The processes responsible for the formation of thick, strong and cold Archean sub-continental lithospheric mantle (mantle keels) beneath Archean cratons remain elusive. Here, the dynamics of some such processes are studied by forward numerical modeling of the thermo-mechanical evolution of continental lithosphere undergoing collision and orogenesis under Neoarchean-like conditions. The numerical experiments illustrate that depending on the composition of the crust and the degree of radioactive heat production (RHP) in the crust, three dominant modes of mantle lithosphere deformation evolve: (1) pure-shear thickening; (2) imbrication; (3) and a mode best described as underplating. All three modes of deformation result in the thickening and emplacement of plate-like mantle lithosphere to depths between 200 km and 350 km. The transition from pure-shear thickening to imbrication is largely dependent on the degree of RHP in the crust, while the transition from the imbrication style to the underplating style is dependent on the composition of the lower crust.
 Archean sub-continental lithospheric mantle (SCLM) is composed of highly refractory peridotitic mantle that exists as a keel of thickened lithosphere beneath most Archean cratons [Carlson et al., 2005]. Models of SCLM formation are poorly resolved, but generally fall into two categories (reviewed by Lee ). The “plume model” suggests that partial melting in a plume head is the mechanism responsible for the formation of Archean SCLM. However, this hypothesis is problematic because the protoliths of cratonic peridotites underwent partial melting at pressures less than their current equilibration pressures [Lee, 2006]. In the second model, imbrication of buoyant oceanic lithosphere during lateral tectonic accretion is responsible for forming SCLM [e.g., Helmstaedt and Schulze, 1989]. However, such a lateral tectonic accretion model implies a greater amount of eclogite in SCLM than is implied by mantle-derived xenoliths [Schulze, 1989]. A third, though less popular, model envisages SCLM as a result of continental collision [e.g., Jordan, 1978]. However, previous studies on Archean continental collision have focused largely on increased continental geotherm induced lateral gravitational-driven flow of lithosphere [e.g., Rey and Houseman, 2006] and the dynamic interaction between the mantle and lithosphere [Cooper et al., 2006].
 In this study, we investigate the nature of continental lithosphere during collision under Neoarchean-like conditions as a potential process for creating thick SCLM, specifically considering the dynamical interaction between the continental crust and mantle. We conduct a series of computational geodynamic experiments that test the sensitivity of continental mantle lithosphere deformation to a buoyant mantle lithosphere; varying crustal compositions, reflective of crustal composition variability in Archean cratons, and degrees of radiogenic heat production (RHP) in the crust. As we demonstrate, these tectonics are significantly modified from present-day plate behavior owing to different thermal/compositional conditions during this time. We do not preclude the possibility of other processes (e.g., plume-related), but rather focus on horizontal tectonics as the most widely cited.
2. Modeling Neoarchean Continental Collision
 In the numerical experiments we used SOPALE, a plane-strain visco-plastic finite-element code [Fullsack, 1995]. The experiments model an idealized collision between continental Archean plates (Figure 1). Several aspects of the configuration tailor the model to Neoarchean conditions: (1) the initial geotherm in the experiments is significantly elevated from present-day estimates; (2) the internal heat production in the crust is based on the interpreted thermal state of Archean cratons ∼2.6 Ga [e.g., Pollack, 1997]; (3) the crustal configurations are chosen based on variability that exists in Archean cratons [Arndt, 1999; Musacchio et al., 2004]; and (4) the densities of the mantle lithosphere and the sub-lithospheric mantle are based on mantle xenolith/xenocryst studies that have placed constraints on the secular variation in the density of the SCLM [Poudjom Djomani et al., 2001]. The more depleted mantle lithosphere in the Archean and assumed paleo-geotherms lead to more buoyant cratonic mantle lithosphere than the underlying mantle. For the experiments, we vary the first three factors to consider how they control the behavior of the convergent mantle lithosphere.
3. Experimental Results
Figure 2a shows results of Run1 after 111 Myr of imposed convergence, corresponding to 3330 km (i.e., 70%) of lithospheric shortening. Precambrian accretionary orogens typically underwent ∼50% (∼80 Myr) shortening (reviewed by Chardon et al. ). These experiments are not intended to recreate mantle lithosphere structures beneath Archean cratons, but to investigate the nature of Neoarchean continental lithosphere during collision; the results presented in Figures 2– 4 are still valid when 50% shortening has occurred. At this stage, five shear zones have developed in the mantle lithosphere (labeled sequentially in the order of their development). Four of the shear zones (SZ1,3–5) dip towards the right margin of the box, while SZ2 dips towards the left. The buoyancy of the mantle lithosphere compared to the underlying sub-lithospheric mantle, is preventing it from subducting into the sub-lithospheric mantle as shortening progresses and allowing it to thicken significantly. Rather, imposed convergence is causing the mantle lithosphere to imbricate. Plate-like mantle lithosphere (log(ηeff.) ≥ 23) is present at depths of ∼200–390 km below all shear zones. Some of the shortening has been accommodated in the mantle lithosphere by pure-shear thickening (e.g., to the left of SZ4) and folding (e.g., to the right of SZ1), but this is minor compared to the imbrication. Imposed convergence is dominantly accommodated in the crust by pure shear-thickening. The disjointed grid lines in the sub-lithospheric mantle are the result of extensive mantle convection-driven deformation of the Lagrangian grid.
 Except for the composition of its crust, the set-up of experiment Run2 is identical to Run1. In Run2, a diabase rheology [Ranalli, 1995] (to simulate wet mafic volcanics) is used for the upper-crust and a felsic granulite rheology [Ranalli, 1995] is used for the lower crust. The motivation for this crustal set-up is to replicate Neoarchean crust prior to a possible crustal inversion event. Figure 2b shows the results of the numerical experiment after 111 Myr of imposed convergence. Run2 shows a similar imbricating mantle lithosphere to Run1 – even the relative timing of the shear zones is the same – but the vergence of the structures is more uniform in Run2. However, the vergence of the shear zones (e.g., the opposite polarity of SZ2 in Run1) may be a rather arbitrary event in the model where small-scale perturbations lead to high strain in the strain-softening materials. The wet quartzite upper-crust in Run1 is weaker than the wet diabase upper-crust in Run2, and as a result, imposed convergence in Run2 is accommodated in the upper crust by pure-shear thickening as well as folding. The difference suggests that the crust in Run1 is at least partially coupled to the mantle lithosphere whereas there is a more complete decoupling in Run 2, as evidenced by the highly deformed Lagrangian grid. Below SZ1-5, plate-like mantle lithosphere is present at depths of ∼200 to ∼390 km.
 Experiment Run3 is identical to Run1, except now a granite rheology [Ranalli, 1995] is used for the upper crust and a mafic granulite rheology [Ranalli, 1995] is used for the lower crust. The motivation for this crustal set-up is to replicate Neoarchean crust after a possible crustal inversion event. After 111 Myr of imposed convergence, pro-side (Figure 1) mantle lithosphere has slid beneath retro-side mantle lithosphere along SZ1 for a distance of ∼950 km (Figure 2c). This style of deformation is best described as “underplating”. This behavior occurs because the mafic granulite lower crust is stronger than either a wet quartzite or a felsic granulite lower crust. Consequently, there is a greater degree of coupling between the lower crust and mantle lithosphere, resulting in a stronger lithosphere. As a consequence of this enhanced coupling, localized thrusting and exhumation of pro-side mafic granulite lower crust onto adjacent retro-side mafic granulite lower crust has occurred above SZ1. Plate-like mantle lithosphere is present at depths of ∼350 km.
 Experiment Run4 (Figure 3a) is identical to the reference model Run1, but RHP in the crust is increased to 6.76 × 10−10 W/kg. Initially, imposed convergence is accommodated in the mantle lithosphere by the development of a shear zone (SZ1). Increased RHP in the crust results in increased lower crustal temperatures, which in turn warms the mantle lithosphere. Higher mantle lithosphere temperatures lower its effective viscosity. Consequently, after the development of SZ1, imposed convergence is no longer accommodated in the mantle lithosphere by the development of shear zones but by pure-shear thickening and minor folding. Plate-like mantle lithosphere extends to depths of ∼200–300 km. As in Run1, the overlying crust accommodates imposed convergence by undergoing pure-shear thickening.
 Experiment Run5 (Figure 3b) is identical to Run2, but the RHP in the felsic granulite lower crust is increased to 6.76 × 10−10 W/kg. The main difference between the models is that in Run5 the shear zones do not develop as readily as they did in Run2. For example, there are fewer shear zones in Run5 and these are not as well developed as in Run2 (viz., SZ4). This is owing to the increased RHP in the lower crust that causes higher mantle lithosphere temperatures and a shift to more ductile deformation, mostly in the form of pure-shear thickening combined with the imbrication behavior. As in Run2, plate-like mantle lithosphere extends to depths between ∼200 and 350 km. Similarly, the crust is strongly decoupled from the mantle lithosphere and undergoes folding and pure-shear thickening.
 The setup of experiment Run6 (Figure 3c) is identical to Run3, but the RHP in the granite upper crust is increased to 6.76 × 10−10 W/kg. The most notable difference between Run3 and Run6 is the presence of a second shear zone (SZ2) that has begun to form, of opposite polarity from SZ1. Below SZ1-2, plate-like mantle lithosphere extends to depths greater than ∼400 km. Again, the crust is well coupled to the mantle lithosphere; leading to localized deformation of the crust at the mantle shear zones and exhumation of the lower crustal layer to the surface.
 The results shown represent a small subset of a series of numerical experiments. When the log of effective viscosity of the lower crust (calculated near the retro- margin (Figure 1) of the box) is plotted versus the degree of RHP in the crust for each of the experiments, the modes of mantle lithosphere behavior group in separate domains (Figure 4). The pure-shear thickening style occurs when the degree of RHP in the crust is sufficiently elevated to weaken the lower crust. This style transitions to the imbrication style as the degree of RHP is lowered, allowing the lower crust to strengthen and more effectively couple to the mantle lithosphere than in the pure-shear thickening style. This transition is in contrast to the one that marks the shift from the imbrication style to the underplating style. This transition is less dependent on the degree of RHP in the crust than on the composition of the lower crust. Transitioning from a felsic granulite to a mafic granulite lower crust (possibly accomplished by vertical crustal tectonics) results in a stronger lower crust. This leads to a greater degree of coupling between the lower crust and the mantle lithosphere, in turn increasing the strength of the mantle lithosphere and lithosphere.
4. Conclusions and Discussion
 We tested how the behavior of mantle lithosphere changed with modification only to the composition of the lower crust and the degree of RHP in the crust. Though we recognize that other factors may be important (e.g., varying chemical buoyancy) we consider various first-order interpretations of the state of the Archean lithosphere (i.e., the degree of RHP in the crust and the configuration of the crust). The numerical experiments identify three dominant styles of model Neoarchean-like mantle lithosphere deformation during collision: (1) dominantly pure-shear thickening (Figure 3a); (2) imbrication (Figures 2a, 2b, and 3b); (3) and underplating with localized exhumation of the crust (Figures 2c and 3c).
 The pure-shear-thickening style of deformation occurs when the temperature of the lower crust is sufficiently raised by RHP. This leads to a decrease in the degree of coupling between the lower crust and the mantle lithosphere, in turn enabling a dominantly ductile distribution of deformation in the mantle lithosphere. The imbrication style of deformation occurs in a mantle lithosphere that is overlain by a rheologically weak lower crust (e.g., wet quartzite or felsic granulite) while the degree of RHP in the crust is sufficiently low. This allows the strong upper portion of the mantle lithosphere to progressively underthrust adjacent mantle lithosphere along a weak/decoupling crust-mantle interface. Imposed convergence is accommodated in the mantle lithosphere by underplating when the lower crust is rheologically stronger, e.g., mafic granulite. The stronger lower crust does not decouple from the underlying mantle lithosphere to the extent that it would if it were more felsic. As a result, the lithosphere deforms as a single plate, more akin to oceanic lithosphere. These three styles of deformation result in the emplacement of plate-like mantle lithosphere at depths of ∼200–400 km.
 Recent papers [e.g., Rey and Houseman, 2006; Duclaux et al., 2007] demonstrate that two-dimensional (2D) experiments cannot adequately study Archean continental collision because three dimensional (3D) warm and buoyant lithosphere will undergo orogen-parallel ductile flow during orogenesis. In some of the experiments the lower crust accommodates convergence by undergoing pure-shear shear thickening and if given the opportunity will likely flow somewhat in the out-of-plane direction. Furthermore, our experiments illustrate that continental collision may be dominantly accommodated in the mantle lithosphere along brittle shear zones. We have chosen to focus on the detailed crust-mantle evolution and consequently have chosen a high resolution 2D model while neglecting (recognized) 3D orogenic processes.
 RG acknowledges funding from an Ontario Graduate Scholarship in Science and Technology. The work was supported by a Natural Sciences and Engineering Research Council of Canada Discovery Grant to RNP. The numerical code, SOPALE, was originally developed by Philippe Fullsack. We appreciate careful and thoughtful reviews by two anonymous referees.