Paleomagnetism from multi-orogenic terranes is " not a 1 simple game " : Pyrenees ' Paleozoic warning 2

Paleomagnetism is a versatile tool in the Earth sciences: it provides critical input to geological time scales and plate tectonic reconstructions. Despite its undeniable perks, paleomagnetism is not without complications. Remagnetizations overprinting the original magnetic signature of rocks are frequent, especially in orogens which tend to be the areas with better rock exposure. Unraveling the magnetic history of the rocks is a complicated task, especially in areas that underwent several orogenic pulses. In turn, constraining the timing of remagnetization represents an opportunity to solve post-magnetization structural and tectonic kinematics. Here, we evaluate the magnetization history of Silurian-Devonian carbonates from the Axial Zone of the Pyrenees. The Pyrenees are a multi-orogenic mountain belt where Silurian-Devonian rocks have seen the Variscan collision (late Paleozoic), the opening of the Atlantic / Bay of Biscay (early Cretaceous) and the Alpine orogeny (late Cretaceous to Miocene). Our results show widespread remagnetization(s) carried by magnetite and pyrrhotite in the Silurian-Devonian series of the Pyrenees. The majority of the samples show a post-folding but pre-alpine tilting magnetization. Considering the equatorial inclinations found in such samples, we suggest that they likely acquired their magnetization during the late Carboniferous and early Permian times. Two of the studied sites (located at the western Axial Zone) were subsequently remagnetized at the end of the Alpine orogeny. The paleomagnetic results constrained that the Variscan orogeny was responsible for the main folding event affecting Paleozoic rocks in the Axial Zone, whereas the Alpine orogeny produced the large-scale thrusting and antiformal stacking of these units. In addition, we observed a general clockwise rotational pattern which could be related with the formation of the Cantabrian Orocline and/or rotations associated with the Alpine orogeny. The Silurian-Devonian carbonates are thus useful to understand the tectonic evolution of the Pyrenean mountain range after a systematic combination of paleomagnetism with structural and petrological observations. In contrast, the secondary character of magnetization and complications associated with the Variscan tectonics indicate that a reassessment of Siluro-Devonian poles from the Variscan elsewhere in Europe might be appropriate.


42
The Earth's magnetic field has left a remnant signature in the geological record through eons. These 43 magnetic signals in the rock archive have been crucial to almost any field of Earth Sciences, from the 44 development of plate tectonics (e.g., Vine and Matthews, 1963), to the development of global time 45 scales (e.g., Kuiper et al., 2008) or the origin and evolution of the core (e.g., Biggin et al., 2015). 46 Paleomagnetism is still the only available technique that can quantify pre-Jurassic paleolatitudes 47 (Domeier and Torsvik, 2019), intensities of the past magnetic field, or global reference times 48 through reversals. The paleomagnetic imprint in rocks can last billions of years but may be also 49 fragile. For example, remagnetizations that overprint or even delete the original magnetic signature 50 are ubiquitous, especially in orogenic belts (e.g., Pueyo et al. 2007Pueyo et al. , 2016a; Van  synorogenic sedimentation processes that evolved to post-orogenic and intracontinental style basins 134 during the Permian (e.g., Oliveira et al., 2019). 135 The trend of the Variscan belt in north Iberia follows a "C" shape known as the Cantabrian 136 Orocline (e.g., Pastor-Galán et al., 2020). The Cantabrian Orocline seems isoclinal, formed by a 137 northern and a southern E-W-trending limbs, but this is likely the product of a retightening during

294
We drilled in 19 limestone sites from the Silurian or Devonian, one site of a late Carboniferous-early 295 Permian granite (OG01, Panticosa intrusion) and one Permian dyke that intruded the surroundings 296 of site OG12 (OG12dyke; Fig. 1B; Table 1) with a petrol-powered drill, in total 240 cores. We also 297 collected 6 oriented hand samples (from the OG07 and OG08 sites, three samples each). Sites are 298  Tait et al. (2000). Several sites allowed field tests: five site-scale fold-tests 302 could be obtained (OG2; OG11; OG13; OG14; OG19); two tilt tests between sites within the same 303 thrust unit (OG3-4; BN1-OR15), and two sites with a baked contact test (OG12 and OG17). We 304 performed all analyses at Paleomagnetic Laboratory Fort Hoofddijk, Universiteit Utrecht, The 305 Netherlands. 306 Our sample collection comes from fresh, non metamorphic and weakly or non-internally deformed 307 sites. Limestones were sedimented in shallow waters usually associated with clastic and pelitic 308 sediments (Casas et al., 2019). Most limestone sites contained variable amounts of organic matter 309 visible while drilling. A few of these sites show a spaced solution cleavage and evidence of 310 recrystallization. We collected the bedding orientation and, when observable, that of the pressure-311 solution cleavage (Table 1). 312

Alpine tilt estimation 313
The sampled rocks were affected by both Variscan and Pyrenean orogenies (Alpine). In the Axial Its geometry has been extensively investigated in previous studies that reconstruct the Axial Zone 316 structure using a combination of surface data and seismic profiles (Labaume et al., 1985;Cámara and 317 Klimowitz, 1985;Muñoz, 1992;Teixell, 1996 Teixell, 2018). These studies reveal that the Axial Zone evolves laterally from an imbricate thrust 320 system in the West (Fig. 2a, Teixell, 1996) to an antiformal-stack in the East (Fig. 2c, Muñoz, 1992). 321 Alpine thrusts are related to kilometric-scale basement folds in their hangingwalls. Basement folding 322 is well recorded by the Mesozoic units unconformably overlying the Paleozoic rocks and resulted in 323 dominantly southward and northward tilts in the southern and northern part of the Axial Zone, 324

respectively. 325
We used the available geological maps (GEODE, Robador et al., 2019) , including numerous 326 bedding data, and published cross-sections to discriminate the effects of Variscan and Pyrenean 327 orogenies and estimate the alpine tilt related to basement thrusting in our sampling sites. In the sites 328 located in the southern part of the Axial Zone, alpine dip directions and dips (Table 1)  Axial Zone (OG11 to OG19) are to far away from Mesozoic cover units. In these sites, we defined 332 dip directions as perpendicular to alpine thrusts in the Paleozoic rocks. Following the dip direction 333 we projected our sampling points into the traces of previously published cross-sections to obtain the 334 dip (Fig. 2). Alpine dips were also estimated considering the Mesozoic geometries reconstructed 335 above the topography in such cross-sections (Fig. 2). Figure 2 provides a general picture of the 336 structural position of the sampled sites (see corresponding thrust units in Table 1) and the regional eventually also a key to unravel magnetization timings. In this research we have performed a series 353 of rock magnetic analyses to fully characterize the magnetic mineralogy of the studied samples as a 354 step towards understanding the magnetization process and timing. We measured 18 high-field 355 thermomagnetic runs in an in-house-built horizontal translation-type Curie balance with a sensitivity 356 of approximately 5*10 -9 Am 2 (Mullender et al., 1993) and one in an AGICO KLY-3 susceptibility 357 bridge with a CS2 furnace attachment with nominal sensitivity (5*10 -7 SI) and air forced into the 358 tube. This latter analysis included two heating and cooling cycles. We also analyzed 20 magnetic 359 hysteresis loops and one first order reversal curve (FORC) diagram. They were measured at room 360 temperature with an alternating gradient force magnetometer (MicroMag Model 2900 with 2 Tesla 361 magnet, Princeton Measurements Corporation, noise level 2 × 10 −9 Am 2 ). Finally, we obtained 88 isothermal remanent magnetization (IRM) acquisition curves from our Pyrenean limestone samples.  Anisotropy of magnetic susceptibility (AMS) measures the induced magnetization in a rock when 381 applying a magnetic field in different directions, defining an ellispsoid (e.g. Parés, 2015). The shape 382 of the AMS ellipsoid depends on the crystallographic preferred orientation of the minerals; the 383 shape, size, and preferred orientation of mineral grains; the occurrence of microfractures, its not visually obvious, but it is also a powerful method to investigate the effect of deformation on the 386 NRM. We determined the composite fabric of the paramagnetic, diamagnetic and ferromagnetic 387 grains by measuring the AMS of 148 samples from our collection with an AGICO MFK1-FA 388 susceptometer (nominal sensitivity 2*10 -8 SI). 389 a sudden increase at ~300°-320° C followed by a sharp decrease afterwards (OG8 in Fig. 3; OG06 402 and OG07 in SF1). All samples showing pyrrhotite contained a less important, but observable, 403 content of pyrite. In the susceptibility vs. temperature curve (Fig. 3, BN1), pyrrhotite is observable 404 during cooling but it likely formed as a secondary mineral during one of the heating cycles. 405

Magnetic hysteresis 406
Representative samples with masses ranging from 20 to 50 mg were measured using a P1 phenolic 407 probe. Hysteresis loops were measured to determine the saturation magnetization (Ms), the 408 saturation remanent magnetization (Mrs), and the coercive force (Bc). These parameters were 409 determined after correcting for the paramagnetic contribution. The maximum applied field was 0.5 410 T. The field increment was 10 mT and the averaging time for each measurement was 0.15 s. We 411 found different loop shapes (Fig. 4, SF-2): (i) Loops that do not saturate at 0.5 T with a pseudo-412 single-domain like shape which points to the presence of a relatively hard magnetic carrier likely 413 pyrrhotite (Fig. 4, OG08) and (ii) typical magnetite-like pseudo-single domain loops (Fig. 4, OG19). 414 We performed a first order reversal curve (FORC) diagram (Fig. 4, BN1-3) that shows a mixture 415 between superparamagnetic and single domain behaviour. 416

Isothermal Remanent Magnetization (IRM) 417
Before the actual IRM acquisition, samples were AF demagnetized with the static 3-axis AF protocol 418 with the final demagnetization axis parallel to the subsequent IRM acquisition field, a procedure that 419 generates IRM acquisition curves with a shape as close to a cumulative-lognormal distribution as After VRM removal, the samples show a single NRM component (Fig. 6), generally trending to the 475 origin, regardless of the mineral, magnetite (usually fully demagnetized at 40-60 mT and 500-580° C) 476 and/or pyrrhotite (fully demagnetized at 330° C and little to barely demagnetized in AF). This 477 characteristic remanent magnetization (ChRM) clusters well in all the sites (concentration parameter 478 k > 8, but generally over 15; Table 2; Supplementary file SF3-B) with the exceptions of sites OG02 479 and OG08 (Figs. 8 and 9; Table 2). In addition, there are three sites with less than seven samples  (Tables 1 and 3). 488 In geographic coordinates sites that passed the quality filter show clusterings that range from k 489 (concentration parameter) ~ 8 (OG10 and 11) to k ~188 (OG19). Site average declinations range 490 from 125° to 297°, the majority of them in the south quadrants with sites OG15 and BN1-OR15 491 (combined sites separated by 100 m) being the only exceptions ( Fig. 8 and 9; Table 2). Inclinations 492 range from -50° to 50°. Bedding correction significantly changes the distribution of the site averages, 493 but the scattering in declinations (from 111° to 289°) and inclinations (-65° to 56°) remains (Table  494 2), which means that magnetization timing is not the same for all samples and/or structural 495 complications are larger than folding. 496 OG13 and OG19) were performed in folds with weakly plunging axes (Table 1; Fig. 11) with the 498 exception of OG11, which in turn is the only one that is not negative (Fig. 10). OG11 shows a 499 better clustering (t1) after tilt correction, however, the fold-axis in site OG11 is steeply plunging (the 500 only case; Fig. 10). The performed fold-tests restore deformation as if axes were horizontal. Steeply 501 plunging axis' folds, therefore usually yield false positives and negative foldtests (e.g. Pueyo et al., 502 2016a). A possibility to decipher the magnetization timing is pre-correcting the plunge of the fold 503 axis before the fold-test. The declination results, in this case, will bring spurious rotations and 504 wouldn't be trustworthy. After back-tilting the plunging-axis in OG11 (azimuth/plunge = 005/42), 505 the fold test remains indeterminate, in this case with a greater clustering (t1) before tilt correction 506 (Figs. 10 (structural correction panel) and 11). The statistics of all sites that pass our quality criteria 507 (n ≥ 7 and k > 8) yield close to random distributions both considering all specimens (k = 1.63 and 508 K = 1.65; Supplementary File SF3-A) and the mean of site averages (k = 1.95 and K = 2.5; Table 4). 509 As expected from the negative within-site fold-tests, the concentration parameter does not change 510 after bedding correction neither in all specimens together (k = 1.51 and K = 1.59; Supplementary 511 File SF3) nor the mean of site averages (k = 1.57 and K = 1.64; Table 4). 512 Inclination only statistics are independent to differential vertical axis rotations since declinations are 513 not taken into account (e.g. Enkin and Watson, 1996). Inclination only statistics were performed on 514 site averages to avoid weighting based on number of specimens (Table 4). The concentration 515 parameter (k) equals to 0 in geographic coordinates and 2.17 in tilt corrected coordinates, both 516 figures representing very poor clusterings. k becomes close to 4 if we do not account for OG3 and 517 OG4, which follow a different trajectory and may represent a different magnetization event (see Alpine tilt, the inclination only concentration parameter is ~4, but becomes ~14 when excluding 520 OG3 and OG4 (Table 4). An inclination only tilt test without OG3 and OG4 shows a best fit for a 521 110% correction, both using the Enkin and Watson, 1996 approach (with a maximum clustering 522 around k ~12) and a stepwise untilting following Arason and Levi (2010) inclination only statistics 523 with a maximum at k ~15 (Fig. 12). OG3 and OG4 share a common true direction in geographic 524 coordinates and their Alpine tilt correction does not change them too much (Fig. 13) 525

526
We measured the anisotropy of the magnetic susceptibility in 148 samples from most sites to 527 explore possible causes for the variety of ChRM directions found ( Table 2). The degree of 528 anisotropy (P) appears to be generally low (< 1.05; Fig. 14  The paleomagnetic and rock magnetic results obtained from the Silurian-Devonian limestones in the widespread remagnetization event. Many samples contain a VRM that is similar to the recent 541 geoaxial dipole for recent times in the Pyrenees (Fig. 7). Apart from this VRM, all rocks, regardless 542 of their magnetic mineralogy, show a single stable component heading to the origin, with the 543 exception of samples not delivering results ( Fig. 6 and 8). This component is not deviated toward 544 bedding/cleavage planes (as inferred from AMS patterns, Fig. 14) and displays negative fold tests 545 Pyrenees. The two end-member model with a reasonably high r 2 value of 0.65 is our preferred 555 model. Models with 3 to 9 end members evidently show slightly better fits (r 2 = 0.73 to 0.88 556 respectively). However, neither the fit improves significantly, nor the shape of the end members 557 shows more or less anticipated IRM acquisition curves for any particular mineralogy (Fig. 5B). In 558 addition, most of the additional end members seem to represent the variable coercivity windows of 559 magnetite (e.g. the 4 endmember solution in Fig. 5B: three of the endmembers (EM1-3) represent 560 magnetite and do not deliver any particular meaningful result). All samples contain a significant 561 amount of those additional endmembers (varying from 10% to 60%) indicating that a variable grain-562 size or compositional (Ti-)magnetite is present in virtually every sample. The two end-member model further distinguishes a 42 mT component with a DP = 0.36 (C1), which is typical for 564 magnetite and a 200mT component and DP = 0.35 (C2), which we interpret as pyrrhotite (Fig. 5C). 565 The two end members are in agreement with individual sample fits, but end-member IRM 566 acquisition curves describe much better the IRM properties of each magnetic phase (Fig. 5). We 567 interpret the soft component (C1) as magnetite varying from coarse to very fine grained (i.e. lower 568 and higher coercivity respectively) as supported by hysteresis loops (Fig. 4 and SF2). It is reasonable 569 that the high coercivity component (C2) reflects the observed pyrrhotite in the thermomagnetic 570 curves as SD pyrrhotite has a rather high coercivity (Dekkers, 1989). In pyrrhotite, the magnetic easy direction is confined to the basal crystallographic plane which 596 implies an intrinsically strong anisotropy because of the 'hard' crystallographic c-axis (Schwarz and 597 Vaughan, 1972; Schwarz, 1974). When pyrrhotite grows oriented in a preferred fabric (e.g. S1 or S0), 598 the direction of the magnetic remanence can be biased towards the fabric plane (Fuller, 1963). The 599 studied samples occasionally show pressure solution cleavage and a widespread presence of 600 pyrrhotite as a partial or main carrier of the NRM. AMS fabrics revealed that pyrrhotite is not 601 oriented according to the S1 fabric, and therefore preclude major biases in the magnetic remanence 602 of such secondary sulfides. Besides, although our AMS results are frequently consistent with bedding 603 (S0), the magnetic remanence is not contained within bedding planes ( Fig. 14; Table 2) which also 604 impedes a deviation of the ChRM towards S0.  ChRMs are not within bedding planes) suggesting that any potential bias caused by a preferred 618 orientation of pyrrhotite particles is not significant. Therefore, a post Variscan folding (i.e. late 619 Carboniferous) is the oldest possible age for the magnetization since no earlier folding event has 620 been described in the Pyrenees. The OG11 fold, which has a steep plunging axis (Fig. 11), yielded an 621 inconclusive fold-test, but shows better clustering after tilt correction (Fig. 10). Classical fold-tests 622 assume horizontal axes and performing them in steeply plunging axis' folds is unreliable (e.g. Pueyo When considering all site paleomagnetic directions, we found that declinations and inclinations are 630 only compatible in geographic coordinates for those sites within the same tectonic unit (Figs. 1, 2, 8,  631 9; Table 2). Such a directional pattern may be indicative of (a) different timing of NRM acquisition differential tilting between units or (d) a combination of the previous processes. To distinguish 634 between these options inclination only statistics are appropriate (e.g. Enkin and Watson, 1996). 635 Inclination only statistics do not consider declinations and therefore are independent of variations 636 due to differential vertical axis rotations. In order to evaluate potential timing of magnetization we 637 performed statistical analyses in geographic coordinates, after bedding correction, and also after 638 correction of the tilt related to the emplacement of Alpine basement thrusts (Figs. 2, 12, 13; Tables 3  639 and 4). The concentration parameter of inclination data (k) is 0 in geographic coordinates (Table 4), 640 which could mean that (i) sites magnetized at significantly different geological times when Iberia was 641 at very different latitudes, and/or (ii) Alpine tilting postdates the magnetization and therefore it has a 642 strong influence on the inclinations. Inclination only k is still too low (minimum k ~ 8 to consider 643 an acceptable clustering) after bedding correction (~2) but also after Alpine tilt correction (~4). 644 Two sites from the Gavarnie thrust unit (OG03 and OG04; Fig. 8 Table 4). After the Alpine tilt correction, the mean inclination is 5° ± 11 (Table 4) and all 654 included sites show SW to SE declinations (Fig. 12). Despite the positive result, our Alpine tilt 655 correction should be taken cautiously. We considered only regional tilt values, which were inferred 656 from the average orientation of the overlying Mesozoic units and thrust slopes in geological maps and cross-sections (Fig. 2). Our estimated values took into account the kilometric-scale, thrust-658 related folding of the basement but can not consider the potential contribution of Alpine, outcrop-659 scale folding of the Paleozoic strata. 660 We believe, however, that the inclination only k value of 14.42 together with the obtained shallow 661 inclinations and southerly declinations are sufficiently convincing to argue for a common timing of 662 NRM acquisition for the samples included in the tilt test (Fig. 12). The results imply a postfolding 663 but a pre-Alpine tilt NRM. The shallow inclinations suggest that Iberia was located at equatorial 664 latitudes and the southerly declinations suggest that this occurred during a reverse chron. We and early Permian times after Alpine tilt correction, with the exception of OG03 and 04 (Fig. 15 A).
the expected general good fit if remagnetization had happened after the Alpine orogeny, when Iberia 681 was tectonically stable (Fig. 15 B). In geographic coordinates only three sites fit in both paleolatitude 682 and declinations with the post-Alpine GAPWaP segment: OG3, OG4, and OG16. However, we 683 interpret OG16 as a Permian remagnetization since it fits much better with its neighboring sites 684 OG17, 18 and 19 after the correction of the alpine tilt. 685 OG03 and OG04 show a negative fold test (Fig. 10) and a common true mean direction in 686 geographic coordinates (Fig. 13). Their paleomagnetic direction is, however, significantly different 687 from OG01 that is a late Carboniferous-early Permian site located in the same thrust sheet. OG03 688 and OG04 show declinations to the south (both in geographic coordinates and after Alpine tilt 689 correction) and upward inclinations of -43° and -48° (geographic coordinates) or -60° and -68º (after 690 the restoration of the inferred Alpine tilt respectively). Both geographic and Alpine tilt corrected 691 data indicate a remagnetization when Iberia was located at latitudes between 25° and 50° during a 692 reverse chron (Fig. 15A, B). OG03 and 04 results fit best with a remagnetization that postdates the 693 Cretaceous normal superchron (e.g. Izquierdo-Llavall et al., 2015 and references therein) (Fig. 15). 694 Inclinations >60° (i.e. after Alpine tilt corrections) would only be possible when Iberia was located 695 at similar latitudes as at present, fairly long after the Alpine orogeny which ended at Miocene times. 696 In such case, no correction at all should be applied. Therefore, the magnetization has to be syn-to  (Fig. 9), but within-site clustering worsens after bedding correction (Table 2). After 716 Alpine tilt correction, however, a very good fit results with the majority of our Silurian-Devonian 717 collection. In addition, site BN1-OR15 contains pyrrhotite (Fig. 3), a feature common to all the 718 other samples studied (Fig. 9)  activity, we suspect that they could be remagnetized as well. We request using those paleolatitudes 724 (Tait et al., 1994;1999) with caution; their paleomagnetic veracity warrants to be reassessed. 725 Despite the inherent loss of information due to the remagnetization, especially regarding the interesting insights. Based on the inclination data we interpret that the sampled rocks mostly 728 range from a few degrees to ca. 90° (Fig. 1). We note that such results are also in line with the may be opposite to and/or in the same sense as the late Carboniferous clockwise rotations (Fig. 15). 759  • Given the generally good paleomagnetic quality of the Devonian carbonates, they could be 794 targeted to study the Alpine imprint on Paleozoic rocks and thus, to unravel the rotational 795 history of basement thrusts.

Conclusions and caution for paleomagnetists
• The widespread remagnetizations found in the Paleozoic of the Pyrenees indicate that 797 paleolatitudes inferred for Silurian and Devonian times from the studied rocks are very 798 unlikely original and should be taken very cautiously. We urge a reassessment of Siluro-799 Devonian poles from the Variscan in Europe. 800 • Paleomagnetism from multi-orogenic areas is NOT A SIMPLE GAME. However, the 801 systematic combination of paleomagnetism with detailed structural observations, seems to 802 be a foremost way to unravel complex tectonic evolutions.                        Figure 7: Viscous remanent magnetization (VRM) from all samples is compatible with the present-day field (geographic coordinates). Red dots are those that fall outside of the 45° cutoff. Uncertainty envelope is in both cases VGP A95. The rather large scattering is likely due to the small number of demagnetization levels containing the VRM (3-4) and the possible migration of the VRM during transport, storage and analysis. Orri thrust S dipping ramp (Fig. 2c) Bielsa thrust S dipping ramp (Fig. 2b) Bielsa thrust S dipping ramp (Fig. 2b) Gavarnie thrust N dipping ramp (Fig. 2b) Gavarnie thrust N dipping ramp (Fig. 2b) Gavarnie thrust flat to S dipping ramp (Fig. 2a) Gavarnie thrust flat to S dipping ramp (Fig. 2a) Gavarnie thrust flat to S dipping ramp (Fig. 2a) Gavarnie thrust flat to S dipping ramp (Fig. 2a) Gavarnie thrust flat to S dipping ramp (Fig. 2a Nogueras thrust (Fig. 2c) Nogueras thrust (Fig. 2c) Orri thrust flat ramp (Fig. 2c) Orri thrust N dipping ramp (Fig. 2c) Orri thrust (Fig. 2c) Orri thrust N dipping ramp (Fig. 2c) Orri thrust S dipping ramp (Fig. 2c) Orri thrus S dipping ramp (Fig. 2c) Orri thrust N dipping ramp (Fig. 2c) Figure 9: Directional and VGP results in geographic coordinates of sites OG12 to OG19 and BN1-OR15. Uncertainty envelope is in both cases VGP A95. Red dots are those that fall outside of the 45° cut-off. Sites OG02, 05 and 08 did not provide statistically meaningful results and were not interpreted.