The formation of Laurentia : Evidence from shear wave splitting

Abstract The northern Hudson Bay region in Canada comprises several Archean cratonic nuclei, assembled by a number of Paleoproterozoic orogenies including the Trans-Hudson Orogen (THO) and the Rinkian–Nagssugtoqidian Orogen. Recent debate has focused on the extent to which these orogens have modern analogues such as the Himalayan–Karakoram–Tibet Orogen. Further, the structure of the lithospheric mantle beneath the Hudson Strait and southern Baffin Island is potentially indicative of Paleoproterozoic underthrusting of the Superior plate beneath the Churchill collage. Also in question is whether the Laurentian cratonic root is stratified, with a fast, depleted, Archean core underlain by a slower, younger, thermally-accreted layer. Plate-scale process that create structures such as these are expected to manifest as measurable fossil seismic anisotropic fabrics. We investigate these problems via shear wave splitting, and present the most comprehensive study to date of mantle seismic anisotropy in northern Laurentia. Strong evidence is presented for multiple layers of anisotropy beneath Archean zones, consistent with the episodic development model of stratified cratonic keels. We also show that southern Baffin Island is underlain by dipping anisotropic fabric, where underthrusting of the Superior plate beneath the Churchill has previously been interpreted. This provides direct evidence of subduction-related deformation at 1.8 Ga, implying that the THO developed with modern plate-tectonic style interactions.

tons, seismic anisotropy, Laurentia.  When a radially-polarised shear wave encounters seismically anisotropic media, it 56 will split into two orthogonal shear waves polarised along the fast and slow axes of 57 the material. The splitting parameters (the polarisation direction of the fast shear 58 wave) and t (the delay time between the two waves) can then be used to characterise 59 the crust and mantle anisotropy beneath the receiver (Silver and Chan, 1991).  Our study region contains large portions of 3 Archean provinces (Rae, Hearne, Su-75 perior: Figure 1) that comprise much of the Canadian Shield (Ho↵man, 1988). The   (Jackson and Berman, 2000). This generally north-south 96 striking fold belt has been linked to the east-west oriented plate-scale Nagssugtoqidian 97 Orogen (NO) of southern Greenland (e.g., Connelly et al., 2006). The combined 98 Rinkian-Nagssugtoqidian orogen is similarly asymmetric, and potentially similar in 99 scale, to the THO to the south. Seismic anisotropy refers to the directional dependence of seismic wavespeed. When 102 a shear wave encounters an anisotropic medium, it will split into two shear waves, or-103 thogonally polarised, one travelling faster than the other (e.g., Silver and Chan, 1991).

104
The polarisation direction ( ) of the fast shear wave, and the delay time ( t) between 105 them can be used to characterise the seismic anisotropy of the material. P-to-S con-106 verted phases such as SKS and SKKS, are ideally-suited for upper-mantle shear-wave 107 splitting studies; they are radially polarised at the core-mantle boundary and thus 108 record no source-side anisotropy (e.g. Silver and Chan, 1991;Long and Silver, 2009).

109
Olivine, the most common mineral in Earth's upper mantle, is highly anisotropic. A   analysis is utilised to find measurements that are stable over many di↵erent windows.
All splitting parameters were determined after analysis of 100 di↵erent windows. An 138 example of the analysis is shown in Figure 2.

140
The results obtained for SKS splitting in northern Canada displayed visually in Fig-141 ure 1 are summarised in Table 1; a full list of all splitting results and associated errors 142 is included in supplemental materials. The average signal to noise ratio (as defined 143 by Restivo and Hel↵rich, 1999) for these data is 15.5 with a standard deviation of 7.1.

144
Quality control was enforced by visual inspection of each split to ensure linearisation 145 of the particle motion. We also enforced data error upper limits of ±15 in and 146 ±0.5 s in t, although most were much lower (see supplemental materials for com-147 plete dataset). This is a much stricter limit than the previous study of Snyder et al.  (Table 1). Each basic class of model (single layer, two-layer, 151 dipping layer) has a distinctive backazimuthal pattern in their splitting parameters. model has peak-to-peak variations 90 , as opposed to interfaces or two-layer mod-156 els, which have sharper changes in and peak-to-peak variations approaching 180 .
The model-class for each station in Table 1 is chosen to be the simplest possible that 159 explains the observations. For many of the stations in Figure 1 a single, horizontal 160 anisotropic layer could adequately explain the data. Data from such stations were 161 stacked using a procedure based on the method of Wolfe and Silver (1998) to obtain 162 single pairs of splitting parameters (e.g., CTSN, DORN, MARN, in Figure 1; Table   163 1). We cannot, however, preclude the possibility that this assumption is invalid for 164 stations where backazimuthal coverage of earthquakes is insu cient to resolve more 165 complex dipping or multi-layer patterns of anisotropy (e.g., Silver and Savage, 1994).  Table 1 compares the number of splits used to define each station in our dataset 174 directly to that of Bastow et al. (2011), in brackets. Across our study area, we commonly observe t 1 s ( Figure 1, Table 1), implying 179 a mantle contribution to the observations: regional continental crust is ⇠40 km thick      asthenosphere boundary in this region is at ⇠240 km depth; a value much deeper 367 than the lower anisotropic layer used to produce the model in Figure 9. Further, the 368 similarity between the FCC and ARVN data in Figure 9 links the structure beneath these stations, thus precluding the Lynn Lake fault (local to station FCC) from being  and slow waveforms before correction, after correction, and after correction without 437 normalised amplitudes. Bottom L-R: particle motion before and after correction.