<<

Tectonophysics 358 (2002) 121–134 www.elsevier.com/locate/tecto

Images of lithospheric heterogeneities in the Armorican segment of the Hercynian Range in

S. Judenherca,*, M. Graneta, J.-P. Brunb, G. Poupinetc, J. Plomerova´ d, A. Mocquete, U. Achauera

a Ecole et Observatoire des Sciences de la Terre, UMR 7516, Strasbourg, France b Ge´osciences Rennes UPR 4661, Rennes, France c Laboratoire de Ge´ophysique Interne et de Tectonique, UMR 5559, Grenoble, France d Geophysical Institute, Czech Academy of Sciences, Prague, Czech Republic e Laboratoire de Plane´tologie et de Ge´ophysique, FRE-CNRS 2129, , France Received 4 September 2000; received in revised form 4 April 2001; accepted 15 June 2002

Abstract

The Armorican is located in western France. It is a part of the Hercynian Range formed in several phases of SE–NW compression from to Carboniferous time (ca. 400–300 Ma). The main tectonic features are the North and South Armorican shear zones with WNW–ESE strike. In the framework of the Ge´oFrance3D program, we have installed 35 seismological stations along two NS lines crossing the main Hercynian WNW–ESE fault systems. The data set collected during the experiment allows us to compute a 3D P-velocity model and to map S-wave seismic anisotropy at depth using teleseismic shear waves splitting measurements. The South Armorican shear zone (SASZ) is characterized by a strong (4–5%) velocity contrast and a fast shear wave azimuth parallel to its strike. The North Armorican shear zone (NASZ) does not show any significant lithospheric signature. The tomographic image and anisotropy results suggest that the Armorican lithosphere consists of the juxtaposition of distinct lithospheric blocks assembled in a type process predating the Hercynian . D 2002 Elsevier Science B.V. All rights reserved.

Keywords: Seismic tomography; Seismic anisotropy; Continental collision; Hercynian lithosphere

1. Introduction such as the Himalayas or, in Europe, the or the . The geophysical imaging of these active Formation of mountain ranges is one of the most regions often shows that not only the crust but the spectacular effects of plate tectonic. Thus, it is not sur- whole lithosphere is involved in the mountain building. prising to see many Earth science studies, in particular, This paper focuses on the Armorican Massif (north- in the field of geophysics, focusing on active structures western France), which is a preserved segment of the Hercynian Orogeny (400–250 Ma). The main question arising when considering this old and presently stable * Corresponding author. Dipartimento di Scienza Fisiche, Com- plesso Universitario Monte S. Angelo, Via Cintia, Naples 80125, region concerns the present-day structure of the litho- Italy. sphere. Does the lithosphere beneath a 250-Ma-old E-mail address: [email protected] (S. Judenherc). still keep the signature of the related

0040-1951/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved. PII: S0040-1951(02)00420-1 122 S. Judenherc et al. / Tectonophysics 358 (2002) 121–134 geodynamic events? Do the tectonic features, such as two major strike–slip fault zones of Carboniferous age fault systems, involve the whole lithosphere? (c 350–300 Ma). The Hercynian Range is a continental (1) The North Armorican Domain (NAD) is dom- collision range. It extends in western Europe from inantly constituted of late Precambrian age formations southern Spain to the . In France, strongly deformed and metamorphosed during the the main remnants include the Armorican Massif, the Cadomian Orogeny (650–540 Ma) (Cogne´ and and the Mountains (Matte, Wright, 1980). (2) The Central Armorican Domain 1986). The Hercynian Range was built from the con- (CAD) is separated from the NAD by the North vergent motion of two continental blocks: Armorican Shear Zone (NASZ) and is bounded to to the South and to the North. The present-day the south by the South Armorican Shear Zone (SASZ). Armorican Massif was located on the southern border The CAD is mainly constituted of Paleozoic sediments of Laurasia and was bounded to the south by the lying unconformably on Brioverian sediments, rang- northward subduction of oceanic lithosphere (Matte, ing in age from Neoproterozoic to (c 600 1986; Burg et al., 1987). At the beginning of Devonian Ma) (Brun et al., 1991). (3) The South Armorican ( c 400 Ma), the closing of this oceanic basin lead to Domain (SAD) mainly consists of metamorphic and the continental collision. The latest events related to the plutonic rocks that have undergone an intermediate to collision are of Permian to Triassic age (250–200 Ma). high-pressure associated with crustal thickening during Devonian and Carboniferous (400– 1.1. Geological settings 300 Ma). The Paleozoic formations of the CAD are only af- The Armorican Massif (Fig. 1) is organized into fected by anchizonal to epizonal metamorphism indi- three E–W-trending elongated domains separated by cating very shallow burial (Le Corre, 1977). Together

Fig. 1. Map of the main Hercynian features of the Armorican Massif. The shaded area in the inset represents the approximate extension of the European Variscides (simplified after Matte, 1986). The locations of the temporary stations installed during the ARMOR2 experiment (1997) and the three short-period permanent stations (SGMF, GRR, and QUIF) used for the normalization process are shown. The legend at the lower left corner indicates the instrument type (1C: vertical sensor, 3C: three-component sensor), the eigenperiod of the instrument (BB stands for broad band) and the seismological recorder type (trig.: triggered station, cont.: continuous recording). S. Judenherc et al. / Tectonophysics 358 (2002) 121–134 123 with the Brioverian sediments, the Paleozoic forma- high beneath the Armorican Massif, ranging from 8.15 tions are affected by a single event of deformation to 8.23 km/s (Sapin, 1973; Hirn, 1977). Comparable during the Carboniferous characterized by long-wave- Pn velocities were obtained by Judenherc et al. (1999) length and low-amplitude folding across the whole using data from the permanent seismic network sta- CAD associated to a vertical slaty cleavage and a nearly tions. This analysis of the anisotropic tomography also horizontal stretching lineation. This deformation pat- shows that the Pn velocity is higher when the waves tern results from dextral simple shear related to strike– propagate parallel to the NASZ in the NAD and slip displacement along the SASZ (Gapais and Le northern CAD and parallel to the SASZ in the SAD. Corre, 1980). Dating of synkinematic emplacement At greater depth, a detailed study of P arrival times and and mylonitisation of (Vidal et al., shear wave polarization from active-source experi- 1984) along the SASZ gives an age of late Devon- ments led Hirn (1977) to propose the existence of an ian–early Carboniferous for the dextral simple shear of anisotropic layer between 60 and 100 km beneath the CAD. It is especially important to note that the basal central . In summary, one can state that the envelope of unmetamorphosed formations knowledge of the deep structure of this region is (c 450 Ma) remains nearly horizontal across the whole limited, in particular, the deep extension of the Hercy- CAD. The above arguments indicate that the sediments nian tectonic features is still an open question. Thus, of the CAD have not been affected by thrusting and the main goal of this study is to characterize the consequently suggest that the CAD could have escaped Hercynian lithospheric seismic structure at depth and significant crustal thickening during the Hercynian to image the deep geometry of the two main Hercynian continental collision. This is in contrast to the presence shear zones. Both would put new constraints on the of eclogitic rocks immediately south of the CAD long-term evolution of this mountain range. between the two branches of the SASZ (Fig. 1). Since the end of the Hercynian Orogeny, the massif has recorded a series of discrete second-order tectonic 2. The seismological study events related to the opening of the North Atlantic during the early and to continental exten- Tomographic images contain information about the sion during the Oligocene. Recent tectonics is moder- mineralogy and thermal state of the study area (e.g. ate; it is characterized by diffuse seismicity with Sobolev et al., 1996). These images reveal the deep earthquake magnitudes rarely exceeding 4.0, and 0.5 geometry of the geological units observed at the sur- to 1.1 mm/year of uplift of the northwestern border of face with a wavelength (c 10 km) compatible with the the Armorican Massif as demonstrated by river inci- lithospheric structural heterogeneities. Moreover, seis- sion during the Quaternary (Bonnet et al., 1998) and mic anisotropy studies provide information on the leveling measurements (Lenoˆtre et al., 1999). Other- deformation pattern and large-scale fabric at depth wise, no major tectono-thermal events have occurred (Babusˇka and Cara, 1991; Vinnik et al., 1992; Silver, in the past 250 My. 1996). In addition, it provides better constraints on the tectonic history of the study area. 1.2. Geophysical settings In order to map the lateral seismic velocity and seismic anisotropy variations beneath the Armorican Geophysical knowledge of the upper mantle of the Massif, a temporary seismic network was operated Armorican Massif is rather poor. Several active-source from March to December 1997 (Fig. 1). The geometry seismic sounding experiments have been run in north- of the network was designed to record teleseismic P- ern and central Brittany (e.g. Bitri et al., 1996) that all and S-waves across the NASZ and SASZ. The array show a quite flat Moho (average depth of 31–32 km) consisted of two 150-km-long N–S-oriented, nearly beneath the massif and, in particular, across the North parallel profiles, with a station spacing of about 10 km Armorican shear zone. Some long-range seismic pro- and a distance of 40 km between the profiles (Fig. 1). files were performed in the early 1970s running from The experiment provided a set of 850 reliable P or the northwest to the southeast of France. The Pn PKP arrival times associated to 130 teleseismic events velocity deduced from this refraction profiles is rather with an epicentral distance >30j. In addition, a set of 124 S. Judenherc et al. / Tectonophysics 358 (2002) 121–134

15 SK[K]S good-quality waveforms on three-compo- computed for each source region. This is achieved by nent instruments was retrieved, suitable for the study using the normalization process described in Appendix of lithospheric anisotropy. A. The procedure uses a subnet of reference formed by the permanent stations SGMF, GRR, and QUIF (Fig. 2.1. P-wave velocity tomography 1) belonging to the seismic network of the Commis- sariat a` l’Energie Atomique—Laboratoire de De´tec- 2.1.1. Teleseismic P-wave travel time residuals tion Ge´ophysique. P and PKP arrival times at the temporary and per- The main advantage of this particular normalization manent reference stations were determined by wave- process is that it allows to include in the data set all the form correlation. This technique leads to an accuracy events recorded by more than one station of the better than 0.1 s for the phase picks. We then computed temporary network and one station of the reference P-wave travel times for the IASP91 model (Kennett subnetwork. In fact, we used only events recorded by and Engdahl, 1991) using the source parameters pro- at least three stations of the array and one reference vided by the NEIC (USGS, 1997). The travel time station; this selection leads to a final data set of 603 residuals (the difference between observed and com- relative P and PKP travel time residuals associated to puted travel times) are related to velocity perturbations 96 events with an epicentral distance >30j recorded by in the Earth and also to errors in source location and the temporary array. Fig. 2 shows the epicentral dis- origin time. In order to reduce the systematic bias due tance and azimuthal distribution of these retained to source parameter error, we have computed relative events together with calculated average relative resid- residuals by subtracting a reference residual function ual curves (RRC) along the profiles. The event dis-

Fig. 2. Map, centered on Brittany, showing the geographical distribution of the 96 events used for the inversion procedure. On each panel is represented an example of significant RRC computed and averaged for different seismic source regions. The labels correspond to the site numbers (blue for the western profile, red for the eastern profile). The RRC are displayed as a function of the latitude of the recording site (see right bottom panel). The error bars represent the standard error. S. Judenherc et al. / Tectonophysics 358 (2002) 121–134 125 tribution is characteristic for the worldwide average image. In our case, the crust is made of two layers. seismicity as recorded in Western Europe: gaps are The upper one (depth 0–10 km) consists of cones of observed for south and northwest azimuths while an rays, each one associated with a single station. The important number of events has north–northeastern lower layer (10–30 km) consists of 15 15 km back-azimuths (North and South Pacific events). The blocks. In the mantle, we used four 30-km-thick layers amplitude of variations of the RRC is 0.6–0.8 s across and blocks with dimensions of 20 20 km (see Table a distance of less than 150 km. It is of the same order as 1). Ray tracing is performed by propagating the rays what has been observed in other Paleozoic regions from the stations down to the bottom of the target such as the western rim of the Bohemian Massif volume in the IASP91 1-D model. (Plomerova´ et al., 1998), in SW Ireland (Masson et al., 1998) and in the (Poupinet et al., 2.2. Anisotropy measurements 1997). For all the northern seismic source regions (groups 3, 4, 5 and 6), the RRC show a maximum The measurement of seismic anisotropy is a basic between latitudes 48jN and 48.3jN and a minimum tool for the detection of deformation at depth. Aniso- between latitudes 47.5jN and 47.7jN. The shape of tropy has been interpreted and modeled as the result of the RRC for events from the west (groups 1 and 2) is strain-induced lattice preferred orientation (LPO) of almost linear while for the events from east (group 7), mantle material, mainly olivine (Nicolas and Christen- the RRC pattern shows a smooth maximum between sen, 1987; Ribe and Yu, 1991; Silver, 1996). In this latitudes 47.7jN and 48.3jN. paper, we address the seismic anisotropy in the Armor- ican Massif by combining two different methods 2.1.2. Tomographic inversion associated with Pn and SKS waves. Azimuthal aniso- Generally, P-wave relative residuals integrate the tropy in the mantle lid can be mapped using Pn travel effect of velocity perturbations located along the whole time tomography (e.g. Bamford, 1977; Enderle et al., ray path. Considering a set of P-wave originating from 1996; Hearn, 1996). These results provide the magni- the same distant event recorded by a small network, we tude and direction of P-wave anisotropy based on Pn- can suppose that the ray paths are separated only in a waves propagating subhorizontally in the mantle lid. relatively shallow region beneath the network, for In addition to azimuthal anisotropy evident in Pn- example, the crust and the uppermost mantle. This is waves, anisotropy in the mantle is evident in teleseis- one of the basic assumptions of the tomographic mic shear-wave splitting (Fukao, 1984; Vinnik et al., method used in this study (Aki et al., 1977). The 1984; Silver and Chan, 1988). In this work, we used mathematical formulation, capabilities and limitations SKS and SKKS core phases. These seismic phases result of this so-called ACH method have been reviewed by from a P-toS-conversion at the core–mantle boundary, Evans and Achauer (1993). The 3-D structure is so that they are radially polarized when entering back divided into a limited number of layers, which are themselves partitioned into nonoverlapping blocks. Table 1 Within each block, the velocity perturbation is calcu- Block geometry used for the tomographic inversion lated in order to explain the relative residuals. As Layer Z Depth NE firstly mentioned by Aki et al. (1977), the initial model does not influence significantly the inversion results. 1 10 10 Special layer 220301515 The vertical extent of the target region is limited to the 330602020 aperture of the array (Evans and Achauer, 1993).We 430902020 therefore have computed a 3-D velocity perturbation 5 30 120 20 20 model down to 150 km. This depth limit is induced by 6 30 150 20 20 the ray geometry which is not able to provide reliable Z, Depth, N and E refer to the thickness, bottom depth, N–S perturbations at greater depth. Some tests have shown extension, and E–W extension of the blocks (km). The first layer (0–10 km) is made by a set of cones, each one associated to a single that extending the model downward by adding a station. Hence, one cone contains all the rays recorded at the station. deeper layer introduces strong coupling between these The choice of the block size is conditioned by the stations’ spacing two deepest layers without changing the shallower and the ray density. 126 S. Judenherc et al. / Tectonophysics 358 (2002) 121–134 into the mantle. Due to their travel in the outer core, Babusˇka et al. (1984) have proposed another SKS and SKKS phases only detect anisotropy along the approach for determining 3-D anisotropy in the litho- upgoing part of the ray path. In an anisotropic region, sphere using P-wave travel time residuals. It is based such shear waves split into two quasi-S waves, which on the azimuthal normalization of the data that is propagate with different velocities. Hence, observation intended to extract the directional term of relative of shear wave splitting from SKS or SKKS phases is an residuals in order to map so-called P-residual spheres. evidence of anisotropy. The method used to retrieve the We have applied the method on our data set and splitting parameters is based on Fukao (1984).It observed that the ray geometry does not allow us to consists in finding a coordinate system in which the discriminate between respective contributions of lat- fast and slow waves are well separated. The polar- eral heterogeneities and anisotropy in the P-residual ization of the fast shear wave is described in 3-D by its spheres (Plomerova´ et al., 2000; Judenherc, 2000). incidence and azimuth (/) and the parameter yt repre- sents the time delay between the fast and slow quasi-S 2.3. Results waves. yt is a direct indicator of the product of the thickness of the anisotropic layer and of the amount of The P-wave velocity perturbations are presented in anisotropy. The technique we used does not provide Fig. 3. Layers 1 and 2 (i.e. the crust) are not displayed as any error estimation for single measurements. How- they mainly show site effects or local and short wave- ever, previous studies (e.g. Granet et al., 1998) show length perturbations. The maximum velocity contrast is that 10j and 0.2 s are reasonable error bars for the observed in the deepest layer (from about 3% to polarization angles and yt, respectively. + 3% in layer 6) suggesting a smearing of some deeper

Fig. 3. Smoothed P-velocity perturbations in %. The blue and red colors indicate positive and negative perturbations, respectively. In order to avoid the effects of discretisation, the tomograms are computed using an offset and average technique (Evans and Achauer, 1993). We also have applied a low pass filter with a cutoff wavelength equal to 1.5 times the block size. The top panel shows the plane views. The contour lines are drawn every 1.5%. Superimposed are the main fault zones detailed in Fig. 1. S. Judenherc et al. / Tectonophysics 358 (2002) 121–134 127 anomalies in this layer and a contribution of neglected Table 2 anisotropy. However, as previously explained, due to List of the distant events used in this paper and their splitting parameters the aperture of the network, extending the model y downward is not reasonable. In layer 3 (30–60 km), Obs Region Date Site h D / t perturbations are rather small. While some southeast- 1 Kermadec Islands 1997.05.25 05 356 164 133 0.5 ern and northwestern areas of the model are charac- 2 Southern Bolivia 1997.01.23 20 237 90 160 0.6 3 Kermadec Islands 1997.05.25 20 356 164 152 0.8 terized by negative perturbations (down to 2%), 4 Banda Sea 1997.09.26 20 61 120 Linear positive perturbations (+ 2%) are observed in the west- 5 Vanuatu Islands 1997.05.21 27 17 152 161 0.7 ern part. The deeper layers (4 to 6) show a simple 6 Kermadec Islands 1997.05.25 27 355 164 152 0.5 pattern with a NW–SE general structural trend. From 7 Banda Sea 1997.07.31 27 60 122 Linear south to north, three domains with alternating pertur- 8 Vanuatu Islands 1997.05.21 09 18 152 128 0.5 9 Kermadec Islands 1997.05.25 09 356 164 118 0.7 bations are observed. They are separated by two 10 Vanuatu Islands 1997.05.21 15 18 152 132 0.7 boundaries clearly visible in the cross-section (Fig. 11 Kermadec Islands 1997.05.25 15 355 165 110 1.2 3). The southern boundary seems associated with the 12 Central Chile 1998.07.29 16 234 101 Linear SASZ and separates the southern domain (negative 13 Philippines 1998.09.02 16 57 110 Linear 14 Southern Peru 1998.10.08 16 244 89 Linear 15 Kyushu, Japan 1999.01.24 16 39 91 Linear h is the back-azimuth, i.e. the azimuth of the incoming wave (mea- sured east positive from the north at the station), D is the epicentral distance, / and yt are the splitting parameters, i.e., the fast S-wave azimuth and the time delay between the fast and slow S-waves.

perturbations of about 2% to 3%) and a central relatively fast region (+ 2% to + 3%). The second boundary is oriented NW–SE, crossing the NASZ with a >30j angle. To the south of this limit is the previously described central positive region; to the north, a large area shows negative perturbations increasing with depth (from 1.5% in layer 4 to 3% in layer 6). The Armorican Pn anisotropic model considered here is based on a more regional model beneath France (Judenherc et al., 1999). This Armorican Pn model is characterized by a 3–4% anisotropy amplitude, with the fast directions oriented E–W, parallel to the NASZ in the NAD and NW–SE, The results of SKS and SKKS splitting measurements are plotted in Fig. 4 and tabu- lated in Table 2. Only a few stations provided good- quality records due to the high level of microseismic noise in this area. In the SAD, fast directions are oriented roughly parallel to the SASZ (N110jE and Fig. 4. Anisotropy results obtained in Brittany from the body wave N130jE for two distinct events at site 15). Measure- studies. Pn anisotropy (3–4%) is symbolized by green bars for which length is proportional to the magnitude of the anisotropy, and the ments in the northern CAD provided N150jE– azimuth of the bar indicates the fast Pn azimuth. Splitting measure- N160jE fast directions. In general, the delay times ments, / and yt, are represented by the azimuth and the length of the are small, between 0.5 and 0.8 s, except for site 15 red bars, respectively. Black arrows indicate null directions: back- where one measurement has provided a delay time of azimuths for which no SKS splitting is observed. One should note the 1.2 s. Such small delay times reflect a thin anisotropic similarity of Pn and SKS fast directions along the South Armorican Shear Zone to the south while to the north, only Pn fast direction layer and/or a small amount of anisotropy (Mainprice aligns with the North Armorican Shear Zone. and Silver, 1993; Ben Ismaı¨l and Mainprice, 1998). For 128 S. Judenherc et al. / Tectonophysics 358 (2002) 121–134 some earthquakes, we observed linear particle motion, The effects of lower mantle heterogeneities have i.e. a clear SKS or SKKS phase with no energy on the been addressed by Masson and Trampert (1997) and transverse component. These null directions (re- Vergne (1998). They estimated the contributions of presented as black single arrows in Fig. 4) can be lower mantle heterogeneities in the observed travel explained either by the absence of anisotropy or by times along a 600-km-long profile in Northern Tibet incoming waves propagating in a direction close to the using a long wavelength shear-wave velocity model. In fast or slow direction of the anisotropic medium (Bar- this particular case, at first order, the lower mantle ruol and Souriau, 1995). We observe null directions at residual curves show a linear shape with a total ampli- sites 16, 20 and 27. In particular, two events from tude < 0.4 s along the profile for epicentral distances N237jE and N61jE opposite azimuths (referenced as >30j. Thus, we can expect a maximum lower mantle observations 2 and 4 in Table 2) were recorded at site contribution of less than 0.1 s for the small Armorican 20. While the NE event provides a null direction, a clear network, i.e. this effect is of second order. However, the anisotropic signature is detected for the SW event. wavelength of the heterogeneities revealed by tomog- raphies based on P-wave travel time (Bijwaard et al., 1996; Bijwaard and Spakman, 2000; Piromallo and 3. Sensitivity of methods and possible bias Morelli, 1997) is generally greater than 100–200 km in the lower mantle. As this size is not significantly The tomographic inversion provides a residual smaller than the network aperture, we cannot expect variance reduction of 70%, which is an average value that those deep and wide heterogeneities would pro- for tomographic problems of this kind and thus is duce any sharp and strong residual variation across the satisfactory. In this work, we estimate the uniqueness network. of the model from the resolution matrix R (Aki and We have defended that the small aperture of the Richards, 1980). The coefficients of R depend mainly temporary network limits the contribution of the lower on the ray geometry, the block model, and the ray mantle heterogeneities, however, the contribution of tracing. A given coefficient Rij represents the coupling the deep upper mantle is more significant. As a between the blocks i and j. In the ideal case, R is the consequence of the limited model thickness, the deep identity matrix and the solution of the inverse problem upper mantle velocity perturbations located out of the is unique. As the full resolution matrix may be difficult target volume are mapped in the last layer. If the to read because of the lack of geographic reference, structures identified in the lower part of the model only the diagonal term of R is usually represented. Fig. extend deeper than the bottom of the model, then the 5 (top) shows the diagonal terms of the resolution amplitudes computed in the lowest layer are over- matrix. A value of 0.5 is usually considered as a estimated. threshold value above which velocity perturbations Finally, as anisotropy is detected in the region, a are correctly retrieved. According to this criterion, trade-off between anisotropy and velocity perturba- the acceptable regions are those colored in pink and tions is possible. This problem has been assessed by red in Fig. 5 (top). This concerns the inner part of the Sobolev et al. (1999). They have shown that the study area, in particular, beneath the stations belonging amplitude of the ‘‘false’’ velocity anomalies mapped to the western profile as they provided most of the due to ignored anisotropy is related to the shape and data. A comparison of the diagonal term of the variation of anisotropy and to the data coverage in the resolution matrix and the data coverage (Fig. 5, bot- study area. In the case of the Armorican Massif, we tom) shows that both parameters are highly correlated. have estimated this bias by testing some synthetic The best resolved regions are those of highest hit-count models (Judenherc, 2000) and we have observed that and highest crossing ray density; this explains the realistic models lead to false velocity perturbations of westward shift of the high-resolution area in the deep the order of 1% to 1.5%. layers of Fig. 5 (top). The standard errors of the Le´veˆque and Masson (1999) pointed out some velocity perturbations is < 1.0% in all the layers, possible misinterpretations due to the artifacts intro- hence, only velocity contrasts higher than this value duced by the usual normalization processes of the should be discussed. residuals. Their work emphasizes that using relative S. Judenherc et al. / Tectonophysics 358 (2002) 121–134 129

Fig. 5. Diagonal terms of the resolution matrix associated to the tomographic inversion (top) and data density in layers 2 to 6 (bottom). Note the strong correlation between the two images. The data density is represented by the hit-count: number of ray hitting each blocks. 130 S. Judenherc et al. / Tectonophysics 358 (2002) 121–134 residuals, one computes velocity perturbations with as a realistic upper limit, we can explain about 1% respect to an unknown reference velocity within each more in the velocity perturbation. About 2% lateral layer. Consequently, a negative perturbation cannot be velocity change still remains, particularly in the central directly interpreted as a low-velocity region since the part of layer 90–120, which is the best resolved part of ‘‘normal’’ velocity remains unknown within a layer the model as demonstrated by the resolution map and (the same is true for positive perturbations). Moreover, the data density (Fig. 5). Consequently, we would the connections between perturbations in different la- likely associate 1% to 2% in the velocity variation to yers are not obvious. Le´veˆque and Masson (1999) a compositional origin. propose to use a ‘‘true’’ one-dimensional model in Following these estimates, the tomographic which absolute velocities are recomputed to avoid results are interpreted as images of lithospheric these bias, but the choice of this model depends on blocks of different composition and probably differ- the a priori knowledge of the regional structure which ent tectonic history. However, the question of the is not sufficient in our case. origin of the central high velocity body arises. We propose to interpret it as the relic of lithospheric material brought down into the asthenosphere by a 4. Discussion and interpretation northward subduction process known to have occurred during the early stage of the collision The major question raised by the presented results (c 400 Ma, e.g. Burg et al., 1987; Dias and Ribeiro, concerns the origin of the velocity perturbations. We 1995). This is illustrated in Fig. 6. The anisotropy have discussed the possible bias that can explain the pattern observed on the SASZ (site 15, Fig. 4) could high amplitudes. In particular, the deepest layer (120– then be related to the Carboniferous (350–300 Ma) 150 km) may be affected by heterogeneities located intense dextral transpression regime suggested by out of the modeled volume and by neglecting aniso- the same authors. In the South Armorican Domain, tropy. We have to explain 4% to 5% of velocity both Pn and teleseismic shear wave fast azimuths perturbation in the rest of the model where the main align parallel to the Hercynian shear direction. This sources of these perturbations are thermal anomaly, supports a vertically coherent deformation, as pro- anisotropy, and compositional change. If the bias posed by Silver (1996), for the southern part of the introduced by anisotropy variations across the region Armorican Massif. In the northern part of CAD and are responsible of 1–1.5% of the observed velocity in the NAD, only the Pn anisotropy direction change (Judenherc, 2000), 2.5–4% still must be correlates with the Hercynian structures while tele- explained by thermal and chemical changes. The heat seismic shear waves splitting parameters and the flow measurements (Vasseur, 1982) reveal an average velocity perturbations do not exhibit any correlation value of 70 mW/m2 for the region of interest and do with the Hercynian structures. In this region, tele- not show any correlation with the velocity image at seismic shear wave splitting measurements can be depth. This is not very surprising as the latest signifi- explained by a high-velocity plane oriented about cant tectono-thermal events are older than 250 Ma. N150jE dipping to the SW. This orientation is Nevertheless, according to Davies (1999), the cooling deduced from a method proposed by Sˇ´ıleny´ and time of a 100-km-thick lithospheric thermal anomaly is Plomerova´ (1996) to retrieve the 3-D anisotropy of the order of 300 Ma, suggesting that thermal orientation. In our case, due to the limited number anomaly remains. Such remnant thermal anomaly of data, we can only propose a model that explains contributes to the velocity perturbation but, if the the observations but we are not able to discuss the lateral velocity changes were entirely of thermal ori- uniqueness of the model. In particular, a NE-dipping gin, a temperature anomaly of the order of 300 to 400 low-velocity symmetry axis is compatible with the jC would be required to explain the 2.5–4% velocity observations at site 20 of both the null direction change (Sobolev et al., 1996; Goes et al., 2000). Such a from the back-azimuth N61jE and the SKS splitting thermal anomaly in the upper mantle is not realistic in from the back-azimuth N237jE. We hypothesize a tectonically stable region as the Armorican Massif. that the associated lithospheric fabric may character- By considering a 100 jC lateral change of temperature ize the Cadomian lithosphere. Moreover, a single S. Judenherc et al. / Tectonophysics 358 (2002) 121–134 131

Fig. 6. SSW–NNE interpretative cross-section along profile A–B in Fig. 3 (same color scale). Inclined lines depict the fabric as deduced from seismic anisotropy studies (Fig. 4). anisotropic layer with dipping symmetry axis does mian lithosphere, has escaped the Hercynian defor- not explain the Pn fast direction that is parallel to mation. The seismological images show the remnant the NASZ. An additional thin subcrustal layer with of a precollisional subduction of at least a Devonian an E–W fast azimuth is needed to fit the Pn data age (350 Ma) followed by a syncollisional trans- set. We thus propose the presence of a subcrustal pression regime during Carboniferous (350–300 detachment horizon that progressively decouples the Ma). In a more general framework, the results crust and the mantle towards the north, as it was presented in this paper illustrate the complexity of proposed by Dia`z et al. (1996) beneath the Hercy- the collage of several continental and oceanic pieces nian SW Iberia. that were accreted during successive stages that led to the Hercynian Range.

5. Concluding remarks Acknowledgements We have obtained a seismic image of the litho- spheric structure of an eroded collision zone in The authors thank W.D. Mooney and P. Ledru for Brittany that represents a preserved part of the their suggestions and improvements on a former Hercynian Range. We focused on characterizing the version of the manuscript and S. Goes, and J. South and North Armorican Shear Zones in terms of Trampert for their constructive and interesting com- seismic velocity variations and seismic anisotropy ments. Field investigations have benefited from the using data from a temporary array. Our seismological efficient help of the Laboratoire de Plane´tologie et data show two distinct stable lithospheric domains in Ge´odynamique (FRE-CNRS) of Nantes University the upper mantle beneath the Armorican Massif. with special thanks to Eric Boeuf and Fre´de´ric While the southern domain is strongly affected by Deschamps. All figures were created using Generic the Hercynian Orogeny, the northern one, a Cado- Mapping Tool (Wessel and Smith, 1995). 132 S. Judenherc et al. / Tectonophysics 358 (2002) 121–134

Appendix A. Normalization process recorded by at least one of the reference stations. A virtual reference residual is built with the regionalized The normalization process used for the tomo- reference residual function used as a replacement for graphic inversion allows to use data from events the reference stations that have not recorded the event. recorded by only a few stations. In the following Thus, the normalized residual for a station j recording equations, tij represents the travel time from source i an event i is given by: to station j, and superscripts ‘a’ and ‘r’ stand for X absolute and relative, respectively. According to these n a a ! rij ¼ rij dik ½rik Rk ðnik Þ ð6Þ notations, the absolute residual for ray ij is written as k the difference between observed (obs) and calculated where d equals 1 if reference station k has recorded (cal) travel time: ik the event and equals 0 if not.

a obs cal rij ¼ tij tij : ð1Þ References Then, choosing a total of N reference stations (k), we define a set of relative residuals for the events (i) Aki, K., Richards, P.G., 1980. Quantitative Seismology. Freeman, recorded by all of the N reference stations: San Francisco. Aki, K., Christofferson, A., Husebye, E.S., 1977. Determination of the three-dimensional seismic structure of the lithosphere. J. Geo- 1 XN rr ¼ ra ra : ð2Þ phys. Res. 82, 277–296. ik ik N ik Babusˇka, V., Cara, M., 1991. Seismic Anisotropy in the Earth. k¼1 Kluwer Academic Publishing, Norwell, MA. ˇ These relative residuals are used to build a regional- Babusˇka, V., Plomerova´, J., S´ıleny´, J., 1984. Large-scale oriented structures in the subcrustal lithosphere of Central Europe. Ann. ized reference residual function depending on the ! Geophys. 2, 649–662. propagation vector n of the upcoming waves for each Bamford, D., 1977. Pn velocity anisotropy in a continental upper reference station k. The function is built as a weighted mantle. Geophys. J. R. Astron. Soc. 49, 29–48. sum of the relative residuals for the events recorded by Barruol, G., Souriau, A., 1995. Anisotropy beneath the Pyrenees all the reference stations: range from teleseismic shear wave splitting: results from a test experiment. Geophys. Res. Lett. 22, 493–496. Ben Ismaı¨l, W., Mainprice, D., 1998. An olivine fabric database: an 1 XM R ð!nÞ¼ w rr ; ð3Þ overview of upper mantle fabrics and seismic anisotropy. Tec- k ik ik tonophysics 296, 145–157. W i 1 ¼ Bijwaard, H., Spakman, W., 2000. Non-linear global P-wave tomo- where M is the number of event recorded by station k graphy by iterated linearized inversion. Geophys. J. Int. 141, 71–82. and wik is a Gaussian weighting function defined as Bijwaard, H., Spakman, W., Engdahl, E.R., 1996. High resolution follows: global delay time tomography. Eos Trans. AGU, 77, Fall Meet. ! Suppl., F482. !ˆ! 2 Bitri, A., Brun, J.P., Guennoc, P., Marthelot, J.M., Nicollin, F., ðn; nikÞ r r Perrin, J., Truffert, C., 1996. Imagerie sismique de la structure wik ¼ rik exp 2 rik ; ð4Þ 2r profonde du domaine cadomien de Bretagne Nord. 16e` Re´union des Sciences de la Terre, p. 19. W being the sum of the weights for the direction !n: Bonnet, S., Guillaucheau, F., Brun, J.P., 1998. Relative uplift meas- ured using river incisions: the case of the armorican basement XM (France). C. R. Acad. Sci. 327, 245–251. Brun, J.-P., Ballard, J.-F., Le Corre, C., 1991. Identification of W ¼ wik ð5Þ Ordovician block-tilting in the hercynian fold belt of Central i¼1 Brittany (France): field evidence and computer models. J. Struct. and r being the smoothness of the regionalized refer- Geol. 13, 419–429. j Burg, J.P., Bale´, P., Brun, J.P., Girardeau, J., 1987. Stretching lin- ence residual function fixed to 15 . eation and transport direction in the ibero-armorican arc during Finally, the normalized residuals for any event can the siluro-devonian collision. Geodin. Acta 1, 71–87. be computed with a common reference if the events are Cogne´, J., Wright, A.E., 1980. L’oroge`ne Cadomien. Geology of S. Judenherc et al. / Tectonophysics 358 (2002) 121–134 133

Europe from Precambrian to the Post-Hercynian Sedimentary teleseismic delay time tomography. Phys. Earth Planet. Inter. Basins. 26th International Geological Congress BRGM Mem., 102, 21–32. vol. 108. Masson, F., Hauser, F., Jacob, A.W.B., 1998. The lithospheric trace Davies, G.F., 1999. Dynamic Earth: Plates, Plumes and Mantle Con- of the Iapetus Suture in SW Ireland from teleseismic data. Tec- vection. Cambridge University Press, New York. tonophysics 305, 83–98. Dias, R., Ribeiro, A., 1995. The Ibero-Armorican Arc; a collision Matte, P., 1986. Tectonics and model for the variscan effect against an irregular continent? Tectonophysics 246, belt of Europe. Tectonophysics 126, 329–374. 113–128. Nicolas, A., Christensen, N.I., 1987. Formation of anisotropy in Dia`z, J., Hirn, A., Gallart, J., Abalos, B., 1996. Upper-mantle upper mantle peridotites: a review. In: Fuchs, K., Froidevaux, anisotropy in SW Iberia from long-range seismic profiles and C. (Eds.), Composition, Structure and Dynamics of the Litho- teleseismic shear-wave data. Phys. Earth Planet. Inter. 95, sphere–Asthenosphere System. Geodyn. Ser. AGU, Washing- 153–166. ton, DC, pp. 111–123. Enderle, U., Mechie, J., Sobolev, S., Fuchs, K., 1996. Seismic ani- Piromallo, C., Morelli, A., 1997. Imaging the Mediterranean upper sotropy within the uppermost mantle of southern Germany. Geo- mantle by P-wave travel time tomography. Ann. Geofis. 40, phys. J. Int. 125, 747–767. 963–979. Evans, J.R., Achauer, U., 1993. Teleseismic velocity tomography Plomerova´, J., Babusˇka, V., Sˇ´ıleny´, J., Hora´lek, J., 1998. Seismic using the ACH method: theory and application to continental- anisotropy and velocity variations in the mantle beneath Saxo- scale studies. In: Iyer, H.M., Hirara, K. (Eds.), Seismic Tomog- thuringicul –Moldanubicum contact in Central Europe. Pure raphy: Theory and Practice. Chapman & Hall, New York, Appl. Geophys. 151, 365–394. pp. 319–357. Plomerova´, J., Granet, M., Judenherc, S., Achauer, U., Babusˇka, V., Fukao, Y., 1984. Evidence from core-reflected shear waves for Jedlicˇka, P., Kouba, D., Vecsey, L., 2000. Temporary array data anisotropy in the Earth’s mantle. Nature 309, 695–698. for studying seismic anisotropy of Variscan massifs—the Ar- Gapais, D., Le Corre, C., 1980. Is the Hercynian belt of Brittany a morican Massif, French Massif Central and Bohemian Massif. major shear zone? Nature 288 (5791), 574–576. Stud. Geophys. Geod. 44, 195–209. Goes, S., Govers, R., Vacher, P., 2000. Shallow mantle temperatures Poupinet, G., Thouvenot, F., Zolotov, E.E., Matte, P., Egorkin, A.V., under Europe from P and S wave tomography. J. Geophys. Res. Rackitov, V.A., 1997. Teleseismic tomography across the mid- 105, 11153–11169. dle Urals: lithospheric trace of an ancient continental collision. Granet, M., Glahn, A., Achauer, U., 1998. Anisotropic measure- Tectonophysics 276, 19–33. ments in the Rhine Graben area and the French Massif Central: Ribe, N.M., Yu, Y., 1991. A theory for plastic deformation and geodynamic implications. Pure Appl. Geophys. 151, 333–364. textural evolution of olivine polycrystals. J. Geophys. Res. 96,

Hearn, T.M., 1996. Anisotropic Pn tomography in the western 8325–8335. United States. J. Geophys. Res. 101, 8403–8414. Sapin, M., 1973. Structure de la crouˆte de la Bretagne au Massif Hirn, A., 1977. Anisotropy in the continental upper mantle: possible Central. PhD thesis, Universite´ Paris VI, Paris. evidence from explosion in seismology. Geophys. J. R. Astron. Sˇ´ıleny´, J., Plomerova´, J., 1996. Inversion of shear-wave splitting Soc. 49, 49–58. parameters to retrieve three-dimensional orientation of anisotropy Judenherc, S., 2000. Etude et caracte´risation des structures hercy- in continental lithosphere. Phys. Earth Planet. Inter. 95, 277–292. niennes a` partir de donne´es sismologiques: le cas du Massif Silver, P.G., 1996. Seismic anisotropy beneath the continents: prob- Armoricain. PhD thesis, Universite´ Louis Pasteur, Strasbourg ing the depth of geology. Annu. Rev. Earth Planet. Sci. 24, I, Strasbourg. 385–432. Judenherc, S., Granet, M., Boumbar, N., 1999. Two-dimensional Silver, P.G., Chan, W.W., 1988. Implications for continental struc- anisotropic tomography of lithosphere beneath France using ture and evolution from seismic anisotropy. Nature 335, 34–39. regional arrival times. J. Geophys. Res. 104, 13201–13215. Sobolev, S.V., Zeyen, H., Stoll, G., Werling, F., Altherr, R., Fuchs, Kennett, B.L.N., Engdahl, E.R., 1991. Traveltimes for global earth- K., 1996. Upper mantle temperatures from teleseismic tomog- quake location and phase identification. Geophys. J. Int. 105, raphy of the French Massif Central including effects of compo- 429–465. sition, mineral reactions, anharmonicity, anelasticity and partial Le Corre, C., 1977. Le Briove´rien de Bretagne Centrale: e´ssai de melt. Earth Planet. Sci. Lett. 139, 147–163. synthe`se lithologique et structurale. Bull. Bur. Rech. Geol. Min. Sobolev, S., Gre´sillaud, A., Cara, M., 1999. How robust is isotropic 1–3, 254–299. delay time tomography for anisotropic mantle? Geophys. Res. Lenoˆtre, N., Thierry, P., Blanchin, R., Brochard, G., 1999. Current Lett. 26, 509–512. vertical movement demonstrated by comparative levelling in USGS, 1997. Preliminary determination of epicenters. Tech. rep., Brittany (northwestern France). Tectonophysics 301, 333–344. U.S. Geological Survey. Reston, VA, US. Le´veˆque, J.J., Masson, F., 1999. From ACH tomographic models to Vasseur, G., 1982. Synthese des re´sultats de flux ge´othermique en absolute velocity models. Geophys. J. Int. 137, 621–629. France. Ann. Geophys. 38, 189–201. Mainprice, D., Silver, P.G., 1993. Interpretation of SKS-waves using Vergne, J., 1998. Un outil de tomographie te´le´sismique robuste tre`s samples from the subcontinental lithosphere. Phys. Earth Planet. utilise´ et critique´:lame´thode ACH. Analyse critique et ame´li- Inter. 78, 257–280. orations. Me´moire de DEA, EOST, Universite´ Louis Pasteur, Masson, F., Trampert, J., 1997. ACH or how reliable is regional Strasbourg I. 134 S. Judenherc et al. / Tectonophysics 358 (2002) 121–134

Vidal, P., Bernard-Griffiths, J., Peucat, J.J., Cocherie, A., Lefort, Vinnik, L.P., Makeyeva, L.I., Milev, A., Usenko, A.Y., 1992. Global F.P., Sheppard, S.M.F., 1984. Geochemical comparison between patterns of azimuthal anisotropy and deformations in the conti- himalayan and hercynian leucogranites. Phys. Earth Planet. In- nental mantle. Geophys. J. Int. 111, 433–447. ter. 35, 179–190. Wessel, P., Smith, W.H.F., 1995. New version of the generic map- Vinnik, L.P., Kosarev, G.L., Makeyeva, L.I., 1984. Anisotropy of ping tools released. EOS, Trans. Am. Geophys. Union 76, 329. the lithosphere according to the observations of SKS and SKKS waves. Dokl. Akad. Nauk SSSR 278, 1335–1339.