Tectonic interactions during rift linkage: Insights from analog and 1 numerical experiments 2

Abstract


Introduction
5 Pre-existing structures as well as fault interaction across multiple scales disturb the regionally 90 inferred stress orientation (Morley, 2010;Oliva et al., 2022). In return, stress re-orientations 91 within and adjacent to rift segments influence the style of progressive deformation. Ultimately, 92 stress re-orientation may even favor pure dip-slip behavior even for extensional faults with an 93 oblique orientation to the regional extension (e.g., Morley

169
The three seed segments hypothetically merge at the model center. However, we exclude 170 weak seeds in an area with a radius r = 10 cm around the model center to allow free 171 interaction of the propagating rift structures (Fig. 2b). The analog model comprises an initial 172 area of 80 cm by 30 cm and has a total thickness of 6 cm (each layer 3 cm) which represents 173 a 30 km thick continental crust. In accordance with the numerical setup, the effectively

Analog model results 179
In the analog model three different rift segments initiate above the weak seeds and propagate 180 toward each other. Thereby, the two rear segments compete for linkage with the frontal 181 segment. After 30 min (i.e., 5 mm extension; Fig. 3(i)), brittle deformation localizes along two 182 rift boundary faults forming the frontal rift segment. Rifting in the rear segments localizes first 183 along right-dipping rift boundary faults and after 60 min (i.e., 10 mm extension; Fig. 3(ii)) 184 both rear segments develop a set of two conjugate rift boundary faults (Fig. 3a,b (ii)).

185
Interestingly, instead of advancing straight forward, the fault tips deflect and propagate away 186 from each other (Fig. 3b,d (ii)). This is partially due to the rift propagation over the area where 187 no seeds are present where rifting perpendicular to the extension direction is favored.

Numerical model setup 217
We use the open source, finite-element code ASPECT to solve the extended Boussinesq  239 mm yr -1 , resulting in a total extension velocity of 10 mm yr -1 (Fig. 4a,b). After a total model

275
The red line indicates the initial depth of the brittle-ductile interface (as defined by the interface between the two rheological 276   (Zoback, 1992). The RSR value maps possible stress regimes to an interval between 0 and 3.

308
For isotropic and homogenous materials, the standard rules of Andersonian faulting are 310 applied (Anderson, 1905

338
Similarly, the right rear segment shows faulting along the right-dipping rift boundary fault but 339 activity along intra-rift faults is lacking. In the central model domain, formerly distributed 340 deformation localized between the frontal and left rear rift segment (Fig. 5d (ii)). While strain 341 rates indicate a shift from a symmetric to an asymmetric deformation phase, topography is 342 still symmetric which implies that the shift is imminent and has not affected the topography 343 after the first million years (Fig. 5b (ii)).

345
After two million years, deformation is entirely localized along the frontal and left rear

356
With ongoing extension, deformation subsequently localizes along the axial rift zone that links 357 the frontal and left rear segments (Fig. 5a,c,d (iv,v)) and faulting activity along rift boundary 358 faults ceases. The linked structure reaches maximum depth inside of the rift after three million 359 years. After four million years, however, the basin experiences minor uplift due to increase upward motion of the underlying viscous material (Fig. 5d (iv,v)). Note that the basin depth 361 of the right rear rift segment remains stable after two million years and does not experience

392
For experiment with a i-seed configuration (Fig. 6d-f) two opposingly propagating rift branches 393 form. Since the right rear segment is absent, both opposingly propagating rift segments link 394 in the model center where deformation is distributed onto intra-rift faults. The overall strain 395 rate field is localized, and no strain rate deflection occurs.

397
Models with a y-seed configuration (Fig. 6g-i) (Fig. 6a-c). The final model stage after four million years best illustrates differences in rift geometry 423 between the models with different seed geometry and an intermediate angle (Fig. 7). Rift 424 deflection is well visible in v-seed models (Fig. 7 a-c) and most prominent in experiments with 425 a larger intermediate angle (Fig. 7b,c). Above the seeds, two short individual rift segments    (Fig. 7g). Both competing rear seeds are close enough 453 such that they build one rift system rather than two distinct branches. For y-seed models with 454 a larger intermediate angle (Fig. 7h,i)  propagating rifts occurs when deformation is symmetrically distributed along both competing 507 rift branches. This is clearly visible for the v-seed configuration (Fig. 8a-e). Assuming 508 orthogonal extension and isotropic material properties, SHmax is expected to align perpendicular 509 to the extension direction producing pure dip-slip normal faults (Anderson, 1905). However, 510 the model shows an immediate SHmax re-orientation at early deformation stages (i.e., after 0.4 511 million years; Fig. 8a) from a N-S to a E-W orientation in the vicinity of the underlying weak 512 seeds such that dip slip faults are favored over oblique-slip faults with a strike-slip component.

513
With progressive extension (Fig. 8b-e), SHmax re-orientations successively propagate into the 514 isotropic zone without pre-existing structures, concomitant with the rift propagation. There is a distinct difference between stress deflection along weak structures and E-W 519 deflections of SHmax in zones where strain rates are below the set threshold of 10 -16 s -1 . The v-520 seed configuration shows localized strain accumulation along one rift boundary fault per 521 segment (i.e., the outer one) resulting in a rift zone with a broad graben system that subsides 522 (Fig. 8e). SHmax re-orientation inside of the graben is in parts identical to the E-W orientation

SHmax evolution in sub-parallel rift segments (i-seed models) 531
During the early stage (i.e., after 0.4 million years, Fig. 8f

Rift arm competition and deflection (y-seed models) 552
A prominent feature in our models with two competing rift segments is the deflection of rift 553 branches and arcuate strain rate patterns (Fig. 8a-e) in the model with a v-seed configuration.

554
Moreover, the i-seed configuration demonstrates a gradual SHmax re-orientation over a broader 555 pre-weakened zone due to formerly active boundary faults. One could therefore expect that 556 both features should occur in the model with y-seed configuration (Fig. 8k-o).

558
Indeed, early stages (i.e., after 0.4 million years; Fig. 8k) are characterized by a symmetric 559 stress field with re-oriented SHmax values near the two rear rift segments. However, in contrast 560 to the v-seed configuration, SHmax re-orientation also occurs near the frontal pre-existing weak 561 fabric along developing rift boundary faults. In the isotropic zone, SHmax values dominantly 562 show a N-S direction. The general N-S orientation reflects the regional stress field due to an 563 E-W extension as predicted by Anderson (1905) in isotropic areas, into which rift segments 564 have yet to propagate. With ongoing extension, all three rift segments propagate into the 565 isotropic zone and cause a re-orientation of SHmax (Fig. 8l). Note that after 0.8 million years 566 the stress re-orientation occurs symmetrically. This contrasts with the i-seed configuration 567 where SHmax values deflect into either an E-W orientation along active rift boundary faults or 568 gradually turn into a fault parallel direction over a broader weakened zone (see subsection 569 3.7.2.). The early symmetric stress distribution in the y-seed configuration model is 570 28 unarguably due to the symmetric seed configuration (see also Fig. 8a-e). At this stage, dip-571 slip faulting along the competing sub-parallel rift segments is favored over oblique slip faults 572 as in models with a v-seed configuration. It is only after 1.2 million years, when fault activity 573 along the right rear segment ceases that deformation localizes along the left rear and frontal 574 segments and linkage intensifies (Fig. 8m). Successively, localization and linkage occur 575 coevally with a switch from a symmetric to an asymmetric stress distribution and resembles 576 more the stress distribution in the i-seed configuration model (Fig. 8f-j). The model state after 577 1.2 million years (Fig. 8m) also marks the switch from a symmetric to an asymmetric stress

586
The symmetry switch is also visible in rose diagrams of stress orientations within the active 587 faulting zone (i.e., strain rate ≥10 -16 s -1 ). A dominantly N-S oriented SHmax distribution changes 588 to a bimodal distribution with a second E-W orientation (Fig. 8l-n). Similarly, bimodal SHmax 589 distribution is also visible in the experiment with an i-seed configuration but occurs earlier.

590
Since the experiment with an i-seed configuration is never in the state of an early symmetric 591 stress distribution linkage is facilitated and occurs earlier (Fig.8g-i).  Kattenhorn et al., 2000). These experiments suggest that earlier fractures lead to 618 subzones (within a broader damage zone), where stresses subsequently rotate away from the 619 regional stress field. Although our analog and numerical models do not feature elastic 620 deformation, they indicate that stress deflection is an ongoing process, even after elastic 621 material failure. Such a stress deflection further implies that stress orientations in rocks with 30 pre-existing weaknesses can substantially deviate from predicted orientations in isotropic 623 media (Anderson, 1905).

625
It has been proposed that early faulting and propagation in the Rukwa and North Malawi Rifts 626 (Fig. 1c) were guided by pre-existing basement fabrics (Heilman et al., 2019). This region is 627 further shaped by a flip in the boundary fault polarity in the present-day geometry within the 628 interaction zone between Rukwa Rift and North Malawi Rift (Bosworth, 1985). Our i-seed

661
We speculate that, while both rear rift segments in our y-seed models equivalently 662 accommodate strain in the early stages (i.e., when the overall stress distribution is symmetric;  i-seed configuration models, respectively. The switch from a symmetric to an asymmetric 669 stress distribution in y-seed models also marks the switch from a system that was formerly 670 dominated by the competing rear rift segments (i.e., v-seed configuration) to a system that 671 is dominated by the linkage of two obliquely oriented segments (i.e., i-seed configuration).

673 32
In models with a v-seed configuration, however, the symmetric phase prevails and causes 674 coeval SHmax re-orientations and rift deflection that cause divergence of the two propagating 675 rift segments. A similar process of extensional segment interaction via stress rotation is known 676 from mid-ocean ridge settings: Pollard and Aydin (1984) argue that paths of two opposingly 677 propagating oceanic ridges weakly diverge due to shear stresses that divert propagating ridges 678 as they approach each other. Once the two ridges overlap, the stress field changes causing  rear segment coinciding with jumps in the SHmax orientation (Fig. 8m-o). Our modelling results

698
show that stress deflection along rift segment tips is a mechanical consequence of the 33 interaction between weak zones and far-field stresses offering a potential explanation for 700 naturally occurring rift deflection. However, we must emphasize that complexities in natural 701 rift settings pose additional difficulties that require further investigations of stress orientations.

703
An example of rift deflection in nature has been described in the Main Ethiopian Rift.

725
We present a series of analog and numerical rifting experiments. Our results suggest that, 726 even in a relatively simple iso-viscous two-layer crustal setup, pre-existing weaknesses 727 substantially disturb the regional stress pattern, which impacts rift propagation and the overall 728 rift evolution. The complex stress re-orientation is distinct for different seed configurations 729 (i.e., v-seed, i-seed, and y-seed) and closely interacts with the final rift geometry. The most

755
While changes in rift orientation are often used to infer regional palaeo-movements, we 756 demonstrate that local stress field re-orientations can occur under constant plate motions.