Abstract
This work represents a numerical study of the thermal activation for dislocation glide of the [100](010) slip system in MgSiO_{3} postperovskite (Mgppv) at 120 GPa. We propose an approach based on a onedimensional line tension model in conjunction with atomicscale calculations. In this model, the key parameters, namely, the line tension and the Peierls barrier, are obtained from density functional theory calculations. We find a Peierls stress σ_{p} = 2.1 GPa and a line tension Γ = 9.2 eV/Å, which lead to a kinkpair enthalpy (under zero stress) of 2.69 eV. These values confirm that this slip system bears a very low lattice friction because it vanishes for temperatures above approximately 500 K under mantle conditions. In the Earth’s mantle, highpressure Mgppv silicate is thus expected to become as ductile as ferropericlase. These results confirm the hypothesis of a weak layer in the D″ layer where Mgppv is present. Easy glide along [100](010) suggests strong preferred orientations with (010) planes aligned. Highly mobile [100] dislocations are also likely to respond to stresses related to seismic waves, leading to energy dissipation and strong attenuation.
Introduction
The D″ region, which lies just above the coremantle boundary (CMB), is of primary importance in geodynamics because it represents the thermal boundary layer, where heat is transferred by diffusion from the core to the convective mantle. Its viscosity plays a key role in controlling the amount of heat that can be extracted from the core, with significant implications regarding the dynamics of the mantle. The discovery of a phase transition from bridgmanite to a postperovskite (ppv) phase at pressure (P) and temperature (T) conditions close to those of the CMB has opened new perspectives regarding the interpretation of the D″ structure and dynamics. This phase, which exhibits a very peculiar layered structure, has attracted much attention. Based on firstprinciples calculations, Ammann et al.^{1} showed that the diffusion of Mg^{2+} and Si^{4+} is extremely anisotropic in postperovskite, with high diffusion rates along <100>. This finding has led to the idea that the D″ layer could be weaker than the overlying mantle due to high diffusion creep rates in ppv, although creep rates may be controlled by diffusion in the slowest direction^{2}. However, this hypothesis is difficult to reconcile with the strong seismic anisotropy that is the signature of the D″ layer (as first observed in ScS^{3} and S_{diff}^{4} phases) and is usually indicative of dislocation creep rather than diffusion creep.
Making the link between seismic anisotropy observations and putative crystal preferred orientations (CPO) in ppv is not straightforward. In horizontally propagating phases, the observation that horizontally polarized shear waves propagate faster than vertically polarized ones (V_{SH} > V_{SV}) agrees with models involving dominant slip in (010) or (001) in ppv^{5,6}. If only dominant slip on (100) seems to be excluded by seismological observations so far, the respective role of (010)^{6,7} and (001) slip^{8,9} remains debatable. From the mineral physics point of view, the hypothesis of slip in (010), which seems intuitive given the layered structure of postperovskite, is difficult to assess experimentally. Indeed, observation of experimentally produced CPO in diamond cell experiments has also led to conflicting results^{10,11}, which may result from textures formed during phase transformation^{12,13}. Additionally, beyond the geometry of crystal plasticity (from CPO), there is no way currently to obtain experimentally quantitative data of the rheology of silicate postperovskite under relevant P, T (and strain rate!) conditions of the D″.
Multiscale numerical modelling represents an alternative that is currently able to describe plasticity by dislocations of highpressure minerals. Recent applications to wadsleyite^{14}, ringwoodite^{15}, periclase^{16} and bridgmanite^{17} have shown their ability to reproduce laboratory experiment data. The main implication of these studies (both numerical and experimental) is that pressure has a strong effect on the lattice friction opposed to dislocation glide. In bridgmanite, Hirel et al.^{18} showed that lattice friction increases monotonically throughout the lower mantle to reach, in the lowermost mantle, values of approximately 15 GPa. In this context, the behaviour of postperovskite appears remarkable. Shearing the Si octahedral layers (for instance, with the [100](001) slip system) yields comparable values of lattice friction to those of bridgmanite^{19}. However, shearing the structure parallel to the structural layering is much easier: lattice friction opposed to the glide of [100] screw dislocations in the (010) plane is one order of magnitude lower than that in (001). This is the reason why we focus on this slip system in the present study. Lattice friction describes the anisotropic mechanical resistance of the material at 0 K only; it is still necessary to model the thermal activation of dislocation glide to determine how lattice friction evolves in the conditions of the D″ layer.
In this work, we present a theoretical study based on full atomistic modelling of plastic deformation, demonstrating easy glide of [100](010) dislocations in MgSiO_{3} postperovskite (Mgppv) at finite temperature and strain rate of the lowermost mantle.
Results
Dislocation core structure and Peierls barrier
The core structure of screw dislocations with the Burgers vector [100] is computed at the atomistic level using density functional theory (DFT). The atomic configuration of a straight dislocation line is shown in Fig. 1a. The stable core configuration is centred between two neighbouring Mg atoms and mainly spread in {011}. In (010), equivalent stable core configurations are found every ½[001], i.e. separated by a distance a′ of 3.07 Å. These cores correspond therefore to alternative variants labelled (I) spread in (011) and (II) spread in , as displayed in Fig. 1b. Although DFT represents high accuracy calculations, we also employ a pairwise potential to compute the minimum energy path (MEP) between two stable cores associated with dislocation glide in (010) and to carefully investigate finite size effects (as described in the supplementary materials). The MEP between configurations (I) and (II) is computed using the nudged elastic band (NEB) algorithm^{20,21,22}. The observed path reproduces the peculiar <011> zigzag trajectory (Fig. 1b) reported in a previous atomicscale study^{19}. The maximum of the Peierls potential corresponds to the MEP dislocation image when the (011) glide trajectory switches to . This high energy configuration is associated with 39 meV/b and 77.7 meV/b energy barriers for pairwise potential and ab initio simulations, respectively (Fig. 2b). Based on these results, the Peierls stress σ_{p} can be estimated from the maximum slope of the Peierls potential. Thus, DFT calculations lead to σ_{p} = 2.1 GPa, whereas the empirical potential simulations predict a lower value of 1 GPa (Fig. 2b). This apparent discrepancy is likely related to the known drawback of the pairwise potential parameterization used in this study. Indeed, both simulation techniques provide identical dislocation core structures, but the empirical potential underestimates the elastic stiffness coefficients C_{55} and C_{66} by a factor of 2 and, consequently, the corresponding anisotropic shear modulus μ (173 GPa vs 324 GPa) of the postperovskite^{23}. Qualitatively, one finds therefore the ratio σ_{p}/μ ~5·10^{−3} regardless of the level of atomic description.
Thermal activation and kinkpair mechanism
At finite temperature, the actual motion of a dislocation occurs through the nucleation and propagation of kinkpairs, i.e., a dislocation does not move as a straight line but partly bows out over the Peierls potential, as illustrated in Fig. 3a. Unlike common silicates, dislocation glide in (010) of Mgppv is associated with a relatively low energy barrier (in metals for example, a σ_{p}/μ ratio of 10^{−3} is usually expected). Under an applied stress σ_{a}, a straight dislocation will move upwards of the Peierls barrier to reach an equilibrium configuration from which a bulge can form. Energy minimization of the bowed configuration is obtained from a balance between the local forces on the dislocation, classically called the line tension, i.e., the force resulting from the Peierls barrier and the applied force. However, assuming that the fluctuation of the Peierls potential is small compared to the energy at rest leads to the standard line tension (LT) model for describing the kinkpair mechanism^{24,25,26}. Within the LT formulation, a screw dislocation line can be represented as a 1D function y(x), which describes its position y in the glide plane at each x coordinate along the dislocation line (Fig. 3a). Then, the dislocation line enthalpy H_{LT} can be estimated according to the following expression:
where Γ is the line tension (representing the stiffness of the dislocation line) and V_{p}(y(x)) − σ_{a}by(x) corresponds to the socalled “substrate enthalpy” in the 1DFrenkelKontorova model^{27,28}.
Computing line tension of a dislocation
To link the LT model with atomicscale simulations, we follow the work of Dezerald et al.^{29} and discretize the integral in Eq. (1) into n segments {Y_{n}} of length b:
where the sum over n accounts for the periodic boundary condition along the dislocation line.
To compute the line tension Γ, the energy cost associated with a dislocation bowout consistent with the first stage of kinkpair formation should be estimated. Bending a dislocation line requires breaking up the 1b translational symmetry of the simulation cell. Thus, the length of the supercell along the dislocation line is increased up to 2b. We consider a dislocation line that consists of two segments of length b: segment S_{1} remains in the Peierls valley (I), while the other segment S_{2} bows out towards the next valley (II), as shown in Fig. 4c. To compute this process at the atomic scale, the evolution of atomic displacements Δx along [100] during the dislocation glide from (I) to (II) is analysed, relying on MEP structural information obtained from NEB simulations. Dealing with a complex material, we mostly focus on the cation sublattice and allow anions to adapt to the local displacement of cations. Along the MEP, one Mg and four Si atoms exhibit the largest displacements Δx along the dislocation line among the cations in the crystal (Fig. 4a,b). The selected Mg atom is located directly between the two Peierls valleys and bears the maximum displacement amplitude (approximately 0.8 Å). The Δx amplitudes of the four Si atoms are two times smaller (Fig. 4b). Once the evolution of atomic positions along the MEP is defined, one can create the dislocation bowout while applying the exact displacements Δx consistent with the MEP (at zero stress) to the selected atoms belonging to the segment S_{2} of a dislocation line (Fig. 4d). The corresponding cations of segment S_{1} are fixed to their regular positions in the Peierls valley, constraining the degrees of freedom along the dislocation line. The computed change in energy, , related to the gradual disposition of line segment S_{2} is shown in Fig. 4b as a function of the reaction coordinate along [001]. Fitting the curvature of quadratic function ΔE_{LT} provides the line tension Γ = 9.2 eV/Å for DFT simulations (vs Γ = 7.1 eV/Å for empirical potential). The anisotropic line tension Γ_{el} = 2.08 eV/Å, calculated within Stroh formalism^{30} using the set of elastic constant C_{ij} of Mgppv^{31} computed with the same generalized gradient approximation (GGA) and pseudopotentials, is notably lower. Previous studies of bcc metals based on a similar simulation approach for computing LT at the atomic scale^{29,32} report comparable discrepancies between the LT values predicted atomistically and from elastic theory. Indeed, the latter does not account for the large effect of the dislocation core contribution, which leads to drastically underestimated Γ_{el} values.
Kinkpair activation enthalpy
Once line tension Γ is computed at atomic scale and the Peierls barrier V_{P} is known, the equilibrium kinkpair shape at a given stress and the corresponding critical kinkpair enthalpy H_{LT} can be calculated using Eq. (1). To solve Eq. (1), we rely on a trial function y(x) that describes the equilibrium shape of a symmetric kinkpair based on a combination of hyperbolic tangents^{28}:
where, as previously mentioned, a′ is the periodicity of the Peierls potential and α and m are variable parameters.
The saddle point on the H_{LT}(α,m) energy landscape ultimately defines both the enthalpy H_{LT} and the equilibrium kinkpair configuration y(x). As illustrated in the Supplementary materials, kinkpair configurations are characterized by extremely large widths in the range of 35–40b (approximately 100 Å) resulting from the very low Peierls barrier. With applied stress, this width changes moderately while the height of the kinkpair decreases rapidly as the straight part of the dislocation line moves upwards along the Peierls barrier. Note that such a wide kink shape justifies our choice of the LT model with respect to unreasonable direct atomistic computation of bowed configuration lines (which would require too many atoms). Figure 3b shows the computed enthalpy H_{LT} as a function of applied stress σ_{a}. As one expects from dislocation theory^{33}, the kinkpair enthalpy is maximum under zero stress, with a value of 2.69 eV, corresponding to twice the energy of a single kink H_{k}, and it vanishes when the applied stress is equal to the Peierls stress. The normalized kinkpair enthalpy 2H_{k}/μb^{3} is found to be approximately 5 · 10^{−2}, confirming the relatively low lattice friction borne by the [100](010) slip system in postperovskite.
Discussion
The dislocation core structure computed here confirms the spreading of the [100] screw core in {011}, as observed in previous semiempirical simulations^{19}. The extension of the core (with a halfwidth of approximately 1.8 Å) is found to be in reasonable agreement with the first results of the dislocation core determined using the PeierlsNabarro model^{31}. In Carrez et al.^{31}, the classical PN model, based on firstprinciples calculations (with the same GGA approximation and pseudopotentials as in this study) of generalized stacking fault energy, showed that the [100] screw dislocation should be compact in {011}, with a halfwidth of 1 Å. Discrepancies in core size between the present results and those of the PN model can be largely attributed to the GSF calculations method. As shown in Goryaeva et al.^{19}, the GSF computed by Carrez et al.^{31} involved atomic layers above the actual spreading layer of the core, delimited by neighbouring Mg rows. Consequently, the GSF energies used in the PN model were overestimated, leading to a narrower core. Nevertheless, note that despite a spreading in {011}, the easiest glide plane of [100] screw dislocations is (010). This is the result of a glide alternating between (011) and , defining a global macroscopic glide plane (010). This nonstandard behaviour could be evidenced only by full atomistic calculations. Based on calculation of the maximum height of the Peierls potential, we find a Peierls stress of 2.1 GPa, which is almost forty times smaller than the value reported in Carrez et al.^{31}. The reason for this discrepancy is that the PN calculation of σ_{p} did not rely on the right lattice periodicity a′. The lattice periodicity a′ is found here to be ½[001] without any ambiguities based on the computation of the exact core energy. Moreover, even considering the correct lattice periodicity for the Peierls potential, the PN model would certainly not be accurate in evaluating the Peierls potential for such a rearrangement of atoms around the line during the glide. Indeed, the dislocation glide process (through a zigzag scheme) violates one of the intrinsic hypotheses of the PN model, which is that the Peierls potential is evaluated in the plane of dislocation core spreading.
In postperovskite, the Peierls stress of [100](010) glide at 120 GPa is remarkably low. This result is consistent with the observation of [100](010) dislocations in several experimental studies of lowpressure CaIrO_{3} and CaPtO_{3} postperovskite analogues^{34,35,36,37}. Because Mgppv is unquenchable, experimental studies remain scarce^{10,11}. From the experimental point of view, in CaIrO_{3} postperovskite, the few TEM studies^{36} do not show strong evidence of lattice friction for [100] dislocations (as observed for olivine deformed at low temperature, where dislocation lines tend to be aligned along a particular direction). This can be supported by several results found here, i.e., the low normalized values of Peierls stress and kinkpairs enthalpy, and from the computed line tension. As recently demonstrated for perovskite material^{17}, kinkpair enthalpy evolution as a function of stress can be used to infer the evolution of the critical shear stress for dislocation glide as a function of temperature:
In Equation (4), p and q describe the evolution of the kinkpair enthalpy through the following relationship: ΔH_{LT}(σ_{a}) = 2H_{k} (1 − (σ_{a}/σ_{p})^{p})^{q}. Temperature T_{a}, often called “athermal temperature”, corresponds to the critical temperature at which lattice friction vanishes. Generally, kinkpair energy 2H_{k} scales with T_{a} according to 2H_{k} = CkT_{a}, where k is the Boltzmann constant; C is a function of a strain rate ε′, of the dislocation density ρ, and of the kink geometry (for more details see the corresponding section of the Supplementary Materials). C is classically found in a range of 20–30 (this is verified in metals^{38} and also in oxides^{39}). Therefore, taking 2H_{k} = 2.69 eV, p = 0.73 and q = 1.31 from the DFT calculations (considering experimental conditions, i.e., strain rates of 10^{−5} s^{−1} and dislocation density of 10^{12} m^{−2}), we find that lattice friction vanishes if the temperature is raised above 1,100 K.
More importantly, as demonstrated recently for bridgmanite^{17}, the previous equation can be used in Earth mantle conditions by adjusting the scaling factor of T_{a} to strain rates characteristic for convection in the Earth’s mantle. Assuming a typical value of 10^{−16 }s^{−1}, the corresponding temperature evolution for the critical stress for the glide of [100] dislocations is as shown in Fig. 5.
Implications
Our results on dislocation glide in Mgppv, including certain unexpected results, shed new light on the rheology of highpressure mantle phases. Indeed, all recent studies, either experimental or theoretical, of wadsleyite^{14,40,41,42}, ringwoodite^{15,43,44}, periclase^{16,45} and bridgmanite^{17,46} consistently show that pressure in the transition zone and lower mantle range leads to a significant increase of lattice friction, which inhibits dislocation glide as a strainproducing mechanism. In particular, these results have important implications regarding the (non)formation of seismic anisotropy from the deformation of the abovementioned phases. In this context, it is surprising to find that crystal chemistry and the formation of a layered structure can lead to a completely different behaviour. Our results demonstrate that the presence of weak {010} Mglayers containing a very short <100> lattice repeat of 2.5 Å leads to dislocation structures that can easily glide. We find that lattice friction is overcome at a critical temperature T_{a} far below the temperatures expected in the D″ layer (3,700–4,400 K^{47,48}). This finding has several unexpected consequences. The relative ease of slip between Mgppv and periclase suggests that the latter could become the stronger phase in the D″ layer. We are not yet in a position to fully establish this fact because additional deformation mechanisms must be activated in Mgppv to ensure compatibility of plastic deformation in an aggregate. However, the fact that diffusion is also fast in this phase^{1} suggests that complementary deformation mechanisms involving diffusion should be easily activated. Mgppv being the dominant phase in this assemblage, it is expected that the D″ layer in regions dominated by the Mgppv should exhibit a very low viscosity compared to the overlying mantle.
The implications of such a low viscosity layer have already been considered and discussed^{49,50,51}. The way that slabs behave when ultimately reaching the CMB is clearly affected, as is the broad dynamics of the CMB. However, the strongest implication is probably the enhancement of heat transfer from the core across the CMB, as earlier predicted by Buffett^{49} and more recently investigated numerically^{50,51}. The most testable implication of our results is, of course, the strong (010) crystal preferred orientation, which should develop upon flow in this weak layer. This is an important parameter because the D″ layer has long been recognized as being highly anisotropic. Although no consensus has yet been reached (see, for instance, Cottaar et al.^{9}), our finding that Mgppv exhibits dominant easy glide along (010) is consistent with the most recent studies of Nowacki et al.^{6} and Ford and Long^{7}.
In addition to a low viscosity, a low lattice friction in Mgppv may have important implications regarding seismic wave attenuation. A seismic (body) wave corresponds to strains in the range of 10^{−8}–10^{−6}, with periods in the range of 1–10 s. These values correspond to stresses of a fraction of a MPa at most, applied at a strain rate of 10^{−6} s^{−1} or lower. Under these conditions, the athermal temperature T_{a} will be greater than in Fig. 5, but, in any case, lower than 1,400 K (the value corresponding to a strainrate of 10^{−5} s^{−1}, constraining the dislocation density at 10^{8} m^{−2}; see supplementary Figure S5). This result shows that Mgppv will be in the athermal regime under seismic loading conditions at temperatures of the D″ layer, with dislocations moving freely without lattice friction.
This situation has not been considered up to now for seismic attenuation because most discussions have been driven by the example of olivine^{52,53}. Olivine exhibits lattice friction; thus, dislocations are prescribed to stay in their Peierls valleys, and dislocation damping can only result in a limited contribution from kink migration^{52,53}. For this reason, the most important source of attenuation in olivine has been linked to diffusionally assisted grain boundary sliding^{54}.
Under an applied stress σ_{a} and without lattice friction, a dislocation segment of length l will bow out with a curvature . This bowing gives rise to a reduction of the effective shear modulus, called the modulus defect or relaxation strength: , where ρ is the dislocation density^{55}. Assuming that l scales with , one can conclude that Δ_{0} could be as high as 15% (as an upper bound). Consequently, a shear wave travelling through the postperovskite containing dislocations could encounter a maximum velocity reduction approximately 7% compared to the ideal structure. In this regime, dislocation damping can be described using the vibrating string model^{55} which assumes that under an applied alternating stress, a dislocation characterized by a line tension can execute forced vibrations like a vibrating string. An alternating stress, such as the one associated with a seismic wave, will result in damping and energy dissipation. This model has two important consequences. At sufficiently high frequency, there exists a peak in tanϕ versus ω with a resonance at ω_{0} such that . The frequency ω_{0} is a function of Γ, i.e., the line tension of the dislocation, as previously computed, and of m_{l}, the effective mass per unit length of the dislocation line. This effective mass m_{l} can be computed by summing the squared displacements dq_{i} of all atoms i in a simulation cell in which a dislocation has moved from one Peierls valley to the next one, i.e., by dQ = a′, using the following expression^{56}:
In the previous expression, M_{i} corresponds to the mass of atom i; the effective mass of the dislocation thus incorporates the kinetic energy of surrounding atoms as if they were to respond adiabatically to dislocation motion. Computed from different cell sizes to account for finite size effects, we find (in units of atomic mass per unit length) m_{l} ~ 9.7 u/Å, which results in ω_{0} ~ 3.3·10^{4} s^{−1}. This frequency is higher than that of seismic waves, but it could allow experimental verification in the laboratory. At lower frequencies corresponding to seismic waves, the internal friction is proportional to ρl^{4} and to the frequency ω.
Our proposition of postperovskite being highly attenuating is consistent with the report of higher attenuation in the D″ by Anderson and Hart^{57} and Lawrence and Wysession^{58}. However, attenuation in the D″ layer is still not well constrained; this issue deserves more attention in the future to verify our prediction.
Conclusions
In this study, we modelled the thermally activated mobility of [100](010) slip in Mgppv using the line tension model in conjunction with atomicscale simulations.
We show that under pressure, temperature and strainrate conditions of the lowermost lower mantle, there is no lattice friction opposed to the glide of these dislocations. This easy glide of [100](010) dislocations has several implications:
 Although the exact viscosity of Mgppv cannot be calculated from a single slip system, we can predict a weak behaviour comparable with, if not weaker than, that of periclase.
 This conclusion supports the scenarios that involve a weak layer in the D″ layer with, in particular, enhancement of heat transfer from the core.
 Easy glide along (010) suggests development of marked crystal preferred orientations characterized by alignment of the (010) planes.
 The high mobility of dislocations allows for energy dissipation (vibrating string model) when a seismic wave travels through dislocationbearing Mgppv. We predict that deforming Mgppv should be characterized by strong seismic attenuation.
Methods
In this work, we employ an effective combination of the firstprinciples simulations with pairwise potential modelling. All simulations are performed with an external pressure of 120 GPa. Unit cell parameters for a Mgppv perfect crystal (Cmcm, Z = 4) at relevant conditions are a_{1} = 2.474 Å, a_{2} = 8.112 Å, a_{3} = 6.139 Å for the firstprinciples simulations and a_{1} = 2.521 Å, a_{2} = 8.124 Å, a_{3} = 6.050 Å for the semiempirical calculations.
Ab initio calculations are performed based on DFT within the GGA, as derived by Perdew and Wang^{59}, and the allelectron projector augmentwave (PAW) method, as implemented in VASP code^{60,61}. The outmost core radius for Mg, Si and O is 2.0, 1.9 and 1.52 au, respectively. To achieve computational convergence, we apply a planewave cutoff of 600 eV. The first Brillouin zone is sampled using the MonkhorstPack scheme^{62}, with a 10 × 1 × 1 kpoint grid for 1b simulation cells containing 360 atoms (the exact geometry is described below) and with a 4 × 1 × 1 grid for 2b cells containing 720 atoms. The convergence energy is 10^{−3} meV/atom.
Atomistic simulations within the semiempirical approach are carried out using the Buckingham form of a pairwise potential, with the parameterization derived by Oganov et al.^{63} for MgSiO_{3} perovskite. Transferability of this parameterization has been previously validated for modelling ground state properties and defects in the postperovskite phase^{23}. Molecular statics simulations are performed using the program package LAMMPS^{64}, which relies on Ewald summation methods for Coulombic interactions. Optimization of dislocation core configurations are performed using a conjugategradient (cg) algorithm, followed by a Hessianfree truncated Newton (hftn) algorithm, until the maximum force on an atom drops below 10^{−9} eV/Å (1.602·10^{−18} N). The NEB simulations are performed via fire damped dynamics, as required by the minimization procedure implemented in LAMMPS. The MEP is sampled with 24 points (configuration images), which are bounded with a spring constant of 0.1 eV/Å.
All simulations are performed by employing a quadrupole arrangement of screw dislocations in fully periodic atomic arrays. Such simulation cells contain two dislocations with positive and two dislocations with negative Burgers vectors arranged as a rectangular checkerboard pattern. This geometry allows for cancelling the longrange displacement field produced by a dislocation^{65} and ensures that interaction of the dislocations remains at a quadrupolar level and that the net force on each core is zero due to the periodic arrangement^{66}. The supercell is designed in such a way that [100] dislocation lines are parallel to x and [010] and [001] crystallographic directions are aligned with z and y, respectively. For core energy calculations and evaluation of the Peierls potential, the designed atomic systems are as thin as a single Burgers vector b along x, i.e., dislocation lines are straight and infinite due to the periodic boundary conditions. Because computation of line tension Γ requires bowing out the dislocation line and breaking the translational symmetry along x, we increase the length of the supercell along the dislocation line and employ 2b geometry, following the strategy proposed by Rodney and Proville^{67}. For DFT simulations, we employ the smallest possible atomic array of 36 Å × 48 Å, with quadrupolar arrangement of dislocations, which is further reduced by half to a nonrectangular (but still fully periodic) cell containing a dislocation dipole, following the procedure described by Bigger et al.^{66}. By applying periodic boundary conditions to such a dipole, the rectangular checkerboard pattern arrangement of dislocations (identical to that in the original rectangular cell containing four dislocations) is explicitly reproduced. The reduced simulation cells employed in this work for DFT simulations contain 360 and 720 atoms for 1b and 2b geometry, respectively. For the simulations performed with the pairwise potential, the size of the atomic arrays, containing a quadrupole of <100> screw dislocations, is gradually increased along y and z to track the size effect on the computed substrate enthalpy V_{P} and line tension Γ. A typical rectangular simulation cell has a size of 97 Å × 97 Å and 3840 and 7680 atoms in the case of 1b and 2b geometry, respectively.
Finally, for periodic arrangements of opposite Burgers vector dislocations, we compute the elastic interaction term, in accord with anisotropic elastic theory^{68}, and subtract from the energy computed via the NEB method. The exact location Y_{c} of each dislocation image is defined through analysing the relative atomic displacements near the dislocation cores, which are further used to compute the disregistry function S(Y) using the following expression^{69}:
where b is the Burgers vector, Y_{c} is the coordinate of a dislocation centre, and ζ is an adjustable parameter.
Additional Information
How to cite this article: Goryaeva, A. M. et al. Low viscosity and high attenuation in MgSiO_{3} postperovskite inferred from atomicscale calculations. Sci. Rep. 6, 34771; doi: 10.1038/srep34771 (2016).
References
Ammann, M. W., Brodholt, J. P., Wookey, J. & Dobson, D. P. Firstprinciples constraints on diffusion in lowermantle minerals and a weak D″ layer. Nature 465, 462–465 (2010).
Karato, S. I. The influence of anisotropic diffusion on the hightemperature creep of a polycrystalline aggregate. Phys. Earth Planet. Inter. 183, 468–472 (2010).
Lay, T. & Helmberger, D. V. The shearwave velocity gradient at the base of the mantle. J. Geophys. Res. 88, 8160 (1983).
Vinnik, L. P., Farra, V. & Romanowicz, B. Observational evidence for diffracted SV in the shadow of the Earth’s core. Geophys. Res. Lett. 16, 519–522 (1989).
Cottaar, S. & Romanowicz, B. Observations of changing anisotropy across the southern margin of the African LLSVP. Geophys. J. Int. 195, 1184–1195 (2013).
Nowacki, A., Walker, A. M., Wookey, J. & Kendall, J. M. Evaluating postperovskite as a cause of D" anisotropy in regions of palaeosubduction. Geophys. J. Int. 192, 1085–1090 (2013).
Ford, H. A. & Long, M. D. A regional test of global models for flow, rheology, and seismic anisotropy at the base of the mantle. Phys. Earth Planet. Inter. 245, 71–75 (2015).
Wenk, H.R. R., Cottaar, S., Tomé, C. N., McNamara, A. & Romanowicz, B. Deformation in the lowermost mantle: From polycrystal plasticity to seismic anisotropy. Earth Planet. Sci. Lett. 306, 33–45 (2011).
Cottaar, S., Li, M., McNamara, A. K., Romanowicz, B. & Wenk, H. R. Synthetic seismic anisotropy models within a slab impinging on the coremantle boundary. Geophys. J. Int. 199, 164–177 (2014).
Merkel, S. et al. Deformation of (Mg,Fe)SiO3 postperovskite and D″ anisotropy. Science 316, 1729–1732 (2007).
Miyagi, L., Kanitpanyacharoen, W., Kaercher, P., Lee, K. K. M. & Wenk, H. Slip systems in MgSiO3 postperovskite: implications for D″ anisotropy. Science 329, 1639–1641 (2010).
Walte, N. P. et al. Transformation textures in postperovskite: Understanding mantle flow in the D′ layer of the earth. Geophys. Res. Lett. 36, 3–7 (2009).
Miyagi, L., Kanitpanyacharoen, W., Stackhouse, S., Militzer, B. & Wenk, H. R. The enigma of postperovskite anisotropy: Deformation versus transformation textures. Phys. Chem. Miner. 38, 665–678 (2011).
Ritterbex, S., Carrez, P. & Cordier, P. Modeling dislocation glide and lattice friction in Mg2SiO4 wadsleyite in conditions of the Earth’s transition zone. Am. Mineral. in press (2016).
Ritterbex, S., Carrez, P., Gouriet, K. & Cordier, P. Modeling dislocation glide in Mg2SiO4 ringwoodite: Towards rheology under transition zone conditions. Phys. Earth Planet. Inter. 248, 20–29 (2015).
Cordier, P., Amodeo, J. & Carrez, P. Modelling the rheology of MgO under Earth’s mantle pressure. temperature and strain rates. Nature 481, 177–180 (2012).
Kraych, A., Carrez, P. & Cordier, P. On dislocation glide in MgSiO3 bridgmanite. Earth Planet. Sci. Lett. 452, 60–68 (2016).
Hirel, P., Kraych, A., Carrez, P. & Cordier, P. Atomic core structure and mobility of [100](010) and [010](100) dislocations in MgSiO3 perovskite. Acta Mater. 79, 117–125 (2014).
Goryaeva, A. M., Carrez, P. & Cordier, P. Modeling defects and plasticity in MgSiO3 postperovskite: Part 2—screw and edge [100] dislocations. Phys. Chem. Miner. 42, 793–803 (2015).
Henkelman, G. & Jonsson, H. Improved tangent estimate in the nudged elastic band method for finding minimum energy paths and saddle points. 113, 9978–9985 (2000).
Henkelman, G., Uberuaga, B. P. & Jó, H. A climbing image nudged elastic band method for finding saddle points and minimum energy paths. J. Chem. Phys. 22, 9901–9904 (2000).
Nakano, A. A space–timeensemble parallel nudged elastic band algorithm for molecular kinetics simulation. Comput. Phys. Commun. 178, 280–289 (2008).
Goryaeva, A. M., Carrez, P. & Cordier, P. Modeling defects and plasticity in MgSiO3 postperovskite: Part 1—generalized stacking faults. Phys. Chem. Miner. 42, 781–792 (2015).
Guyot, P. & Dorn, J. E. A critical review of the Peierls mechanism. Can. J. Phys. 45, 983–1016 (1967).
Seeger. In Dislocation (eds. Veyssière, P., Kubin, L. & Castaing, J. ) 141–178 (1984).
Caillard, D. & Martin, J. L. Thermally Activated Mechanisms in Crystal Plasticity. (Pergamon, 2003).
Frenkel, J. & Kontorova, T. A. On the theory of plastic deformation and twinning. Phys. Z. Sowj 13, 1–10 (1938).
Joós, B. & Duesbery, M. S. Dislocation kink migration energies and the FrenkelKontorowa model. Phys. Rev. B 55, 11161 (1997).
Dezerald, L., Proville, L., Ventelon, L., Willaime, F. & Rodney, D. Firstprinciples prediction of kinkpair activation enthalpy on screw dislocations in bcc transition metals: V, Nb, Ta, Mo, W, and Fe. Phys. Rev. B 91, 094105 (2015).
Stroh, A. N. Dislocations and Cracks in Anisotropic Elasticity. Philos. Mag. 3, 625–646 (1958).
Carrez, P., Ferré, D. & Cordier, P. Peierls– Nabarro model for dislocations in MgSiO3 postperovskite calculated at 120 GPa from first principles. Philos. Mag. 87, 3229–3247 (2007).
Proville, L., Ventelon, L. & Rodney, D. Prediction of the kinkpair formation enthalpy on screw dislocations in αiron by a line tension model parametrized on empirical potentials and firstprinciples calculations. Phys. Rev. B 87, 144106 (2013).
Hirth, J. P. & Lothe, J. Theory of dislocations. (Krieger Publishing Company, 1982).
Miyajima, N., Ohgushi, K., Ichihara, M. & Yagi, T. Crystal morphology and dislocation microstructures of CaIrO3: A TEM study of an analogue of the MgSiO3 postperovskite phase. Geophys. Res. Lett. 33, 1–4 (2006).
Miyagi, L. et al. Deformation and texture development in CaIrO3 postperovskite phase up to 6 GPa and 1300 K. Earth Planet. Sci. Lett. 268, 515–525 (2008).
Miyajima, N. & Walte, N. Burgers vector determination in deformed perovskite and postperovskite of CaIrO3 using thickness fringes in weakbeam darkfield images. Ultramicroscopy 109, 683–692 (2009).
McCormack, R. et al. The development of shape and crystallographicpreferred orientation in CaPtO3 postperovskite deformed in pure shear. Am. Mineral. 96, 1630–1635 (2011).
Tang, M., Kubin, L. P. & Canova, G. R. Dislocation mobility and the mechanical response of b.c.c. single crystals: A mesoscopic approach. Acta Mater. 46, 3221–3235 (1998).
Amodeo, J., Carrez, P., Devincre, B. & Cordier, P. Multiscale modelling of MgO plasticity. Acta Mater. 59, 2291–2301 (2011).
Nishihara, Y. et al. Plastic deformation of wadsleyite and olivine at highpressure and hightemperature using a rotational Drickamer apparatus (RDA). Phys. Earth Planet. Inter. 170, 156–169 (2008).
Kawazoe, T., Nishiyama, N., Nishihara, Y. & Irifune, T. Deformation experiment at P–T conditions of the mantle transition zone using DDIA apparatus. Phys. Earth Planet. Inter. 183, 190–195 (2010).
Farla, R., Amulele, G., Girard, J., Miyajima, N. & Karato, S. Highpressure and hightemperature deformation experiments on polycrystalline wadsleyite using the rotational Drickamer apparatus. Phys. Chem. Miner. 42, 541–558 (2015).
Hustoft, J. et al. Plastic deformation experiments to high strain on mantle transition zone minerals wadsleyite and ringwoodite in the rotational Drickamer apparatus. Earth Planet. Sci. Lett. 361, 7–15 (2013).
Miyagi, L. et al. Plastic anisotropy and slip systems in ringwoodite deformed to high shear strain in the Rotational Drickamer Apparatus. Phys. Earth Planet. Inter. 228, 244–253 (2014).
Marquardt, H. & Miyagi, L. Slab stagnation in the shallow lower mantle linked to an increase in mantle viscosity. Nat. Geosci. 8, 311–314 (2015).
Girard, J., Amulele, G., Farla, R., Mohiuddin, A. & Karato, S. Shear deformation of bridgmanite and magnesiowüstite aggregates at lower mantle conditions. Science 351, 144–147 (2016).
Boehler, R. Highpressure experiments and the phase diagram of lower mantle and core materials. Rev. Geophys. 38, 221–245 (2000).
Alfè, D., Gillan, M. J. & Price, G. D. Composition and temperature of the earth’s core constrained by combining ab initio calculations and seismic data. Earth Planet. Sci. Lett. 195, 91–98 (2002).
Buffett, B. A. A bound on heat flow below a double crossing of the perovskitepostperovskite phase transition. Geophys. Res. Lett. 34, L17302 (2007).
Nakagawa, T. & Tackley, P. J. Effects of lowviscosity postperovskite on thermochemical mantle convection in a 3D spherical shell. Geophys. Res. Lett. 38, 1–6 (2011).
Li, Y., Deschamps, F. & Tackley, P. J. Effects of lowviscosity postperovskite on the stability and structure of primordial reservoirs in the lower mantle. Geophys. Res. Lett. 41, 7089–7097 (2014).
Karato, S. & Spetzler, H. A. Defect microdynamics in minerals and solid state mechanisms of seismic wave attenuation and velocity dispersion in the mantle. Rev. Geophys. 28, 399–421 (1990).
Karato, S. A Dislocation Model of Seismic Wave Attenuation and Microcreep in the Earth: Harold Jeffreys and the Rheology of the Solid Earth. pure Appl. Geophys. 153, 239–256 (1998).
Jackson, I. et al. Grainsizesensitive viscoelastic relaxation in olivine: Towards a robust laboratorybased model for seismological application. Phys. Earth Planet. Inter. 183, 151–163 (2010).
Nowick, A. S. & Berry, B. S. Anelastic Relaxation in Crystalline Solids. (Academic Press, 1972).
Vegge, T. et al. Calculation of Quantum Tunneling for a Spatially Extended Defect: The Dislocation Kink in Copper Has a Low Effective Mass. Phys. Rev. Lett. 86, 1546–1549 (2001).
Anderson, D. L. & Hart, R. S. Q of the Earth. J. Geophys. Res. Solid Earth 83, 5869–5882 (1978).
Lawrence, J. F. & Wysession, M. E. QLM9: A new radial quality factor (Qμ) model for the lower mantle. Earth Planet. Sci. Lett. 241, 962–971 (2006).
Perdew, J. P. & Wang, Y. Accurate and simple analytic representation of the electrongas correlation energy. Phys. Rev. B 42, 13244 (1992).
Kresse, G. & Furthmüller, J. Efficient iterative schemes for ab initio totalenergy calculations using a planewave basis set. Phys. Rev. B 54, 11169 (1996).
Blöchl, P. E. Projector augmentedwave method. Phys. Rev. B 50, 17953 (1994).
Monkhorst, H. J. & Pack, J. D. Special points for Brillouinzone integrations. Phys. Rev. B 13, 5188–5192 (1976).
Oganov, A. R., Brodholt, J. P. & Price, G. D. Comparative study of quasiharmonic lattice dynamics, molecular dynamics and Debye model applied to MgSiO3 perovskite. Phys. Earth Planet. Inter. 122, 277–288 (2000).
Plimpton, S. Fast Parallel Algorithms for ShortRange Molecular Dynamics. J. Comput. Phys. 117, 1–19 (1995).
Cai, W. In Handbook of materials modelling (ed. Yip, S. ) 813–826 (Springer, 2005).
Bigger, J. et al. Atomic and electronic structures of the 90° partial dislocation in silicon. Phys. Rev. Lett. 69, 2224–2227 (1992).
Rodney, D. & Proville, L. Stressdependent Peierls potential: Influence on kinkpair activation. Phys. Rev. B 79, 094108 (2009).
Clouet, E. Dislocation core field. I. Modeling in anisotropic linear elasticity theory. Phys. Rev. B 84 (2011).
Peierls, R. The Size of a dislocation. Proc. Phys. Soc. 52, 34–37 (1940).
Acknowledgements
This work is supported by funding from the European Research Council under the Seventh Framework Programme (FP7), ERC Grant No 290424—RheoMan. Computational resources are provided by the CRIUniversité de Lille 1. P.C. acknowledges fruitful discussions at the CIDER 2016 summer program during final writing of the manuscript. Constructive inputs from the reviewers and the editor are also acknowledged.
Author information
Authors and Affiliations
Contributions
P.C. designed the study and supervised it with Ph.C.; A.M.G. performed atomicscale simulations with Ph.C. All authors discussed and interpreted the results and contributed to writing the manuscript.
Ethics declarations
Competing interests
The authors declare no competing financial interests.
Electronic supplementary material
Rights and permissions
This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/
About this article
Cite this article
Goryaeva, A., Carrez, P. & Cordier, P. Low viscosity and high attenuation in MgSiO_{3} postperovskite inferred from atomicscale calculations. Sci Rep 6, 34771 (2016). https://doi.org/10.1038/srep34771
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/srep34771
Further reading

Weak cubic CaSiO3 perovskite in the Earth’s mantle
Nature (2022)

Seismic anisotropy of the D″ layer induced by (001) deformation of postperovskite
Nature Communications (2017)

Modeling defects and plasticity in MgSiO3 postperovskite: Part 3—Screw and edge [001] dislocations
Physics and Chemistry of Minerals (2017)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.