history preserved in the sample. Furthermore, it is possible to (partially) preserve multiple AFT cooling ages or 'peak ages' in samples which fail Pearsons χ 2 test or where single-grain ages show significant dispersion (above an arbitrary value of 25%) which is regarded as a beyond natural spread 57 . In more detail, radial plots often show an open-jaw display of an older event and younger age component that can be related to distinct cooling events 58 . This is especially the case when two apatite populations have significantly different chemistries. In this regard, chlorine concentrations are often used as a chemical discriminator 30 . Elevated chlorine concentrations are known to slow fission track annealing but often show only little differences between apatite populations. This study uses uranium concentrations in addition to chlorine concentrations as a discriminator, which is thought to have a similar effect on fission track annealing. In this regard 34 , described a model of radiation-enhanced annealing for apatites with elevated uranium concentrations from old cratonic rocks. More recent observations 35 ; confirm that alpha-particle induced annealing of radiation damage occurs in apatite and should be considered alongside with the better understood thermal annealing. Thermal history (time-temperature) modelling was performed using the software package QTQt which uses the Bayesian trans-dimensional Markov Chain Monte Carlo statistical method to derive the most likely thermal history models 59 . The input parameters required in our modelling approach include individual AFT ages, their standard deviations, confined track lengths, sample elevations and Dpar used as a proxy for apatite chemistry as kinetic parameter 59 . The final output of each modelling attempt resolves three different models (Max. Likelihood, Max. Posterior and Max. Mode) and generates a weighted mean model or 'Expected' model based on a probability function at each point in the time-temperature space. The probability function is visualised in a colour spectrum ranging from blue (low probability) to red (high probability) where the absolute numerical probability value varies per model ( Supplementary Fig. S5). Further details on the QTQt modelling approach and modelling outputs, can be found in 59 . For comparison purposes, a smooth spline was fitted through the 'maximum mode' for each model output (Supplementary Fig. S5). (Supplementary Fig. S5). These splines are the best representations of the most likely tT path and are compared to each other in Fig. 3. Model parameters are summarised in Supplementary Table 2, following the recommendations by 60 .

Data Availability
All data generated or analysed during this study are included with the initial submission of the article in the form of Supplementary Information and are available on request.