lithospheric and sublithospheric anisotropy beneath

Physics of the Earth and Planetary Interiors 158 (2006) 190–209

Lithospheric and sublithospheric anisotropy beneathcentral-southeastern Brazil constrained by

long period magnetotelluric data

Antonio L. Padilha ∗, Icaro Vitorello, Marcelo B. Padua, Maurıcio S. BolognaInstituto Nacional de Pesquisas Espaciais, INPE, CP 515, 12201-970 Sao Jose dos Campos, Brazil

Received 12 July 2005; received in revised form 12 December 2005; accepted 8 May 2006

Abstract

Electric anisotropy calculated from geoelectric strikes of magnetotelluric (MT) data and seismic anisotropy derived from shear-wave splitting parameters are jointly analyzed to estimate the degree and orientation of strain in the subcontinental mantle ofcentral-southeastern Brazil. High-quality long-period MT soundings are available at 90 sites concentrated along four profiles atthe southwestern and southern borders of the Paleoproterozoic-Archean Sao Francisco craton. This data set is complementedby a previous study of SKS and SKKS splitting measurements available at more than 40 sites covering a slightly wider region.For this study, the MT data were processed with modern techniques, including recovery of the undistorted EM field polarizationsthrough correction of static shift and phase mixing due to galvanic distortions. Three-dimensional forward modelling of MT and GDS(geomagnetic depth soundings) transfer functions supports interpretation of deep electrical anisotropy in the region. Magnetotelluricphase responses of orthogonal propagation modes present slight splitting at long periods for most of the sites, indicative of overallelectrically low mantle anisotropy to depths greater than 250 km. The electrical strike azimuths are parallel to the fast NW directionsof shear-wave splitting along the southern borders of the Sao Francisco craton, suggesting that the seismic anisotropy also resideswithin the same depth range. Since this direction is very different from that of present-day South American westward absoluteplate motion, it is inferred that no significant lateral mantle flow or deformation related to the plate motion is observed under thestudy area, either because it is absent or the rigid lithosphere is much thicker than expected. To explain the prevailing electric andseismic common direction it is suggested that a mechanical coupling between lithospheric and sublithospheric mantle exists. Arelic strain, possibly resulting from ancient continental collision processes, is interpreted to have induced a general alignment ofolivine down to mantle depths beneath the continental rigid plate. The observed slightly enhanced conductive texture could beassociated with hydrogen diffusion along the aligned olivine a-axis, especially within mantle patches subjected to metasomaticprocesses. The coincidence of mantle strikes with the trend of surface deformation pattern also suggests that crust, lithospheric
and sublithospheric upper-mantle have deformed and translated coherently, preserving the regional NW direction since the tectono-thermal event responsible for the deformation. Distinct electric azimuths from the general NW trend are observed in regions probablyperturbed by Cretaceous rift-related intraplate magmatism and in zones of complex deep structures produced by superposed nearly
st of th ancisco craton.
simultaneous oblique Neoproterozoic collisions in the southea © 2006 Elsevier B.V. All rights reserved.
Keywords: Upper mantle; Central-southeastern Brazil; Magnetotellurics; Ele

∗ Corresponding author. Tel.: +55 12 3945 6791; fax: +55 12 3945 6810.E-mail address: [email protected] (A.L. Padilha).

0031-9201/$ – see front matter © 2006 Elsevier B.V. All rights reserved.doi:10.1016/j.pepi.2006.05.006

e Sao Fr

ctrical anisotropy; Lithosphere/asthenosphere coupling

mailto:[email protected]

dx.doi.org/10.1016/j.pepi.2006.05.006

h and P

1

pdlaocta1awwlJboK

tow(dptltayleh2

fiptarpofcca(ttz

A.L. Padilha et al. / Physics of the Eart

. Introduction

Observations of seismic and electrical anisotropyrovide complementary approaches to estimate mantleeformation and ultimately address a wide range of geo-ogical and geophysical problems. Shear-wave splittingnalysis gives information on the azimuthal anisotropyf the upper mantle, which is generally interpreted as aonsequence of the strain-induced crystallographic tex-ure or lattice preferred orientation (LPO) of intrinsicallynisotropic mantle minerals, principally olivine (Silver,996; Savage, 1999). However, the exact location of thenisotropy within the upper mantle is poorly constrained,ith some researchers interpreting it to reside primarilyithin the lithosphere and associated with ancient geo-

ogical processes (Silver and Chan, 1988; Gaherty andordan, 1995), whereas others suggest it is maintainedy shear deformation in the asthenosphere as a resultf mantle convection (Vinnik et al., 1992; Debayle andennett, 2000).

Similarly, electrical anisotropy may be generated inhe upper mantle either by a preferred interconnectionf a highly conducting mineral phase (such as graphite)ithin foliation planes that mark past tectonic events

Mareschal et al., 1995), or by strain-induced hydrogeniffusion along olivine crystals oriented by present-daylate motion (Simpson, 2001). As the depth to an elec-rically conductive anisotropic-layer is well constrained,ong-period magnetotelluric (MT) soundings can be usedo resolve the ambiguity in interpretation of seismicnisotropy measurements. However, a careful data anal-sis must be performed because in some circumstancesithospheric anisotropy estimated from long period geo-lectric strikes can alternatively be explained by crustaleterogeneity (e.g., Korja, 2003; Heinson and White,005).

To date, few MT studies with sounding periods suf-ciently long to resolve mantle anisotropy have beenerformed in areas with available teleseismic observa-ions. In the Superior Province of Canada, Mareschal etl. (1995) identified a very significant difference in MTesponse in the two orthogonal directions without theresence of any vertical field response. SKS data fromne region across the Grenville Front indicated that theast seismic direction was almost parallel to the highonductivity direction, yet there was a small but statisti-ally meaningful obliquity between them (Senechal etl., 1996). This obliquity was interpreted by Ji et al.
1996) as an indicator of the movement sense on duc-ile mantle shear zones. However, a recent collocatedeleseismic and MT study across the Great Slave shearone, northern Canada, did not provide evidence for
lanetary Interiors 158 (2006) 190–209 191

a systematic obliquity between seismic and electricalanisotropy in the upper mantle (Eaton et al., 2004). Forthe North Central craton of Australia, Simpson (2001)demonstrated the presence of an electrical anisotropydeeper than 150 km. The direction of electromagneticstrikes revealed good agreement with the fast directionsof SV-waves (Debayle and Kennett, 2000), but both geo-physical datasets showed significant angular discrepancywith the present-day absolute plate motion (APM).

In central-southeastern Brazil, portable broadbandstations derived upper mantle seismic anisotropy frommeasurements of SKS and SKKS splitting at more than40 sites. The fast polarization directions show consistentorientation over hundreds of kilometres and are gen-erally parallel to the structural trend of the last majororogeny (James and Assumpcao, 1996; Heintz et al.,2003). A long-term MT programme is under way in thesame region, aiming at a large-scale reconnaissance ofmajor conducting geostructures that could have persistedas records of past episodes (Padua, 2004; Bologna et al.,2005, 2006). Ninety high-quality MT soundings havebeen deployed along four profiles mainly located in thesouthwestern and southern borders of the Sao Franciscocraton. At most of the stations, the data period rangesfrom 0.0008 to 13,653 s, which allows the vertical imag-ing of geoelectric structures from the near surface (tensof metres) to great depths into the upper mantle (morethan 250 km).

In this paper, these MT data are analyzed usingthe most current techniques to get band-limited strikedirections for periods most sensitive to differentlithospheric/asthenospheric depths. Because the MTresponses at long periods are sensitive both to deepand to distant structure, dimensionality analysis, tensordecomposition and three-dimensional (3D) modellingare carried out in order to understand the effects of crustalheterogeneity and electrically anisotropic structures inthe crust and in the upper mantle. Selected geoelectricstrikes at typical depths of the crust, upper lithosphericmantle and deeper lithosphere/asthenosphere are thencompared with seismic anisotropy for interpretations ofthe structural deformation below the study area.

2. Geological context

The South American platform is composed of a Pre-cambrian central core, bordered by active subduction-related orogens to the west and northwest, and of a
Mesozoic-Cenozoic passive continental margin to thenortheast and east. The core is formed by several Archeanto Mesoproterozoic blocks amalgamated during the Neo-proterozoic Brasiliano (Pan African) orogeny in the final

192 A.L. Padilha et al. / Physics of the Earth and Planetary Interiors 158 (2006) 190–209

Schobr the M

soprote

Fig. 1. Generalized geological map of the study area (modified fromTocantins province, SFP for the Sao Francisco province, and MQP fodifferent grey tones represent: 1, Archean; 2, Paleoproterozoic; 3, Me

assembly of Gondwana (Almeida et al., 1981; Alkmimet al., 2001). In central-southeastern Brazil (Fig. 1), theSao Francisco craton constituted the centre of the WestGondwana and its margins were initially formed duringthe Archean in several discrete events between 3.2 and2.7 Ga. The Transamazonian orogeny left its imprint onthe southern and eastern parts of the craton between 2.1and 1.9 Ga, creating a NNW-verging fold-and-thrust beltthat was later incorporated in a dome-and-keel province(Alkmim and Marshak, 1998).

Following Machado et al. (1996), four princi-pal stratigraphic units occur around the study areawithin the Sao Francisco craton: the Archean base-ment (granitoid intrusives, gneisses, and migmatitesthat range in age from 3.2 to 2.7 Ga), an Archean(ca. 2.7 Ga) supracrustal assemblage (greenstone andassociated sedimentary sequences), and two Paleopro-terozoic supracrustal assemblages (platformal and deepmarine strata including quartzite, phyllite, carbonate, andLake Superior type banded iron formations, depositedbetween 2.1 and 2.4 Ga, and more recent quartzite andconglomerate). The basement occurs in dome-shapedbodies separated by deep keels (troughs) containingdeformed supracrustal rocks. All margins of the cra-
ton were subjected to Brasiliano deformation during theNeoproterozoic and earliest Phanerozoic (Cambrian),creating fold-thrust belts that generally verge towardsthe interior of the craton.
benhaus et al., 1984). PB stands for the Parana basin, TOP for theantiqueira province. Black dots are the location of the MT sites and

rozoic; 4, Neoproterozoic; 5, Phanerozoic; 6, water.

The Tocantins Province is a large NeoproterozoicBrasiliano orogen developed between three major con-tinental blocks represented by the Amazon and SaoFrancisco/Congo cratons and another continental block,presently covered by the Phanerozoic sedimentary andvolcanic rocks of the Parana basin. The province com-prises three large fold belts: the Araguaia and Paraguaybelts, which border the eastern and southeastern mar-gin of the Amazon craton, respectively, and the Brasıliabelt, established along the western margin of the SaoFrancisco craton.

Proterozoic metamorphic rock units of varied natureand age constitute most of the Brasılia belt, comprisingpassive margin sequences of the Sao Francisco continent,back-arc and fore-arc basin sequences, and a youngerpost-inversion platform sequence deposited in a fore-land basin of the Sao Francisco craton (Pimentel et al.,2001). In the southern segment of the belt, the Brasilianodeformation involved overthrust sheets (nappes), trans-ported eastward at least 150 km towards the platformof the Sao Francisco craton, and subsequently stackedby thrust faults over Neoproterozoic pelitic and carbon-atic sequence (Fuck et al., 1994). In the study area, theBrasılia belt forms a complex structural pattern, locally
characterized by foliations associated with the thrustsheets and long lineaments corresponding to sub-verticallateral ramps or wrench faults originated from differen-tial displacement of the thrust wedges.

h and P

tPanto

cetHvotwcbgla

dTadoimcfswstt5tbtrm(

tfiasTgdt


Extensive Cretaceous magmatism is observed inhe southern portion of the Brasılia belt and northernarana basin, possibly related to several other Cretaceouslkaline-carbonatite provinces that evolved contempora-eously around the Parana basin during the opening ofhe South Atlantic Ocean and subsequent westward driftf the continent (Thompson et al., 1998).

The vast region located between the Sao Franciscoraton and the eastern continental margin of Brazil isncompassed by the northern and central sectors ofhe Neoproterozoic Mantiqueira Province (Almeida andasui, 1984). This N–NE trending structure has beeniewed as part of the Aracuaı-West Congo orogen, devel-ped between the Sao Francisco and Congo cratons inhe course of the Brasiliano assembly of West Gond-ana during the Neoproterozoic (Trompette, 1997). It

an be divided into the largely greenschist and amphi-olite facies Aracuaı belt on the west and the largelyranulite facies Ribeira belt on the east. A pronouncedinear gravity and magnetic anomaly defines the bound-ry between these two belts.

The major tectonic framework of the Ribeira Belt isefined by two distinct terrains (Heilbron et al., 2000).he occidental terrain comprises a pile of superposedllochtonous terrains thrust to the west and subsequentlyeformed in transpressional regime, with large verticalblique shear zones associated with granitic plutons. Its considered as the early Sao Francisco craton passive

argin. The oriental terrain is characterized by large iso-linal recumbent folds, low angle dipping metamorphicoliation, and numerous NW trending ductile-ruptilehear zones containing post-collisional granitoids. To theest, the Ribeira belt deformation partially overprints the

lightly older Brasılia belt. The main collisional event ofhe Brasılia belt occurred around 625 Ma, whereas inhe Ribeira belt the main collision took place around90 Ma (Schmitt et al., 2004). This diachronous evolu-ion is registered by an interference zone between theseelts, to the south of the Sao Francisco craton, where low-o medium-pressure metamorphic mineral assemblages,elated to the Ribeira belt, overprint higher pressureetamorphic mineral assemblages of the Brasılia belt

Trouw et al., 2000).To the west, the Parana basin is a large intracra-

onic basin, developed entirely on continental crust andlled with sedimentary and volcanic rocks ranging inge from Silurian to Cretaceous. Five major depositionalequences (Silurian, Devonian, Permo-Carboniferous,
riassic, and Juro-Cretaceous) constitute the strati-raphic framework of the basin. The first four are pre-ominantly siliciclastic in nature, and the fifth containshe most voluminous basaltic lava flows of the planet.

The depositional history of the basin was closed in theUpper Cretaceous by a package of alluvial, fluvial andeolian sedimentary rocks. Maximum thickness exceeds7000 m in the central depocentre. The sequences areseparated by basin wide unconformities related in thePaleozoic to Andean orogenic events and in the Meso-zoic to the continental breakup and sea floor spread-ing between South America and Africa. The structuralframework of the Parana basin consists of a remark-able pattern of linear features (faults, fault zones, arches)clustered into three major groups (N45–65W, N50–70E,E–W). The northwest- and northeast-trending faults arelong-lived tectonic elements inherited from the Precam-brian basement whose recurrent activity throughout thePhanerozoic strongly influenced sedimentation, faciesdistribution, and development of structures in the basin(Zalan et al., 1990).

3. Magnetotelluric data acquisition andprocessing

Over the last 7 years, the natural source MT methodhas been extensively used in the central-southeasternregion of Brazil to define the southwestern-southernboundary of the Sao Francisco craton and to determinethe electrical properties of the mantle beneath this area.The MT data were acquired along four main profiles,roughly perpendicular to the tectonic grain of differentparts of the craton and its margins (Fig. 2).

On the API profile, 25 stations were recorded along180 km in a WSW–ENE direction. The profile crossesthe cluster of diatremes in the central part of the AltoParanaıba igneous province (APIP), with its westernlimit over the Parana basin and the eastern limit onthe Neoproterozoic sedimentary cover of the Sao Fran-cisco craton (Bologna et al., 2005). A commercialsingle-station broadband MT system (Metronix GMS05)was used at every site in a coordinate system withone of the horizontal axis aligned with the magneticmeridian (N20◦W). The telluric field variations weremeasured with 100 m dipoles, in a cross-configuration,with non-polarizable cadmium-cadmium chloride elec-trodes, whereas the magnetic fields were measured withinduction coils for the two horizontal and one verticalcomponents. The period range of this equipment is of0.0008–1024 s and recording time was typically of atleast 24 h. At 14 of the above stations, other commercialremote-referenced long-period MT systems (Phoenix
LRMT) were operated in the period range of 20 to at least13,653 s with recording time of at least 2 weeks underthe same geomagnetic coordinate system. The horizontaltelluric fields were acquired with lead–lead chloride elec-


udy areO for tare also

Fig. 2. Schematic outline of the main geological provinces in the stBrazilian belt, RB for the Ribeira belt, AB for the Aracuaı belt, and Astructural elements (faults and shear zones). Locations of main cities(BH).

trodes in a cross-configuration with 150-m lengths, whilethe three components of the magnetic field were acquiredwith high-sensitivity ring-core fluxgate sensors. At siteswhere both systems were used, the data was merged toobtain more than seven-decade period range.

The 230-km-long PIU profile, in an ENE–WSWdirection, started on the Brasılia belt, in the western side,and ended on the exposed Archean basement of the SaoFrancisco craton. The main profile comprised 16 single-station broadband and 12 remote-referenced long-periodMT stations. A secondary branch, with four broadbandand long-period measurements, extended the surveyfrom near the centre of the main profile in the northwest-ern direction, crossing a kimberlitic province within theBrasılia belt. The long-period data were recorded withthe same systems used in the API profile, but the broad-band data were measured with a new generation MTsystem (Metronix GMS06). Recording times and use-ful period ranges were the same as in the API profile.For both broadband and long-period systems, the telluricfield variations were measured with lead–lead chlorideelectrodes with 150 m dipoles.

The same systems and field procedures were usedin the 160-km-long SJR profile in a roughly NS direc-tion. The profile extended from Neoproterozoic parag-
neissic rocks in the Ribeira belt, at the southern end,across a Mesoproterozoic pegmatitic province of the SaoFrancisco craton, and ended on the exposed Archeanbasement of the craton. Broadband MT data were
a with the identification of the four MT profiles. BB stands for thehe Atlantic Ocean. Grey traces are the location and direction of main

indicated: Sao Paulo (SP), Rio de Janeiro (RJ), and Belo Horizonte

acquired at 19 sites from which 11 also had long-perioddata.

For the ∼550-km-long IBI profile, a new set of instru-ments was available, including another GMS06 and atime domain electromagnetic (TDEM) system (ZongeGDP-32II). At most stations, time series in the broad-band range were recorded simultaneously at pairs ofsites in order to use remote referencing for noise reduc-tion in the analysis (Gamble et al., 1979) and TDEMdata were acquired for static shift correction of the MTsoundings (Meju, 1996). The EW profile comprised 27broadband and 22 long-period stations and ran from thePhanerozoic sediments and volcanics of the Parana basinto the west, across a regional fold with WNW plung-ing in the Neoproterozoic Brasilia belt in the centre,and the Neoproterozoic/Phanerozoic cover and exposedArchean complex of the Sao Francisco craton to the east.

After rotating the time series to geographic coor-dinates, modern techniques of robust processing wereapplied to the data to yield MT impedance tensor andvertical field geomagnetic transfer function estimates asa function of period for each site. Data from the GMS05system were previously edited to remove noisy seg-ments and robust processing was used to calculate singlesite estimates of the electromagnetic transfer functions
through the commercial software PROCMT (Metronix,Germany). For the GMS06 and the long-period systems,the time series data were processed using remote refer-ence data when available (Egbert, 1997).

h and P

mdttAeep

tfwoslac

Fwasi

Fr


The long-period and broadband MT data were thenerged into single responses for each station. Due to

ifferences in electrode array layout, the apparent resis-ivity curves of the long-period MT data were shiftedo the level of the broadband data at some sites of thePI profile. However, the absolute levels of the appar-

nt resistivity curves at each site (static shifts) werestimated as part of a two-dimensional (2D) inversionrocedure (Bologna et al., 2005).

Due to lack of significant cultural EM noise at most ofhe sites and the high signal levels, high quality responseunction estimates were obtained at most sites for thehole period range, with errors in the impedance phasesf less than 2◦. The only exceptions were observed atome sites in the centre of the API profile, at sites near thearge Furnas Dam in the western side of the PIU profile,nd at the two northernmost sites of the SJR profile, thelosest to the large Belo Horizonte city.

Typical examples of high-quality data are shown inig. 3, including one site from each profile. Site API78
as located close to the border of the Parana basin. Phase
nd apparent resistivity curves in orthogonal directionshowed a small split in the 0.01–100 s period interval,ndicative of a departure from the one-dimensional (1D)

ig. 3. MT responses derived from data acquired at sites representative ofesistivities and phases are from electric dipole oriented at geographic NS (tri


condition in the crust. IBI121 was positioned at the west-ern end of the IBI profile, over a thicker sedimentary-volcanic package of the Parana basin. Sites in this regionpresented a phase split at periods between 100 and 1000 s(not seen in the unrotated data of Fig. 3, but shown inFig. 4) related to structural or intrinsic anisotropy at thebase of the crust. PIU06 was located near the centre ofthe PIU profile, over a thin Phanerozoic cover of the SaoFrancisco province and close to one of the teleseismicstations. A significant split was observed in the phase andapparent resistivity at short periods, indicative of mul-tidimensional structure at shallow depths. SJR03 wassituated in the southern part of the SJR profile, overthe metamorphic rocks of the Ribeira belt. The short-est periods showed apparent resistivity split and diver-gent phases (lower than 45◦), probably associated withinductive distortions in the electric field caused by thecontrasting heterogeneous topsoil cover of the resistivePrecambrian rocks.

The most striking result in these graphs is the
behaviour of the phase data for periods longer than 100 sthat are very similar in orthogonal directions. Except-ing the sites in the western end of the IBI profile, thisresult was confirmed for rotated data of all MT soundings
each profile (three first letters as the site identification). Apparentangles) and EW (squares) coordinates.


Fig. 4. Upper panels are the G–B parameters for unconstrained 2D fit to long-period data of site IBI121: (a) normalized chi-square misfit; (b) strike,current channeling, shear and twist azimuths; (c) estimated scaled regional apparent resistivities (triangles in the strike direction and squares in theperpendicular direction); (d) estimated regional phases. Lower panels (e–h) are the fit of the model data to the observed data for all four impedanceelements (squares and solid lines are the real parts, triangles and dashed lines are the imaginary parts).

h and P

air

4

mfbmgmtoHhsls(1oloEme

udtoosepalttafet

(ampsi


nd indicated that the occurrence of electrical anisotropys not very pronounced at upper mantle depths for thisegion of the Brazilian shield.

. Geoelectric strike azimuths

The MT impedance tensor can be used to deter-ine the geoelectric strike direction, a useful parameter

or defining the different geoelectric structures detectedy the data. In the maximum phase split orientationethod, the impedances are rotated to determine the

reatest phase difference between the off-diagonal ele-ents. The geoelectric strike is then determined as

he orientation at which the phase difference betweenff-diagonal impedance tensor elements is maximized.owever, electric charge accumulation on near-surfaceeterogeneities can distort the measured MT responseo that the azimuth determined in this way may noonger accurately represent the regional conductivitytructure (Jones and Groom, 1993). In the Groom–BaileyG–B) tensor decomposition method (Groom and Bailey,989), the geoelectric strike is determined simultane-usly with the near-surface distortion effects using aeast-squares approach. The distortion model assumed isf a galvanic 3D distortion sheet overlying a regional 2Darth. Because of the number of parameters fitted in thisethod, stable estimates of the regional strike require its

stimation over relatively broad period bands.The geoelectric strikes in the study area were derived

sing the G–B tensor decomposition at nearly one-ecade-wide bands (six or seven adjacent periods) inhe intervals of 4–27 s, 40–320 s, 430–2560 s, and peri-ds longer than 3400 s. The tensor decomposition codef McNeice and Jones (2001) was used in the single-ite, multi-frequency option on the MT responses fromach of the sites, with an assumed error floor of 2◦ inhase. Fig. 4 shows the G–B model parameters and fit forn unconstrained best-fitting 2D parameterization of theong-period data of site IBI121. The upper panels showhe statistical residual (χ2) for the model, the G–B distor-ion parameters, and the 2D regional apparent resistivitynd phase estimates in the resulting strike-coordinaterame. The details of the misfit are given in the lower pan-ls, which show a scaled fit of the estimated impedanceensor under the G–B model compared with the data.

In Fig. 4, the χ2 variable presents a value less than 4horizontal line in Fig. 4a) for all periods, indicative ofn acceptable model of distortion for reliable error esti-
ates (Groom et al., 1993). The twist parameter is almost
eriod independent with values within ±5◦, whereas thehear is slightly larger at long periods but within the±15◦nterval. The strike angle is around 20◦ (or −70◦ due to


the 90◦ ambiguity) at shorter periods and decreases toa long-period azimuth of about 0◦ (or 90◦). Estimatedregional apparent resistivities and phases vary smoothlyover the whole period range, with scaterers and largererror bars at periods longer than 10,000 s due to lowsignal/noise ratio. The previously referred phase split atperiods between 100 and 1000 s is now clearly seen inthe rotated data. All off-diagonal and diagonal tensorelements are well fit by the distortion model.

Similar analyses were performed at every site. Gener-ally a good fit was observed to each element of the mea-sured impedance tensor at most of the sites. Chi-squaredmisfits were very small and shear and twist angles wereless than 15◦, an indication that the sites were relativelyundistorted and that the geoelectric structure in the sur-vey region fitted well with the 3D/2D approximation ofthe G–B decomposition. Exceptions were observed atthe noisy sites previously described and at some local-ized sites in different profiles. At these sites, twist andshear angles were very high, sometimes approaching thelimits of 60◦ and 45◦, respectively. An additional anal-ysis of the dimensionality of the impedance tensors wasmade using the Bahr’s classification (Bahr, 1991). Theresults indicated that most of the data fit a 2D modeldistorted by local heterogeneities with only a weak tostrong galvanic response (classes 3–5), consistent withthe G–B decomposition.

The strike azimuths from single site decompositionsare shown in Fig. 5 for the four bandwidths previouslydefined, with the lengths of the bars scaled by the averagephase difference over the band between the conductive(strike) and resistive directions. For the shortest periodband, the azimuths can be compared to the geologicaland tectonic features present at the surface as displayedin Fig. 2, which indicate that the main structural ele-ments trend predominantly NW and NNW directionsalong the western border of the Sao Francisco cratonand E–W to ENE on the southern border of the craton.Geoelectric strike directions in this band are, in general,consistent with these predominant surficial geologicaltrends of the region. The map shows stronger site-to-site variation, associated with the complex upper crustalstructure that includes transport of large allochtonous ter-rains during the Neoproterozoic remobilization. Largestlateral variations are observed in the API profile, with thegeneral NW direction prevailing in the southwestern endof the profile, oscillating from WNW to E–W beneathsites in the central area, and NNE off to the northeast-
ern region. This anomalous feature is interpreted to berelated to the extensive magma emplacement during theCretaceous period (Bologna et al., 2005). For the nextthree bands, strikes were defined through the direction


Fig. 5. The unconstrained G–B regional strike from four period bands: (a) 4–27 s; (b) 40–320 s; (c) 430–2560 s; (d) >3400 s. The azimuths of theeriod bl 2D im

solid bars illustrate the preferred geoelectric strike direction for that pdifference between the off-diagonal elements of the recovered regiona

of maximum phase. The maps show good site-to-siteconsistency, with a remarkable coherence in the strikedirections from the different profiles. However, a largenumber of sites exhibit small phase differences in orthog-onal directions, implying a weak lateral variation of theelectrical resistivity beneath the area. This behaviour ismore evident at the sites on the western side of the SaoFrancisco craton, along IBI and PIU profiles. One signif-icant rotation of the strike directions is observed at thesouthernmost sites of the SJR profile, where azimuthschange from nearly E–W in the three first bands to NWat the longest period band.

Other ways of assessing geoelectric strike and dimen-sionality of the region are provided by induction vectors.They are determined independently only from the mag-netic field variation recorded at each MT sounding. Overa 2D structure, the real part of the induction vectors areorthogonal to the geoelectric strike and can be used toresolve the 90◦ ambiguity in the strike directions derivedfrom tensor decomposition. Over more complex struc-ture, induction arrows can be influenced by regional
conductivity anomalies outside the MT study region gen-erating a lack of orthogonality between strike and arrowsorientation (Simpson and Bahr, 2005). Fig. 6 shows mapsof the real induction vectors at periods representative of
and and the lengths of the bars are proportional to the average phasepedance tensor.

the four bands for which geoelectric strikes were calcu-lated. The vectors were derived from magnetic variationsmeasured with fluxgate magnetometers (long period MTsystems). The maps for the two longest periods alsoinclude data from a regional magnetometer array study(Subba Rao et al., 2003), also using fluxgate magnetome-ters (Chamalaun and Walker, 1982).

At the two shortest periods the vectors are almostidentical, indicating significant current concentration inthe region of the Alto Paranaıba volcanics on the APIprofile and pointing towards larger conductance to thecentre of the Parana basin, on the western side of theIBI profile. The other soundings over the Brasılia beltand the Sao Francisco craton have negligible verticalresponse at 20 s, and point towards a conductive structurebetween the PIU and SJR profiles at 106 s. Vectors fromsites over the Ribeira belt points to the south, parallel tothe SJR profile. At the two longest periods, vectors arenear zero all over the northwestern side of the study areaand point southeast to the central and southeastern area,towards the electric currents concentrated at the ocean-
land boundary. The significance of these results will bediscussed in more details in a later section that willinclude a 3D forward modelling of a sketchy regionalconductance map.

A.L. Padilha et al. / Physics of the Earth and Planetary Interiors 158 (2006) 190–209 199

F k bars)a (for GD

5

wTsagHttiEroBots2

dsTm

ig. 6. Real induction vectors calculated from long-period MT (blacre: (a) 20 s; (b) 106 s; (c) 1280 s; (d) 6826 s (for MT data) and 7680 s

. Depth estimation

MT short periods sample near-surface structureshereas long periods are sensitive to deeper structures.herefore, the period dependence of the geoelectrictrike can be used to distinguish between lithosphericnd sublithospheric effects, helping to resolve the ambi-uity associated with the depth of seismic anisotropy.owever, due to the diffusive nature of EM propaga-

ion and the vast range in electrical conductivity insidehe Earth, one must be careful that the MT and seismicnformation derives from the same depths. Also, there arearth structures for which EM signal penetrations have

adically different depths for the two orthogonal modesf propagation (as for instance, site PIU06 of Fig. 3).ecause of this, comparisons of the MT responsesbserved at the same period range from different loca-ions with seismic anisotropy derived from shear-waveplitting analysis are not straightforward (see also Jones,006).

MT depth estimations can be hindered by near-surface
istortions of the electric field amplitudes, causing atatic shift in the impedance magnitudes. For sites whereDEM data were unavailable, the static shifts were esti-ated as part of the 2D inversion procedure for each
along the profiles and GDS (grey bars) from a wider survey. PeriodsS data). Vectors are reversed to point towards current concentrations.

profile (Bologna et al., 2005; Padua, 2004). The approachused was to fit well the phase data at all stations and theapparent resistivities from stations without static shift,yet allowing a larger misfit to the apparent resistivitydata of static-shifted stations (Wu et al., 1993). The staticshift values estimated in this way were used to correcteach site prior to depth estimation.

However, one cannot assume that the electric fieldis distorted only by these near-surface heterogeneitiesbecause conductive structures present in the middle andlower crust can affect the electric field at longer peri-ods in similar way (Bahr et al., 2000). As the galvanicdistortions of these structures at different depths are sum-marized at long periods, they can be interpreted as asingle distorter. The total static shift effect can then becorrected using a global sounding curve (Vanyan et al.,1980) or magnetic transfer functions derived from Sqvariations (Bahr et al., 1993). These techniques for cor-rection of the distorted apparent resistivity data are oflimited applicability in complex 3D medium inhomo-geneities and when array data are not available.

To avoid inaccurate correction for these galvanic dis-tortions we opted for a conservative approach, excludingfrom the analysis, at a given period band, all geoelectricstrikes affected by conductive bodies at shorter periods.

h and P
200 A.L. Padilha et al. / Physics of the Eart
Also, because G–B decomposition gives stable resultsonly if the phase split is much greater than the errors ofphase measurements, we excluded from the analysis thestrikes for which the 2D regional impedances do not havesignificantly different phases. Following Berdichevsky(1999) and considering the mean deviation of 2◦ in theimpedance phases, we chose 7◦ as the minimum phasesplit for a valid G–B decomposition.

The penetration depths of the MT signals in the fourperiod bands were estimated using skin depth relationsand the method of Schmucker and Jankowski (1972).Calculations were checked with an approximate equiv-alent depth using the formulation of Niblett and Sayn-Wittgenstein (1960).

Periods of 4–27 s correspond to penetration to thebase of the crust, excepting at the westernmost sites of theIBI profile where larger conductance of the Parana basinlimit penetration to the middle crust. MT strikes in thisband are therefore representative of the middle and lowercrust. Periods of 40–320 s typically penetrated to depthsbetween 50 and 100 km into the upper mantle. Again,penetration is smaller at some sites to the centre of theParana basin. For the 430–2560 s band, most sites in theRibeira and Brasılia belts and some isolated sites close to
the border of the Parana basin have EM fields penetratingto the base of the lithosphere (depths of 150–200 km).Surprisingly, most sites over the exposed Archean rocksof the Sao Francisco craton do not have penetration to
Fig. 7. Azimuths and amplitudes of shear-wave splitting (open bars) and geobetween 10 and 40 km. For electrical measurements, the solid bars indicate thphase difference between the most conductive and orthogonal directions. Forthe fast shear wave and their lengths are proportional to the delay time betwee(APM) of the South American plate is also indicated.

lanetary Interiors 158 (2006) 190–209

such depths at these periods. This is associated withan anomalously large conductivity of the upper man-tle beneath this region, probably affected by Mesozoictectono-thermal reactivation (Padua, 2004). For periodslonger than 3400 s, some sites in the Ribeira and Brasıliabelts, two sites close to the Parana basin border and onlyone site over the Archean Sao Francisco craton havepenetration deeper than 250 km, providing informationprobably about sublithospheric depths.

5.1. Crustal strike

Fig. 7 presents a compilation of seismic anisotropyderived from measurements of SKS and SKKS split-ting in Southeast and Central Brazil. Description ofdata acquisition, processing and details of interpreta-tion were given by James and Assumpcao (1996) andHeintz et al. (2003). The figure also shows the direc-tion of present-day absolute plate motion (APM) forthe HS3-NUVEL1A model (Gripp and Gordon, 2002),which is determined assuming that hotspots are station-ary relative to the deep mantle below the asthenosphere.There is a significant discrepancy between the seismicanisotropy and the APM direction, with the fast polar-
ization direction of the seismic waves showing preferredorientation generally parallel to the last major oroge-nies. In the Brasılia belt, this direction has a NW trend inagreement with a proposed collision between a cratonic
electric anisotropy (solid bars), calculated for EM penetration depthse most conductive directions and their lengths are proportional to theteleseismic data, the open bars indicate the polarization direction ofn the two split waves. Direction of present-day absolute plate motion

h and P

bctwo(sStmcet

f(fbbtgcmsasateft


lock beneath the Parana basin and the Sao Franciscoraton, in the earliest stages of Gondwana amalgama-ion, by which the southern section of the Brasilia beltas formed (Alkmim et al., 2001). In the southern partf the Ribeira belt, the anisotropy appears to be stronger>1 s) with a WSW direction, parallel to the trend of con-picuous transcurrent shear zones. To the south of theao Francisco craton, in the interference zone between

he Brasılia and Ribeira belts, the anisotropy pattern isore complex, with no predominating direction. In the

entral part of the Parana basin, where also there are nolectromagnetic data, the anisotropy direction tends tourn to E–W.

Fig. 7 also presents the selected geoelectric strikesor EM signals with typical penetration at crustal depths10–40 km). As previously explained, they were derivedrom the period band between 4 and 27 s. At the southernorder of the Sao Francisco craton and along the Brasıliaelt, geoelectric strike is consistent with the main direc-ions of the seismic anisotropy and the regional structuralrain. The general NNW trend in the centre part of theraton is in agreement with the directions of emplace-ent of mafic dykes in this region. There are no tele-

eismic data in the APIP volcanic region (API profile)nd the conspicuous lateral variation of the geoelectrictrike is related to differences in the Mesozoic intrusivend extrusive magmatism affecting mid-crustal struc-
ures in different ways along that MT profile (Bolognat al., 2005). No electric strike information is availableor most of the Ribeira belt except along the region nearhe southern cratonic border.
Fig. 8. The same as in Fig. 7, for geoelectric anis


A significant disagreement between seismicanisotropy and geoelectric strike directions is observedat sites over the northeastern Parana basin. In thiscase, the geoelectric data follow the known structuraltrend beneath the basin, given by gravity anomalies.The position and direction of the largest geoelectricstrike bars are consistent with a steep gravity gradientinterpreted as an ancient suture zone (Lesquer et al.,1981). Yet, towards the central part of the Parana basin,no electrical azimuth is available.

5.2. Topmost upper mantle strike

Fig. 8 presents geoelectric strikes derived from EMsignals with penetration at depths between 50 and100 km. Results in the southern part of the Sao Franciscocraton are similar to the ones from the crust, rotatingapproximately from WSW in the limits of the Ribeirabelt to E–W or WNW inside the Sao Francisco province.At sites on the exposed Archean rocks of the craton, thegeoelectric strikes point NW, in accordance with the seis-mic data from this region. In spite of the rigorous dataselection, some stations over the Passos nappe (PIU pro-file) still appear to be affected by noise and the generalNW trend is not easily seen in this region.

A remarkable alignment in the WNW direction is seenin the electrical azimuths of the IBI stations over the
Brasılia belt. This direction is the same as for the faultzones in the area. The gravity gradient is still sensed atthese depths, indicating that the suture zone extends deepinto the upper mantle. To the west of the gradient, the
otropy at depths between 50 and 100 km.


ric aniso
Fig. 9. The same as in Fig. 7, for geoelect
vectors rotate smoothly to the NW direction, tending toalign with the seismic anisotropy direction.

5.3. Lowermost lithosphere strike

Fig. 9 shows the geoelectric strikes at depths in theinterval of 150–200 km. There is an excellent correlation
with seismic anisotropy at every region where both datasets are available. The regional NW main direction isonly disturbed at the southern part of the Sao Franciscocraton, where the E–W trend observed at lower depths
Fig. 10. The same as in Fig. 7, for geoelectric a

tropy at depths between 150 and 200 km.

still persists, and at the northeastern end of the API pro-file, where rift-related Mata da Corda volcanics outcropsand the direction of the geoelectric strike seems slightlyrotated to NNW.

5.4. Sublithospheric strike

Fig. 10 presents geoelectric strikes at the sites withEM signals reaching depths greater than 250 km, pre-sumably in the sublithospheric mantle. Generally, thedirections are still in agreement with the seismic data

nisotropy at depths greater than 250 km.

h and P

btwaptarrsNmt

6r

dthtqsro

ticwebTdt2bePsdtadadeia

a


ut the general NW trend is significantly different fromhe absolute plate motion and is therefore inconsistentith interpretations in terms of plate-scale mantle flow or

sthenospheric deformations. Local differences in com-arison to the seismic anisotropy are observed in theransition area where the southern Sao Francisco cratonbuts the western Ribeira belt, characterized by a clearotation in the direction of the geoelectric strike fromoughly E–W at lithospheric depths to NW at the deeperublithospheric mantle. Also, in the APIP region, theNW electric trend does not agree with the E–W seis-ic anisotropy observed at one station about 150 km to

he north of the API profile.

. Numerical simulation of GDS and MTesponses

To associate unequivocally the observed MT strikeirections with the presence of anisotropy, it is necessaryo disqualify possible effects related to regional-scaleeterogeneities. A 3D forward modelling was used toest different conductivity structures in an attempt to fitualitatively the available GDS transfer functions andome characteristics of the MT responses. The surfaceesponse of the 3D models was calculated using the codef Mackie et al. (1994).

The model was built with a first layer representinghe upper crust, from the surface to a depth of 5 km,ncluding ocean bottom sediments and seawater. Theontinental shield consists of a highly resistive crust,ith about 30 S of total conductance. It was not nec-

ssary to include continental sediments in the modelecause they have minor influence on long period data.his surficial layer was embedded in a layered 1D model,erived as an approximate average from the models ofhis portion of the Brazilian shield (Bologna et al., 2005,006; Padua, 2004). The resistive continental shield isordered by the highly conductive Atlantic Ocean to theast and southeast (conductance of 10,000 S, followingadilha et al., 2002). To test the anisotropy hypothe-is, an anisotropic layer was included at upper mantleepths, represented by alternating conductive and resis-ive dyke-like medium. The ratio of integrated lateralverage conductivities (across and along strike) repro-uces the value of the anisotropic layer and the strikengle of the best conductor is around N65◦W, in accor-ance with the seismic anisotropy. To simulate the het-rogeneity hypothesis, 2D regional-scale structures were
ncluded at the continental lower crust. A trial-and-errorpproach was used in an attempt to fit the available data.
The results of the numerical calculations for thenisotropy hypothesis are summarized in Fig. 11. In


spite of the quite simple 3D model chosen to representthe regional structure there is a remarkable agreementbetween synthetic and experimental data. The inductionarrows at most of the sites are clearly controlled by thebehaviour of the coast effect over most of the study area.The GDS response is large close to the boundary betweenthe seawater and the land mass and decrease away fromthat boundary, vanishing at the distant sites. The fig-ure also shows geoelectric strikes determined from theimpedance tensor response at a selected site in the sur-face of the model. Consistent with the 1D layered model,the derived strike directions tend to align around −65◦(N65◦W) at periods sensing the anisotropic layer. Thisdirection is not normal to the observed and calculatedinduction arrows at this site (difference of 20◦ betweenthe azimuths of the arrows and the geoelectrical strike).These results indicate that the ocean accounts for thetipper magnitude and direction, whereas the anisotropiclayer in the upper mantle is responsible for the MT strike.

On the other hand, simulations with the crustal het-erogeneity model were unable to reproduce satisfactorilythe GDS and MT results. In fact, the coast effect that con-trols the magnetic transfer functions is not overprintedby any significant inland regional conductivity anomaly.It must be observed that localized crustal conductors areundetected at the regional scale of these data (spacingaround 100 km between stations), but small structuresare unlikely to generate the regional phase split patternobserved at long periods (see Fig. 5).

As the influence of large-scale lower crustal conduc-tors appears to be negligible in this region of the Brazilianshield, upper mantle anisotropy is favoured to explainstrikes and phase split in the long period data. Thisresult is at variance to that observed, for instance, in theFennoscandian shield (Korja, 2003). That shield hostshigh conducting elongated belts, possibly signatures ofprevious tectonic processes, concentrating induced cur-rents that can generate the great majority of the strikesand apparent anisotropy observed around them. Simi-lar structures are not observed in our study area whereconductive lithospheric bodies appear as isolated pock-ets, most of them related to the voluminous magmatismthat occurred from Early Cretaceous until Eocene times(Bologna et al., 2005, 2006; Padua, 2004). Such charac-teristic is independently confirmed by regional seismictomography and geoid anomaly studies (Schimmel et al.,2003; Molina and Ussami, 1999).

7. Discussions

In general, the spatial behaviour of phase splittingand amplitude of induction vectors help to distinguish


Fig. 11. Summary of 3D modelling to test the anisotropy hypothesis. (a) Map of conductance of the surface layer (0–5 km); (b) lateral plan view) and ating wi1280 s

ey and

of the 1D layered model incorporating the surficial layer (thin sheetN65◦W and resistivities of blocks within the anisotropic layer alternaand calculated (grey arrows) real GDS induction arrows for a period ofaccording to Swift (1967) and Bahr (1991), shown, respectively, as gr

between intrinsic anisotropy and the presence of lat-eral structure as the cause of geoelectric strike. In thelaterally displaced conductivity case, the phase split-ting will decrease with increasing site distance from theanomaly, whereas in deep anisotropy the splitting willremain the same over large distances. On the other hand,magnetic fields are generated by lateral conductivity gra-dients. Therefore, the absence of a significant inductionvector or phase splitting not correlated with the geomag-netic transfer functions supports the hypothesis of deepanisotropy.

MT data imaging crustal depths along the API profileand those crossing a gravity-defined suture zone beneath
the Parana basin have significant lateral variation in thegeoelectric strike and expressive induction vectors atshort periods. Both are clear examples of lateral struc-ture controlling the geoelectric strike. On the other hand,
n anisotropic layer from 100 to 160 km depths; anisotropy strike ofth 300 and 30 � m; (c) comparison between observed (black arrows); (d) strikes from synthetic MT data at the site circled in (c), calculatedblack squares.

the general NW geoelectric strike that pervades from thecrust down to sublithospheric depths and is not associ-ated with locally generated vertical magnetic field at longperiods are here interpreted as being related to regionalgeoelectric anisotropy. Small phase splitting at periodslonger than 100 s at most of the MT soundings suggestsa weak electric anisotropy in the upper mantle beneathmost of the region.

At sites where phase splitting is small even within thecrust, it is possible to use 1D inversion to derive the rangeof anisotropic conductance that can fit the data. An exam-ple of 1D layered-earth inversion is shown in Fig. 12 for asite in the southern part of the Sao Francisco craton, using
data rotated to the upper mantle strike direction. Theapproximate depth transformation (Niblett and Sayn-Wittgenstein, 1960) is also shown and indicates that pen-etration depths of TE and TM data are virtually the same

A.L. Padilha et al. / Physics of the Earth and Planetary Interiors 158 (2006) 190–209 205

Fig. 12. Results of 1D layered-earth inversion of rotated apparent resistivity and phase data for site SJR03 of Fig. 3. Left panel shows comparisonsb -strikee d approS

aaotaudastgpmcflaliBneds

its2

etween the observed and calculated along-strike (black) and acrossarth models that fit the data (grey zone is the anisotropic zone) anayn-Wittgenstein, 1960).

t depths around 100 km. An anisotropic layer is requiredt depths greater than 120 km, with a conductance ratiof about 3. This is a typical result, with anisotropy fac-ors not exceeding 3 for most of the sites. Such smallnisotropy is readily explained by olivine LPO in thepper mantle, with conductivity locally enhanced by theiffusion of hydrogen resulting from addition of smallmounts of water in the olivine crystal lattice at regionsubjected to mantle metasomatism. It must be notedhat this hypothesis requires that some remnant hydro-en would be retained in the upper mantle for the longeriod since the last tectono-thermal event liable for theetasomatism. Other alternative sources of enhanced

onductivity in the upper mantle include a free aqueousuid phase, conductive minerals (graphite, sulphides)nd partial melts (Jones, 1999). However, they are lessikely to explain the observed conductivity distributionn this region of the Brazilian shield (see discussion inologna et al., 2006). These qualitative observations stilleed to be confirmed by quantitative modelling, linkinglectric and seismic anisotropy to anisotropic hydrogeniffusion in olivine, considering the characteristics of thetudy area (e.g., Simpson, 2002).

The observed close resemblance between both results
s indicative of coincidental parallel seismic and geoelec-ric anisotropies in this region of the Brazilian shield,imilar to what is seen in Australia and Canada (Simpson,001; Eaton et al., 2004). The spatial coherence between
(grey) responses. Right panel shows the linearized inverse layered-ximate depth transformation of the experimental data (Niblett and

both anisotropies and surface structural features at thewestern boundary of the Sao Francisco craton sup-ports also a model of coherent lithospheric deforma-tion derived from the last significant tectonic episoderather than related to the mechanisms of absolute platemotion. Similar results observed elsewhere have beeninterpreted as evidence that both crust and upper man-tle have remained undeformed since the last lithosphericdeformation episode, despite plate motion (Silver, 1996).

Simple asthenospheric flow is therefore rejected asthe source of mantle anisotropy around this region of theBrazilian shield. Absolute motion of the South Ameri-can plate has a well-defined trend, approximately E–Wfor most of the continent (see Figs. 7–10). Observed fastseismic azimuths and geoelectric strikes around the SaoFrancisco craton show no correlation with this directionbut display a striking correlation with the surface geol-ogy. Towards the centre of the Parana basin, SKS fastpolarization directions are nearly parallel to APM. How-ever, towards the western portion of the basin and overthe Rio Apa block (not shown in our figures) the seis-mic anisotropy is again aligned with the N–S directionof crustal blocks bordering the western side of the basin(James and Assumpcao, 1996). It could be argued that
the asthenospheric flow would be channelled around thetopography at the base of the lithosphere, as suggested tooccur in central Europe (Bormann et al., 1996). However,modelling of asthenospheric flow at the southern termi-

h and P
nation of the craton failed to explain the sharp curvatureof the anisotropy pattern in the Ribeira and Brasılia belts(Heintz et al., 2003).

The geoelectric information can be used to solve theproblem of lack of vertical resolution of the seismic data.Our results suggest that the NW anisotropy around thesouthern and southwestern border of the Sao Franciscocraton extends beneath the continental plate, possibly atleast to the 410 km olivine–spinel transition. Very deeplithosphere or coupling between lithospheric and sub-lithospheric mantle since the last major tectono-thermalevent is implied from the apparent absence of sublitho-spheric lateral mantle flow or deformation due to thepresent-day westward motion of the South Americanplate.

This interpretation is consistent with an independentevidence for a mechanically coherent translation of theSouth American plate and its upper mantle over thelast 130 Ma to explain an upper mantle low velocityanomaly beneath Brazil revealed by seismic tomography(VanDecar et al., 1995; Schimmel et al., 2003). Also,Heintz et al. (2003) in their study at the Ribeira beltfound two stations with very high delay times requir-ing either an intrinsic anisotropy of mantle rocks muchlarger than commonly described in petrophysical studiesor an anisotropic layer thicker than the lithosphere. Onthe other hand, it must be considered that temperaturesin the 250–410 km depth range exceed 1000 ◦C, and areprobably too high to allow sustained frozen anisotropyin a mechanically coherent lithosphere/asthenosphere lidon geologically relevant time scales (Vinnik et al., 1992).A possible explanation for the anisotropy to continuedown to these depths could be a variation in grain sizewith depth, allowing to an alignment of mineral grainsthrough dislocation creep (Ji et al., 1994). Furthermore,the notion that the upper mantle has remained mechani-cally unaltered since the last tectono-thermal event downto the transition zone places strong constraints on pro-posals of plate motion driven by a plate basal drag.For example, Russo and Silver (1996) proposed for theSouth American plate a mechanism to move this non-subducting plate through basal tractions arising fromdeep convective flow of the mantle because the stressesrequired to form and maintain the Andes chain are toohigh to be accounted only by ridge-push.

Another point to consider is the role of the Creta-ceous intraplate magmatism. The NNW trend of thedeep geoelectric strike in the northernmost MT profile
is suggested to be associated with rift-related volcanics(Bologna et al., 2006). Recent studies have shown thatrifting does not necessarily erase the inherited seis-mic anisotropy of a previously deformed lithosphere

(Vauchez and Garrido, 2001). Instead of that, rift ori-entation seems to be controlled by the preexisting litho-spheric mantle fabric revealed by deep geophysical data(Tommasi and Vauchez, 2001). Geoelectric anisotropyrelated to this rift structure could be then related to pre-existing mantle fabric, possibly reactivated by the nearbycollisional orogen but not necessarily correlated with itsstress perturbations to the lithosphere because of the sig-nificant time difference between both events.

Deep structures beneath the southern part of theSao Francisco craton and surrounding fold belts areextremely complex. As a consequence, information com-ing from different geophysical methods is sometimesdifficult to reconcile. For example, surface and bodywave tomography have shown that this part of the cra-ton is characterized by high velocities down to at least200 km, indicating the presence of a thick and unde-formed lithosphere beneath this region (Schimmel etal., 2003). On the other hand, a possible explanationfor a circular (600–800 km in diameter), positive, 8 m-amplitude geoid anomaly, detected in the same region(Molina and Ussami, 1999) where MT data along profilesPIU and SJR points to a circular-shaped high conductiv-ity anomaly at upper mantle depths beneath the Archeanrocks of the craton (Padua, 2004), requires a regionalthermal anomaly of a few tens of degrees at upper man-tle depths to compensate for a modelled thinner crust.Such results imply that the base of the lithosphere couldhave been eroded through reheating.

Two-dimensional inversions of the MT data along thefour profiles have shown that the effects of the Neo-proterozoic Brasiliano orogeny seems to be restrictedto the brittle and intermediate crust, while the under-lying lithosphere remained unaffected. Moreover, theborders of the blocks involved in the continental col-lision that generated the marginal belts are not detectedby the geophysical studies. Excepting anomalous areasclearly affected by Mesozoic thermal events, the litho-sphere beneath these belts presents high resistivities upto 200 km (Padua, 2004; Bologna et al., 2006). Theseresults are at variance with tomographic seismic modelsthat suggest a lithosphere thickness of around 100 kmbeneath the belts surrounding the craton (Schimmel etal., 2003).

Rotation of geoelectric strikes at depths sensing litho-spheric and sublithospheric mantle over the Ribeira beltbrings another piece to these puzzling results. The senseof rotation, with an E–W to WSW azimuth at the deep
lithosphere and a NW azimuth below it, is exactly theopposite of what should be expected in the most com-monly used models of two-layer anisotropy. These mod-els would consider a lower layer corresponding to the

h and P

AaBlt(

caPamcarcstvtiitat

8

aatttcdaitthrctEramgdtm


PM direction in the sublithospheric mantle (N253◦E)nd an upper layer related to the orogenic direction of therasılia belt (roughly NW). However, numerical simu-

ations carried out with this model could not reproducehe observed seismic variations at stations of this regionHeintz et al., 2003).

Complexity of this region is certainly related to theollision pattern of the Brasiliano orogens. They includeroughly NW zone of collision of the Sao Francisco andarana blocks, in the period between 630 and 600 Ma,nd the roughly ENE zone of collision of a near-coasticroplate with the southeastern border of the Sao Fran-

isco paleocontinent, near simultaneously between 605nd 550 Ma (Heilbron et al., 2000). Frozen anisotropy iselated to the last tectonic event but in this case it must beonsidered that minerals cannot reorient instantaneously,o that the preferred orientation is a complicated func-ion of the strain history, depending on how the responsearies with time and with the amount of strain, in addi-ion to other factors such as temperature, strain rate, andnitial conditions (Wenk and Christie, 1991). Intricatenteractions between both collisions, in the zone wherehey join in southern border of the craton, have gener-ted a complicated type of deformation that is difficulto solve with the available data.

. Summary and conclusions

Complementary anisotropy measurements of seismicnd magnetotelluric (MT) techniques carried out sep-rately in central-southeastern Brazil help to constrainhe manner anisotropy varies with depth, a key informa-ion concerning mantle motions and ancient continen-al fabrics. Forward modelling of a simplified regionalonductance map supports predictive interpretation ofeep electrical anisotropy in this region. MT responsest long periods indicate that the conductive structures only weakly anisotropic at upper mantle depths sohat alignment of olivine with the influence of addi-ional mechanisms at localized regions (e.g., diffusion ofydrogen) is a ready explanation of most of the observedesults. The southwestern border of the Sao Franciscoraton presents resistive lower crust and upper man-le, thus allowing the deep penetration of EM signals.lectric strikes derived at different depths along this

egion match the fast polarization direction of S-wavesnd show that lithospheric and sublithospheric defor-ation is vertically coherent with the surficial tectonic

rain, in a general NW direction. There is a significantiscrepancy between this direction and that related tohe present-day absolute plate motion, an indication that

antle deformation is not dominated by strains associ-


ated with the basal drag linked to the westward motionof the South American plate. The presence of a fossilanisotropy since the last tectono-thermal event is inter-preted as related to a very deep lithosphere or a cou-pling between lithospheric and sublithospheric mantlebeneath this region. Thus, the strain-induced crystallo-graphic fabric formed in response to large-scale tectonicprocesses in the ancient past probably extends to verylarge depths. The long term persistence of this deep fossilanisotropy has serious implications to mantle rheologyand the dynamics of plate motions.

Different electric azimuths from the general NWtrend are observed in the northern and southern part ofthe study area. In the north, the orientation of electricstrikes seems to be aligned with rift-related magma-tism, whereas in the south, a complex deep structureis observed, probably related to nearly simultaneousoblique collisions. More detailed geophysical studies atboth areas and also covering a wider geographic regionwill be necessary to substantiate the present conclusions.

Acknowledgements

This study was supported by research grants andfellowships from FAPESP (95/0687-4, 00/00806-5,01/02848-0 and 03/10817-2) and CNPq (142617/97-0,350683/94-8 and 351398/94-5). Relevant logistical sup-port from SOPEMI S.A. (De Beers Brasil Ltd.) is alsoacknowledged. The authors are grateful to Sergio Fontes,for loaning the long-period MT systems of ON/MCT forthe first fieldworks, Francois Chamalaun, for loaning thefluxgate magnetometers of Flinders University for theGDS array, and the dedicated field and lab crew. Theoriginal version was improved by careful reviews fromtwo anonymous reviewers. Special thanks to Alan Jonesand Dave Eaton, guest editors of this issue, for encour-agement and useful suggestions.

References

Alkmim, F.F., Marshak, S., 1998. Transamazonian orogeny in thesouthern Sao Francisco craton region, Minas Gerais, Brazil:evidence for Paleoproterozoic collision and collapse in theQuadrilatero Ferrifero. Precamb. Res. 90, 29–58.

Alkmim, F.F., Marshak, S., Fonseca, M.A., 2001. Assembling WestGondwana in the Neoproterozoic: clues from the Sao Franciscocraton region, Brazil. Geology 29, 319–322.

Almeida, F.F.M., Hasui, Y., 1984. O Pre-Cambriano do Brasil. EdgardBlucher Ltd., Sao Paulo, 378 pp.

Almeida, F.F.M., Hasui, Y., Brito Neves, B.B., Fuck, R.A., 1981.Brazilian structural provinces: an introduction. Earth Sci. Rev. 17,1–29.

Bahr, K., 1991. Geological noise in magnetotelluric data: a classifica-tion of distortion types. Phys. Earth. Planet. Interiors 66, 24–38.

h and P
Bahr, K., Olsen, N., Shankland, T.J., 1993. On the combination ofthe magnetotelluric and the geomagnetic depth sounding methodfor resolving an electrical conductivity increase at 400 km depth.Geophys. Res. Lett. 20, 2937–2940.

Bahr, K., Bantin, M., Jantos, C., Schneider, E., Storz, W., 2000. Elec-trical anisotropy from electromagnetic array data: implications forthe conduction mechanism and for distortion at long periods. Phys.Earth Planet. Interiors 119, 237–257.

Berdichevsky, M.N., 1999. Marginal notes on magnetotellurics. Surv.Geophys. 20, 341–375.

Bologna, M.S., Padilha, A.L., Vitorello, I., 2005. Geoelectric crustalstructures off the SW border of the Sao Francisco craton, centralBrazil, as inferred from a magnetotelluric survey. Geophys. J. Int.162, 357–370.

Bologna, M.S., Padilha, A.L., Vitorello, I., Fontes, S.L., 2006. Tec-tonic insight into a pericratonic subcrustal lithosphere affectedby anorogenic Cretaceous magmatism in central Brazil inferredfrom long-period Magnetotellurics. Earth Planet. Sci. Lett. 241,603–616.

Bormann, P., Gruenthal, G., Kind, R., Montag, H., 1996. Upper man-tle anisotropy beneath Central Europe from SKS wave splitting:effects of absolute plate motion and lithosphere–asthenosphereboundary topography? J. Geodyn. 22, 11–32.

Chamalaun, F.H., Walker, R., 1982. A micro-processor-based digitalfluxgate magnetometer for geomagnetic deep sounding studies. J.Geomagn. Geoelectr. 34, 491–507.

Debayle, E., Kennett, B.L.N., 2000. The Australian continental upper-mantle: structure and deformation inferred from surface waves. J.Geophys. Res. 105, 25, 423–25, 450.

Eaton, D.W., Jones, A.G., Ferguson, I.J., 2004. Lithospheric anisotropystructure inferred from collocated teleseismic and magnetotelluricobservations: Great Slave Lake shear zone, northern Canada. Geo-phys. Res. Lett. 31, L19614, doi:10.1029/2004GL020939.

Egbert, G.D., 1997. Robust multiple-station magnetotelluric data pro-cessing. Geophys. J. Int. 130, 475–496.

Fuck, R.A., Jardim de Sa, E.F., Pimentel, M.M., Dardenne, M.A.,Soares, A.C.P., 1994. As faixas de dobramentos marginaisdo Craton do Sao Francisco: sıntese dos conhecimentos. In:Dominguez, J.M.L., Misi, A. (Eds.), O Craton do Sao Francisco.SBG-SGM-CNPq, Salvador, pp. 161–185.

Gaherty, J.B., Jordan, T.H., 1995. Lehmann discontinuity as the baseof an anisotropic layer beneath continents. Science 268, 1468–1471.

Gamble, T.D., Goubau, W.M., Clarke, J., 1979. Magnetotellurics witha remote magnetic reference. Geophysics 44, 53–68.

Gripp, A.E., Gordon, R.G., 2002. Young tracks of hotspots and currentplate velocities. Geophys. J. Int. 150, 321–361.

Groom, R.W., Bailey, R.C., 1989. Decomposition of magnetotelluricimpedance tensor in the presence of local three-dimensional gal-vanic distortion. J. Geophys. Res. 94, 1913–1925.

Groom, R.W., Kurtz, R.D., Jones, A.G., Boerner, D.E., 1993. A quanti-tative methodology to extract regional magnetotelluric impedancesand determine the dimension of the conductivity structure. Geo-phys. J. Int. 115, 1095–1118.

Heilbron, M., Mohriak, W.V., Valeriano, C.M., Milani, E.J., Almeida,J., Tupinamba, M., 2000. From collision to extension: the roots ofthe southeastern continental margin of Brazil. In: Mohriak, W.V.,
Talwani, M. (Eds.), Geology and Geophysics of Continental Mar-gin, vol. 115. AGU Geophysical Monograph, pp. 1–31.
Heinson, G., White, A., 2005. Electrical resistivity of the NorthernAustralian lithosphere: crustal anisotropy or mantle heterogeneity?Earth Planet. Sci. Lett. 232, 157–170.


Heintz, M., Vauchez, A., Assumpcao, M., Barruol, G., Egydio-Silva,M., 2003. Shear wave splitting in SE Brazil: an effect of activeor fossil upper mantle flow, or both? Earth Planet. Sci. Lett. 211,79–95.

James, D., Assumpcao, M., 1996. Tectonic implications of S-waveanisotropy beneath SE Brazil. Geophys. J. Int. 126, 1–10.

Ji, S., Zhao, X., Francis, D., 1994. Calibration of shear-wave split-ting in the subcontinental upper mantle beneath active orogenicbelts using ultramafic xenoliths from the Canadian Cordillera andAlaska. Tectonophysics 239, 1–27.

Ji, S., Rondenay, S., Mareschal, M., Senechal, G., 1996. Obliquitybetween seismic and electrical anisotropies as a potential indicatorof movement sense for ductile shear zones in the upper mantle.Geology 24, 1033–1036.

Jones, A.G., 1999. Imaging the continental upper mantle using elec-tromagnetic methods. Lithos 48, 57–80.

Jones, A.G., 2006. Electromagnetic interrogation of the anisotropicEarth: looking into the Earth with polarized spectacles. Phys. EarthPlanet. Interiors, this issue.

Jones, A.G., Groom, R.W., 1993. Strike angle determination from themagnetotelluric tensor in the presence of noise and local distortion:rotate at your peril! Geophys. J. Int. 113, 524–534.

Korja, T., BEAR Working Group, 2003. Upper mantle conductivityin Fennoscandia as imaged by the BEAR array. In: Joint Euro-pean Geosciences Union—American Geophysical Union Meeting,Nice, Abstracts, EGU, 2 pp.

Lesquer, A., Almeida, F.F.M., Davino, A., Lachaud, J.C., Maillard, P.,1981. Signification structurale des anomalies gravimetriques de lapartie sud du craton de Sao Francisco (Bresil). Tectonophysics 76,273–293.

Machado, N., Schrank, A., Noce, C.M., Gauthier, G., 1996. Ages ofdetrital zircon from Archean-Paleoproterozoic sequences, implica-tions for greenstone belt setting and evolution of a Transamazonianforeland basin in Quadrilatero Ferrıfero, southeast Brazil. EarthPlanet. Sci. Lett. 141, 259–276.

Mackie, R.L., Smith, J.T., Madden, T.R., 1994. Three-dimensionalelectromagnetic modeling using finite difference equations: themagnetotelluric example. Radio Sci. 29, 923–935.

Mareschal, M., Kellet, R.L., Kurtz, R.D., Ludden, J.N., Ji, S., Bailey,R.C., 1995. Archean cratonic roots, mantle shear zones and deepelectrical anisotropy. Nature 375, 134–137.

McNeice, G., Jones, A.G., 2001. Multisite, multifrequency tensordecomposition of magnetotelluric data. Geophysics 66, 158–173.

Meju, M.A., 1996. Joint inversion of TEM and distorted MT soundings:some effective practical considerations. Geophysics 61, 56–65.

Molina, E.C., Ussami, N., 1999. The geoid in southern Brazil and adja-cent regions: new constraints on density distribution and thermalstate of the lithosphere. J. Geodyn. 28, 357–374.

Niblett, E.R., Sayn-Wittgenstein, C., 1960. Variation of the electri-cal conductivity with depth by the magnetotelluric method. Geo-physics 25, 998–1008.

Padilha, A.L., Vitorello, I., Brito, P.M.A., 2002. Magnetotelluricsoundings across the Taubate Basin, Southeast Brazil. Earth Plan-ets Space 54, 617–627.

Padua, M.B., 2004. Estudos de inducao eletromagnetica nacaracterizacao de estruturas profundas sob a borda sul do cratonde Sao Francisco. Ph.D. Thesis. Instituto Nacional de Pesquisas
Espaciais.
Pimentel, M.M., Dardenne, M.A., Fuck, R.A., Viana, M.G., Junges,S.L., Fischel, D.P., Seer, H.J., Dantas, E.L., 2001. Nd isotopes andthe provenance of detrital sediments of the Neoproterozoic BrasıliaBelt, central Brazil. J. S. Am. Earth Sci. 14, 571–585.

http://dx.doi.org/10.1029/2004GL020939

h and P

R

S

S

S

S

S

S

S

S

S

S

S

S


usso, R.M., Silver, P.G., 1996. Cordillera formation, mantle dynam-ics, and the Wilson cycle. Geology 24, 511–514.

avage, M.K., 1999. Seismic anisotropy and mantle deformation: whathave we learned from shear wave splitting? Rev. Geophys. 37,65–106.

chimmel, M., Assumpcao, M., VanDecar, J., 2003. Upper man-tle seismic velocity structure beneath SE Brazil from P- andS-wave travel time inversions. J. Geophys. Res. 108, 2191,doi:10.1029/2001JB000187.

chmitt, R.S., Trouw, R.A.J., Van Schmus, W.R., Pimentel, M.M.,2004. Late amalgamation in the central part of West Gondwana:new geochronological data and the characterization of a Cambriancollisional orogeny in the Ribeira Belt (SE Brazil). Precamb. Res.133, 29–61.

chmucker, U., Jankowski, J., 1972. Geomagnetic induction studiesand the electrical state of the upper mantle. Tectonophysics 13,233–256.

chobbenhaus, C., Campos, D.A., Derze, G.R., Asmus, H.E., 1984.Mapa geologico do Brasil e da area oceanica adjacente incluindodepositos minerais, escala 1:2,500,000. DNPM, Brasılia.

enechal, G., Rondenay, S., Mareschal, M., Guilbert, J., Poupinet, G.,1996. Seismic and electrical anisotropies in the lithosphere acrossthe Grenville Front, Canada. Geophys. Res. Lett. 23, 2255–2258.

ilver, P.G., 1996. Seismic anisotropy beneath the continents: probingthe depths of geology. Annu. Rev. Earth Planet. Sci. 24, 385–432.

ilver, P.G., Chan, W.W., 1988. Implications for continental structureand evolution from seismic anisotropy. Nature 335, 34–39.

impson, F., 2001. Resistance to mantle flow inferred from the elec-tromagnetic strike of the Australian upper mantle. Nature 412,632–634.

impson, F., 2002. Intensity and direction of lattice-preferred orienta-tion of olivine: are electrical and seismic anisotropies of the Aus-tralian mantle reconcilable? Earth Planet. Sci. Lett. 203, 535–547.

impson, F., Bahr, K., 2005. Practical Magnetotellurics. CambridgeUniversity Press, Cambridge, 254 pp.

ubba Rao, P.B.V., Padua, M.B., Bologna, M.S., Vitorello, I., Padilha,A.L., Chamalaun, F.H., Rigoti, A., 2003. Preliminary results ofongoing GDS survey in Center-Southeast Brazil. In: Proceedingsof the Eighth International Congress of the SBGf, Rio de Janeiro,Extended Abstracts, SBGf, pp. 1–4.


Swift, C.M., 1967. A magnetotelluric investigation of an electricalconductivity anomaly in the South Western United States. Ph.D.Thesis. Massachusetts Institute of Technology.

Thompson, R.N., Gibson, S.A., Mitchell, J.G., Dickin, A.P., Leonar-dos, O.H., Brod, J.A., Greenwood, J.C., 1998. MigratingCretaceous-Eocene magmatism in the Serrado Mar alkalineprovince SE Brazil: Melts from the deflected Trindade mantleplume? J. Petrol. 39, 1493–1526.

Tommasi, A., Vauchez, A., 2001. Continental rifting parallel to ancientcollisional belts: an effect of the mechanical anisotropy of the litho-spheric mantle. Earth Planet. Sci. Lett. 185, 199–210.

Trompette, R., 1997. Neoproterozoic (∼600 Ma) aggregation of West-ern Gondwana: a tentative scenario. Precamb. Res. 82, 101–112.

Trouw, R.A.J., Heilbron, M., Ribeiro, A., Paciullo, F.V.P., Valeriano,C.M., Almeida, J.C.H., Tupinamba, M., Andreis, R.R., 2000. Thecentral segment of the Ribeira Belt. In: Cordani, U.G., Milani,E.J., Thomaz Filho, A., Campos, D.A. (Eds.), Tectonic Evolutionof South America. DNPM, Rio de Janeiro, pp. 287–310.

VanDecar, J.C., James, D.E., Assumpcao, M., 1995. Seismic evidencefor a fossil mantle plume beneath South America and implicationsfor plate driving forces. Nature 378, 25–31.

Vanyan, L.L., Berdichevsky, M.N., Vasin, N.D., Okulessky, B.A.,1980. On the normal geoelectric profile. Izv. Akad. Nauk SSSRSer. Fizika Zemli 2, 73–76.

Vauchez, A., Garrido, C.J., 2001. Seismic properties of an astheno-spherized lithospheric mantle: constraints from lattice preferredorientations in peridotite from the Ronda massif. Earth Planet. Sci.Lett. 192, 235–249.

Vinnik, L.P., Makeyeva, L.I., Milev, A., Usenko, A.Yu., 1992. Globalpatterns of azimuthal anisotropy and deformations in the continen-tal mantle. Geophys. J. Int. 111, 433–447.

Wenk, H.R., Christie, J.M., 1991. Comments on the interpretation ofdeformation textures in rocks. J. Struct. Geol. 13, 1091–1110.

Wu, N., Booker, J.R., Smith, J.T., 1993. Rapid two-dimensional inver-sion of COPROD2 data. J. Geomagn. Geoelectr. 45, 1073–1087.

Zalan, P.V., Wollf, S., Astolfi, M.A.M., Vieira, I.S., Conceicao, J.C.J.,Appi, V.T., Neto, E.V.S., Cerqueira, J.R., Marques, A., 1990. TheParana Basin, Brazil. In: Leighton, M.W., Kolata, D.R., Oltz, D.F.,Eidel, J.J. (Eds.), Interior Cratonic Basins, vol. 51. AAPG Memoir,Tulsa, pp. 681–708.

http://dx.doi.org/10.1029/2001JB000187

lithospheric and sublithospheric anisotropy beneath

Documents