Surface Science 554 (2004) 222–232 www.elsevier.com/locate/susc Orientation dependence of the stiffness of surface steps: an analysis based on anisotropic elasticity V.B. Shenoy *, C.V. Ciobanu Division of Engineering, Brown University, Box D, Providence, RI 02912, USA Received 31 October 2003; accepted for publication 27 January 2004 Abstract Motivated by recent experimental work that shows that the orientation dependence of the stiffness of steps on Cu surfaces differs significantly from the predictions of Ising and solid-on-solid modes, we study the effect of elastic anisotropy on the stiffness of steps on (1 0 0), (1 1 0) and (1 1 1) surfaces of cubic materials. The contribution to the stiffness due to elastic relaxation is obtained from a regularized continuum model similar to the Peierls–Nabarro model, which is commonly used to study the energetics of dislocations in crystals. Our analysis shows that while the difference the corresponding in the elastic self-energy of steps between the soft and hard orientations is only about 5 meV/A, The latter estimate is found to compare favorably with the difference in stiffness can be as large as 50–100 meV/A. experimental data on Cu(0 0 1) surfaces. Ó 2004 Elsevier B.V. All rights reserved. Keywords: Green’s function methods; Stepped single crystal surfaces; Surface structure, morphology, roughness, and topography; Copper 1. Introduction Self-organized surface nanostructures hold the promise for manufacture of micro-electronic devices and structures with unprecedented performance characteristics. While several mechanisms for self-assembling processes through kinetic instabilities and equilibrium phenomena on surfaces have been identified, the task of actually fabricating these devices reproducibly remains an engineering challenge. Further progress in this direction will, at least in part, depend on the ability * Corresponding author. Tel.: +1-401-8631475; fax: +1-4018639025. E-mail address: shenoyv@engin.brown.edu (V.B. Shenoy). to accurately measure and compute kinetic parameters such as adatom diffusion coefficients and step-edge barriers, as well as energetic driving forces obtained from orientation dependent surface and step energies. While novel experimental techniques and the availability of fast computers have provided important insights into atomic-scale mechanisms that play a role in determining the structure and stability of nanostructures, close agreement between measurements of kinetic and thermodynamic parameters and theory/computations has only been achieved for relatively few material systems. Recently Dieluweit and coworkers [1] investigated the orientation dependence of the stiffness of the surface steps on Cu(0 0 1) surfaces and found 0039-6028/$ - see front matter Ó 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.susc.2004.01.060 V.B. Shenoy, C.V. Ciobanu / Surface Science 554 (2004) 222–232 that the Ising and the solid-on-solid (SOS) models fail to describe the experimental data over a large range of angles that they analyzed. This observation has significant consequences for the fabrication of nanostructures where steps act as growth fronts, since fluctuations of the advancing steps are controlled by their stiffness. Also, in situations where a straight step becomes unstable during growth due to the Bales–Zangwill [2], the kinkrounding [3] or the strain-driven meandering instabilities [4], the wavelength of the fastest growing mode of meandering is determined by the stiffness of the steps. It is therefore important to understand the reasons for the disagreement between experimental observations and widely used theoretical models that are based on near-neighbor interactions. In their paper, Dieluweit et al. [1] have noted that models that go beyond the near-neighbor interactions, by including short-range interactions with other atoms within a finite cut-off distance, cannot explain the trends in the experimental results. They, however, point out that a consistent picture can perhaps be obtained if long-range interactions are included in the analysis. Such longrange interactions are usually mediated by elastic deformations [5] or are attributed to the presence of electric-dipoles in the vicinity of steps and kinks [6]. In the case of anisotropic materials, the elastic fields of surface steps will depend on their orientation: steps aligned along soft-elastic directions can achieve larger relaxation of the atoms at their edges. Similar arguments for the dependence of stiffness on orientation can also be made for the case of electrical interactions. In this paper we present an analysis of the orientation dependence of the stiffness of steps on (0 0 1), (1 1 0) and (1 1 1) surfaces of cubic materials due to anisotropy in elastic properties. We estimate difference in the stiffness of the steps aligned along the soft and hard which is comparaorientations to be 50 meV/A, ble to the deviations of the experimental results from the Ising or SOS estimates. In the case of isotropic elastic materials, the elastic fields of surface steps were first obtained by Marchenko and Parshin (MP) [5] by modeling the steps as force-dipoles. Their work shows that the stress field of a step decays as the inverse-square of 223 the distance from the location of the step. Within linear elasticity, the elastic self-energy of the step, however, does not converge due to the singular nature of the stress fields in the vicinity of the step. A similar situation occurs in the case of dislocations [7], where regularization methods such as the Peierls–Nabarro models [7,8] have been successfully employed to obtain the self-energy of the dislocations. The parameters that regularize the divergences in such models are obtained from atomic-scale calculations, where the atomic-displacements and stresses do not diverge in the ‘‘core’’ of the dislocation. In the present work, we extend the classic model of Marchenko and Parshin in two ways: First, we introduce a regularized model that leads to displacement fields that are well-behaved at the edge of the step and yield the expected dipolar decay at larger distances from the step-edge. The regularization procedure contains a phenomenological length parameter, which is of the order of the height of the step and can be extracted using atomic-scale calculations. Second, using the surface Green’s functions derived by Ting [9], we provide a general framework for obtaining the stress-fields and the self-energy of the steps in anisotropic crystals. The elastic selfenergy of the steps is then used to obtain numerical estimates for the orientation dependence of the stiffness of steps in cubic crystals. The paper is organized in the following way: In Section 2, we introduce a regularized model for the elastic field of surface steps in anisotropic crystals. The results of this section are used to derive the orientation dependence of the elastic self-energy and the stiffness of steps on cubic crystals in Section 3. In Section 4, we provide numerical estimates for the stiffness of steps on (1 0 0), (1 1 0) and (1 1 1) surfaces of Cu as a function of their orientation. A summary of the results and directions of future work are given in Section 5. The array of elastic constants and the surface Green’s functions used in our analysis are given in the appendices. 2. Elastic relaxation of a surface step In this section, we introduce a continuum model, similar to the celebrated Peierls–Nabarro (PN) 224 V.B. Shenoy, C.V. Ciobanu / Surface Science 554 (2004) 222–232 x2 bel ¼ 1 2 þ Z Cijkl ui;j ðxÞuk;l ðxÞ dx R Z ð2Þ sðsÞðsÞ ds; S h(x1) n S m x1 R Fig. 1. Schematic depiction of the profile of the step given by Eq. (1). The surface of the crystal is denoted by S, while the bulk is denoted by R. The unit vectors n and m are locally normal and tangent to the surface, respectively. where the first integral over the volume R occupied by the stepped crystal, is the elastic energy of the bulk atoms and the second term, over the surface of the crystal S, is the work done by the surface stress in stretching or compressing the atomicbonds on the surface. In (2), Cijkl is the array of elastic constants, ui;j ¼ oui =oxj , where ui is the icomponent of the displacement field, p s is the arc-ffi ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi length along the surface, so that ds ¼ dx21 þ dx22 , ðsÞ is the surface stretch and sðsÞ is the surface stress. The surface stretch is related to the material displacement field ui through the relation ðsÞ ¼ mi ðsÞui;j mj ðsÞ; model for dislocations, to obtain the elastic-fields of a surface step. The relaxed configuration of the flat surface prior to the creation of the step is chosen to be the reference state relative to which the elastic displacement fields are calculated. Using the coordinate axes shown in Fig. 1, the profile of the step is assumed to be of the form hs hðx1 Þ ¼ tan1 p x1 ; r0 ð1Þ where hs is the step-height and r0 is a parameter that ensures that the field variables (e.g. stresses, displacements) do not diverge at x1 ¼ 0. As we will see, with this choice of the profile, at distances from the step that are much larger than r0 , the elastic-fields coincide with the standard dipole model of Marchenko and Parshin. The parameter r0 can, in principle, be obtained from atomistic simulations by comparing either the displacement field or the elastic energies, as is done in the case of dislocations [10,11]; intuitively, we expect r0 to be of the order of hs or smaller. When a step is created on a surface by removing the topmost layer of atoms to the left of the step, the remaining atoms will undergo further elastic relaxation, so as to minimize the total energy of the system. The total energy involved in this relaxation process can be written as ð3Þ where mi ðsÞ is a unit vector tangent to the surface S at the point s. In the presence of the step, we assume that the surface stress has the form sðsÞ ¼ s0 þ Dh0 ðsÞ; ð4Þ where s0 is the surface stress of the flat surface, prime denotes differentiation with respect to s and D is a parameter that gives a measure of the deviation of the surface stress in the vicinity of the step relative to s0 . Below, we will show that the constant D which appears along with the derivative of the height with respect to the arclength is related to one of the components of the dipole in the MP model of surface steps. Since the first term in (2) is always positive, the total energy of the stepped surface can only be lowered if the second term is sufficiently negative; we will show that this is indeed the case. To this end, using the relations ui;j mj ¼ oui ox1 oui ox2 dui þ ¼ ox1 os ox2 os ds dmi ðsÞ ¼ kðsÞni ðsÞ; ds and ð5Þ where ni is the unit outward normal to S and kðsÞ is the curvature of the surface, which is taken to be positive if the center of curvature is outside the material, we can rewrite the second term of (2) as V.B. Shenoy, C.V. Ciobanu / Surface Science 554 (2004) 222–232 Z sðsÞðsÞ ds S Z dsðsÞ mi ðsÞ þ sðsÞkðsÞni ðsÞ ui ðsÞ ds: ¼ ds S ð6Þ To obtain the above equation, we have set boundary terms which come from an integral of the type Z d ðsðsÞui ðsÞmi ðsÞÞ ds ð7Þ ds S to zero, since the displacement field of the step vanishes at the infinitely remote boundaries of the system. Mechanical equilibrium requires that the variation in the total energy bel due to small perturbations in the displacement field dui from equilibrium values of ui must vanish to linear order in dui . Using (2) and (6), the first-order variation in energy is Z dbel ¼ dui;j ðxÞCijkl uk;l ðxÞ dx R Z dsðsÞ mi ðsÞ þ sðsÞkðsÞni ðsÞ dui ðsÞ ds: ds S ð8Þ The first term in the above equation can be further simplified to yield Z dui;j ðxÞCijkl uk;l ðxÞ dx R Z Z ¼ rij ðsÞnj ðsÞdui ðsÞ ds rij;j ðxÞdui ðsÞ dx; S R ð9Þ where rij ¼ Cijkl uk;l is the stress tensor. Because the variation in (8) must vanish for arbitrary dui , we obtain the usual equilibrium condition that the divergence of the stress must vanish at each material point, rij;j ¼ 0 in the region R. In addition, traction boundary conditions reflecting the structure of the stepped surface emerge in the form dsðsÞ ; ds ni ðsÞrij ðsÞnj ðsÞ ¼ sðsÞkðsÞ; mi ðsÞrij ðsÞnj ðsÞ ¼ ð10Þ 225 where the first term represents the shear traction on the surface due to the variation of surface stress with position along the surface, while the second term represents the normal traction on S due to the so-called Laplace pressure which arises from surface stress. The relation of these surface tractions to the dipoles introduced by Marchenko and Parshin can be understood by assuming that jh0 ðsÞj 1, which also allows the use of surface elastic Green’s function to obtain the displacement fields. This analysis is carried out in the next paragraph. When the condition jh0 ðsÞj 1 is satisfied, the surface tangent and normal vectors can be written, respectively, as ðm1 ; m2 Þ ¼ ð1; h0 ðx1 ÞÞ and ðn1 ; n2 Þ ¼ ðh0 ðx1 Þ; 1Þ to linear order in surface slope. In this limit, the 1- and 2-components of the surface tractions in (10) can be written as f1 ðx1 Þ r12 ðx1 Þ ¼ D d2 hðx1 Þ ; dx21 d2 hðx1 Þ f2 ðx1 Þ r22 ðx1 Þ ¼ s0 ; dx21 ð11Þ where the second derivative of the step profile can be calculated from (1). When the parameter r0 is small, using the relation lim r0 !0 d2 hðx1 Þ 2x1 r0 ¼ hs lim ¼ hs d0 ðx1 Þ; r0 !0 pðx2 þ r2 Þ2 dx21 1 0 ð12Þ we have fðx1 Þ ¼ dd0 ðx1 Þ where d ¼ Dhs e1 þ s0 hs e2 ð13Þ is the surface dipole introduced by Marchenko and Parshin and ei s are unit vectors in the coordinate directions. The above analysis shows that our continuum description reduces to the well-known MP description for small r0 , but does not yield diverging elastic fields near the step-edge. Indeed, within an atomic-scale description of a step, the displacement field of a step would be large but bounded at the step-edge; the parameter r0 can therefore be obtained by comparing the elastic fields of the present model (which we will derive below) with that of an atomistic simulation. In this respect, the present approach is similar to the PN model, where a closely related length parameter that regularizes the elastic field is derived from 226 V.B. Shenoy, C.V. Ciobanu / Surface Science 554 (2004) 222–232 atomic-scale calculations. Recent work, where abinitio density functional calculations are used to obtain the parameters in the PN model, can be found in Refs. [10,11]. The displacement field can be obtained from the surface tractions in (11) using the relation Z 1 ui ðx1 Þ ¼ Gij ðx1 xÞfj ðxÞ dx; ð14Þ e2 ,e02 1 where Gij is the surface Green’s function for an elastic half-space. This relation can be combined with (6) to express the elastic energy in (2) in terms of the surface tractions: Z Z 1 1 1 bel ¼ fi ðxÞGij ðx x0 Þfj ðx0 Þ dx dx0 : 2 1 1 ð15Þ We note that the elastic relaxation of surface atoms leads to a net gain in the surface contribution to the total energy (second term in (2)), which is twice the energy lost in the elastic deformation of the bulk (first term in (2)), so that there is an overall gain (or reduction) in the total energy of the system. Note that the elastic energy can depend on the orientation of the steps through the orientation dependence of the components of the dipole in (13) and the elastic constants (Cijkl ). Consequently, the contribution of the elastic energy to the stiffness of the steps will also be a function of the orientation of the steps. In what follows, we will use the surface Green’s function to obtain quantitative estimates for elastic self-energy and stiffness of surface steps in cubic materials. It is also important to note that there are other key contributions to the energy and stiffness of surface steps––these include the energy required to break the bonds (prior to elastic relaxation) and the longrange interactions due to electrical dipoles that might be formed at the step-edge. We do not consider these terms here; the aim of the present work is to estimate the contribution of the elastic fields to the stiffness of steps. 3. Orientation dependence of step-stiffness To obtain the surface Green’s function for an arbitrary orientation of the step, we first consider φ e3 φ e10 e03 e1 Fig. 2. Transformation of coordinate axes for computing the elastic constants for arbitrary orientations of the surface step. The step runs along the e3 -direction on a surface whose outward normal is directed in the e2 -direction (refer to Fig. 1). The ei -basis is obtained from the e0i -basis by a counter-clockwise rotation about the e2 -axis through an angle /. The unit vectors and the elastic constants in the e0i -basis for (0 0 1), (1 1 0) and (1 1 1) orientations of cubic crystals are given in Table 1 and Appendix A, respectively. the orientation dependence of the elastic constants. Let e2 and e3 in Fig. 2 denote, respectively, the unit vector that is normal to the surface on which the step is created and the unit vector that is parallel to the direction along which the step runs on this surface. The elastic constants for the crystal oriented in this fashion can be obtained from the elastic constants of the crystal aligned along high symmetry directions (denoted by e0i in Fig. 2) by using the standard transformation rules for fourthorder tensors. The basis vectors e0i that we will use for the (0 0 1), (1 1 0) and (1 1 1) surfaces of cubic materials and the corresponding array of elastic constants are given in Appendix A. With this choice, the ei -basis can be transformed to the e0i basis by a simple rotation about the e02 -direction (refer to Fig. 2); the transformation tensor can be then written as 0 1 cos / 0 sin / Rð/Þ ¼ @ 0 1 0 A; ð16Þ sin / 0 cos / where / is the angle made by the step-direction 0 (e3 ) with the e03 -direction. If Cpqrs are the elastic V.B. Shenoy, C.V. Ciobanu / Surface Science 554 (2004) 222–232 constants in the e0i -basis, the elastic constants in transformed basis can be written in terms of the components of the transformation tensor as 0 : Cijkl ð/Þ ¼ Rip ð/ÞRjq ð/ÞRkr ð/ÞRls ð/ÞCpqrs ð17Þ Following the work of Ting [9], the elastic constants in (17) can be used to express the surface elastic Green’s function in the form 8 p1 ðlog jx x0 jÞI 12 Sð/Þ L1 ð/Þ; > > < 0 x1 x > 0; 0 Gðx x0 Þ ¼ > p ðlog jx x jÞI þ 12 Sð/Þ L1 ð/Þ; > : x x0 < 0; ð18Þ where I is the identity tensor and the expressions for the Barnett–Lothe tensors [12], Sð/Þ and Lð/), are given in Appendix B. The displacement field of the surface step can be obtained by using (18) in (14); the resulting integral can be evaluated in closed form to give uðx1 Þ ¼ 1 x1 L1 ð/Þdð/Þ pr02 1 þ x21 r2 0 1 1 þ Sð/ÞL1 ð/Þdð/Þ: pr0 1 þ x21 r2 ð19Þ 0 In the above equation dð/Þ ¼ ½hs Dð/Þ; hs s0 ð/Þ; 0 denotes the orientation dependent dipole vector of the step. It can be seen that the displacements decay as 1=x1 when x1 r0 , which is expected for a dipolar field, while remaining well-behaved near the edge of the step. The above relation for the displacement field can be used in (15) to obtain the elastic self energy of the step, bel ð/Þ ¼ 1 di ð/ÞL1 ij ð/Þdj ð/Þ; 8pr02 ð20Þ from which the contribution of the elastic fields to the stiffness of steps can be written as [14] 2 ~ ð/Þ ¼ b ð/Þ þ d bel ð/Þ : b el el d/2 ð21Þ An order-of-magnitude estimate of the elastic self-energy of the step can be derived by taking r0 0:3hs , so that bel ð/Þ s20 =E, where E is comparable in magnitude to the Young’s modulus of 227 the material. Taking s0 1 J/m2 and E 100 GPa, which is typical for metals and semicon Since ductors, we find that bel ð/Þ 10 meV/A. the experimentally determined step formation (for SB and energies can vary from 10 meV/A DB steps on Si(0 0 1) [13]) to relatively large (for h1 1 0i steps on values like 100 meV/A Cu(0 0 1) [14]), the elastic contribution to the step energy can be anywhere from 10% to nearly 100% of the formation energy of steps. In the next section, we will focus attention on the stiffness of steps and show that the contribution of the elastic fields to the stiffness of steps can be as large as 50 meV/A. 4. Numerical estimates for Cu In this section we provide numerical results for the orientation dependence of the step formation energies stiffnesses for the (0 0 1), (1 1 0) and (1 1 1) surfaces of Cu. The elastic constants of Cu, referred to the standard h0 0 1i-oriented cubic basis are taken to be C11 ¼ 168:4 GPa, C12 ¼ 121:4 GPa and C44 ¼ 75:4 GPa [7]. As noted earlier, the elastic constants for a given orientation of the step can be obtained using (17), where the C 0 s are tabulated in Appendix A. In order to calculate the elastic self-energy, we require information on the intrinsic dipole, Dhs , whose strength is expected to depend on the details of the atomic bonding configurations at the step-edge and has to be obtained from an atomic-scale calculation. However, atomistic simulations of Shilkrot and Srolovitz [15] for h0 0 1i steps on Ag(0 0 1) and Ni(0 0 1) surfaces show that the magnitude of the intrinsic dipole is very close to the magnitude of the 2-component of the dipole, s0 hs (refer to (13)). In what follows, in order to obtain numerical estimates for the orientation dependence of the stiffness, we will assume that D ¼ s0 . Further, we take the regularization parameter r0 to be 0.33hs in all the cases; this value is comparable to the core parameter that is required to make the dislocation energies computed in atomistic calculations equal that of linear anisotropy elasticity theory in a variety of material systems (refer to Table 8-2, p. 232 of Ref. [7]). 228 V.B. Shenoy, C.V. Ciobanu / Surface Science 554 (2004) 222–232 4.1. Steps on Cu(0 0 1) (in meV/A0) -12 We first focus attention on the orientation dependence of the stiffness of steps on Cu(0 0 1). In general, the elastic self-energy in (20) includes contributions from the two components of the dipole in (13) and the contributions from the interactions between these components. The interaction terms however vanish on the (0 0 1) surface due to the crystallographic symmetries of this surface, so that bel ð/Þ ¼ h2s 2 1 D L11 ð/Þ þ s20 L1 22 ð/Þ 2 8pr0 bD þ bs ; (a) -13 βD(φ) βτ(φ) -14 -15 0 ð22Þ where we have used the fact that the surface stress tensor of the (0 0 1) surface is isotropic [17]. Since we were not able to find experimental data for the surface stress of Cu surfaces, we have used the results of Wan and coworkers [16], calculated by using a modified-embedded-atom potential. The orientation dependence of the self-energy of each component of the dipole, obtained by taking D ¼ s0 ¼ 2:1 J/m2 , is shown in Fig. 3(a). We find that the self-energy of the step that runs along h1 0 0i, which is the soft-elastic direction is lower than the self-energy of the h1 1 0i-step by about 3 The corresponding difference in the stiffmeV/A. ness of steps, shown in Fig. 3(b), however is about one order of magnitude larger. This can be understood by noting that the h1 1 0i-direction is a local maximum of the self-energy (and hence has negative curvature at / ¼ 0), while the h1 0 0idirection is a local maximum with a positive curvature. Since the stiffness is the sum of the selfenergy and its curvature, the energy cost associated with changing the orientation of the step from the soft-elastic direction becomes large. In the above analysis, since the dipoles were taken to be independent of the orientation, the angle dependence of the stiffness stems entirely from the anisotropy of the elastic constants. Recent experimental work in this case has shown that while the stiffnesses of steps close to the h1 1 0i-direction agree with the predictions of the Ising and SOS models, these models underestimate the stiffness of the steps close to the h1 0 0i-direction by about 10 20 30 40 φ (in degrees) (in meV/A0) (b) 50 ∼ ∼ βD(φ)-βD(0) 40 ∼ ∼ βτ(φ)−βτ(0) 30 20 10 0 0 10 20 30 φ (in degrees) 40 Fig. 3. The elastic self-energy (a) and the contribution of elastic fields to the stiffness of steps (b) plotted as a function of the orientation of the step relative to the h1 1 0i-direction on Cu(0 0 1). The subscripts s and D refer to contributions from the 1- and 2-components of the dipole in (13), respectively. 1 As noted earlier, the contribution 36 meV/A. from the elastic fields have to be added to the stiffness obtained from models based on short-range 1 ~ k Þ, in the Fig. 2 of Ref. [1] shows the magnitude of kB T =ðba case of h1 0 0i-steps, to be 0.2 for the experimental data and 0.7 for the Ising/SOS model. When kB T ¼ 26 meV and the we find b ~ to be 51 and 15 meV/ atomic spacing ak ¼ 2:55 A, for the data from measurements and Ising model, respecA tively. V.B. Shenoy, C.V. Ciobanu / Surface Science 554 (2004) 222–232 interactions, to account for the long-range relaxation effects due to atomic displacements. If the long-range interactions are to be invoked to account for the discrepancy between the predictions of the Ising/SOS models and experimental data, the work of Dieluweit et al. would suggest that the difference in the contribution to the stiffness from these interactions between the h1 1 0i and the h1 0 0i This value is indeed directions is about 36 meV/A. comparable in magnitude to the difference in stiff predicted by the elastic fields in ness (50 meV/A) Fig. 3. Given the simplicity of the analysis, this agreement is indeed encouraging. 0 (in meV/A ) -4 4.3. Steps on Cu(1 1 1) On the (1 1 1) surface of cubic materials, the interactions between the two components of the dipole in (13) do not vanish on the basis of symmetry requirements. Consequently, the decomposition of the self-energy given in (22) ceases to hold on this surface. Also, the surface stress tensor of this surface is isotropic due to the presence of a βτ(φ) (a) βD(φ) -6 -8 -10 -12 0 4.2. Steps on Cu(1 1 0) As in the case of the (0 0 1) surface, the interactions between the distinct components of the dipole in (13) vanish for the steps on this surface, so that the decomposition of the self-energy given in (22) remains valid. The surface stress, however, is anisotropic due to absence of a fourfold symmetry axis on this surface [17]. The orientation dependence of the self-energies, calculated by taking s0 ð/Þ ¼ 1:22 cos2 / þ 1:95 sin2 / J/m2 [16], is given in Fig. 4. The self-energies decrease in a monotonous manner from the hard-elastic h1 1 0idirection to the soft h1 0 0i-direction (/ ¼ 90°). The difference in the self-energies between these which is comparable directions is about 8 meV/A, to the behavior on the (0 0 1) surface. Unlike the (1 0 0) surface, the stiffness of the steps on this surface shows a non-monotonic trend, with the stiffness of the h1 0 0i-steps exceeding the stiffness The anof the h1 1 0i-steps by about 40 meV/A. isotropies in both the surface stress tensor and the elastic constants lead to the complex behavior of the stiffness observed in the present case. 229 20 40 60 80 φ (in degrees) (in meV/A0) (b) 40 30 20 ∼ ∼ βD(φ)-βD(0) 10 ∼ ∼ βτ(φ)−βτ(0) 0 0 20 40 60 φ (in degrees) 80 Fig. 4. The elastic self-energy (a) and the contribution of elastic fields to the stiffness of steps (b) plotted as a function of the orientation of the step relative to the h1 1 0i-direction on Cu(1 1 0). The subscripts s and D refer to contributions from the 1- and 2-components of the dipole in (13), respectively. threefold rotation axis [17]. The total elastic selfenergies and stiffnesses obtained by taking s0 ¼ 1:92 J/m2 [16] and D ¼ s0 , D ¼ 0:5s0 and D ¼ 0 are given in Fig. 5. For any non-zero value of D, the self-energy shows a non-monotonic behavior with a minimum at an angle between the h1 1 2i-direction and the h1 1 0i-direction, while the behavior is monotonic when D ¼ 0. The more complicated behavior of the self-energy can be traced to the interaction between the two components of the 230 V.B. Shenoy, C.V. Ciobanu / Surface Science 554 (2004) 222–232 β(φ) (in meV/A0) Once again, the non-monotonic behavior in the orientation dependence of the step-stiffness for D 6¼ 0 can be traced to the interactions between the two distinct components of the step-edge dipole. (a) D=0 -12 D = 0.5τ0 -14 5. Concluding observations -16 -18 -20 D = τ0 -22 -24 0 10 20 30 φ (in degrees) ∼ ∼ β(φ)−β(0) 0 (in meV/A ) (b) D = τ0 D = 0.5τ0 28 24 D=0 20 16 12 8 4 0 0 10 20 φ (in degrees) Fig. 5. The elastic self-energy (a) and the contribution of elastic fields to the stiffness of steps (b) plotted as a function of the orientation of the step relative to the h1 1 0i-direction on Cu(1 1 1). The symbols s and D refer the 1- and 2-components of the dipole in (13), respectively. dipole at the step-edge; the location of the minimum in the self-energy is determined by the ratio of D and s0 (refer to (13)). A key point to be noted here is that the variation of the self-energy is an than the order of magnitude smaller (0.5 meV/A) variations for the (1 0 0) and (1 1 0) surfaces. The step-stiffness, on the other hand, has a variation of for all the cases considered. about 30 meV/A In summary, we have analyzed the orientation dependence of the stiffness of surface steps with a continuum model based on anisotropic elasticity. Our key finding is that the anisotropy in elastic properties and surface stresses can result in a 5 difference in the step self-energies between meV/A the soft and hard orientations, while the difference in the step-stiffnesses can be one order of magnitude higher. This estimate for the variation in stiffness is comparable to the difference in stiffness between the experimental data and Ising/SOS models (35 for the h1 1 0i steps on Cu(0 0 1). In our meV/A) calculations for the Cu(0 0 1) surface, the orientation dependence of stiffness can be solely attributed to the anisotropy in elastic constants. This result suggests that if one considers materials with smaller elastic anisotropy (e.g., tungsten or aluminum), then the orientation dependence of the stiffness of the steps on the (0 0 1) surface of these materials should be negligible; it would be interesting to see how well the experimental data in these systems compares with the predictions of models based on short-range interactions. The present model requires the knowledge of parameters such as the regularization-length r0 and the dipole D, which are expected to depend on atomic-scale phenomena (such as bonding) at the step-edge. In our analysis, we have only used an order-of-magnitude estimate for these parameters; a more realistic description can be obtained with the aid of atomistic simulations. In general, data on step-energies from experiments or quantum mechanical simulations (see for example Ref. [18]) would contain information on both the electrical and elastic interactions; it is usually difficult to separately obtain the contribution of the individual interactions. The present model could also be useful in this regard––if the atomic displacements are known from experiments/atomistic calculations, the magnitudes of the dipoles and hence the V.B. Shenoy, C.V. Ciobanu / Surface Science 554 (2004) 222–232 elastic interactions can be extracted by comparison with our model. 0 0 ¼ C13 ¼ C12 ; C12 231 0 C23 ¼ 12ðC11 þ C12 Þ C44 ; 0 C44 ¼ 12ðC11 C12 Þ; 0 0 C55 ¼ C66 ¼ C44 ; Acknowledgements III. (1 1 1) Surface Support from the MRSEC at Brown University (DMR-0079964), National Science Foundation grants CMS-0093714 and CMS-021 0 095, and the Salomon Research Award from the Graduate School at Brown University is gratefully acknowledged. Appendix A. Elastic constants for different surface orientations 0 0 C11 ¼ C33 ¼ 12ðC11 þ C12 Þ þ C44 ; 0 C22 ¼ 13ðC11 þ 2C12 þ 4C44 Þ; 0 0 ¼ C23 ¼ 13ðC11 þ 2C12 4C44 Þ; C12 0 C13 ¼ 16ðC11 þ 5C12 2C44 Þ; 0 0 0 ffiðC11 þ C12 þ 2C44 Þ; ¼ C16 ¼ C54 ¼ p1ffiffiffi C36 18 0 0 ¼ C66 ¼ 13ðC11 C12 þ C44 Þ; C44 Below, we provide the expressions for the ar0 rays of elastic constant Cpqrs (refer to Eq. (17)) for the (0 0 1), (1 1 0) and (1 1 1) surfaces. The coordinate axes (e0i s in Fig. 2) for each of these orientations are given in Table 1. We have used the matrix notation of Voigt, where the pair of indices 11, 22, 33, 23, 13, 12 are denoted by the labels 1, 2, 3, 4, 5 and 6, respectively. C11 , C12 and C44 refer to standard cubic elastic constants where the cube axes are oriented along h1 0 0i. Note that only non-zero components of C 0 s are given below. 0 C55 ¼ 16ðC11 C12 þ 4C44 Þ: I. (0 0 1) Surface Rik ðh; /Þ ¼ Cijks ð/Þnj ms ; 0 0 ¼ C33 ¼ 12ðC11 þ C12 Þ þ C44 ; C11 0 C22 ¼ C11 ; 0 0 C12 ¼ C23 ¼ C12 ; 0 C13 ¼ 12ðC11 þ C12 Þ C44 ; 0 0 C44 ¼ C66 ¼ C44 ; 0 C55 ¼ 12ðC11 C12 Þ; ¼ C11 ; 0 C22 ¼ 0 C33 ¼ 1 ðC11 2 þ C12 Þ þ C44 ; Table 1 The table shows the unit vectors e01 , e02 and e03 (refer to Fig. 2) that define the coordinate systems for computing the elastic constants in the case of the (0 0 1), (1 1 0) and (1 1 1) surface orientations Surface e01 e02 e03 (0 0 1) p1ffiffi[ 1 1 0] 2 (1 1 0) [0 0 1] p1ffiffi[1 1 2] [0 0 1] p1ffiffi[1 1 0] p1ffiffi[1 1 0] 2 1 0] p1ffiffi[1 2 1 0] p1ffiffi[1 2 (1 1 1) 6 2 p1ffiffi[1 1 1] 3 The Barnett–Lothe tensors Sð/Þ and Lð/Þ [12] in the surface Green’s function (refer to (18)) are derived from matrices Q, R and T whose components are given by Qik ðh; /Þ ¼ Cijks ð/Þnj ns ; ðB:1Þ Tik ðh; /Þ ¼ Cijks ð/Þmj ms ; in which h is a real parameter and ni ¼ ½cos h; sin h; 0; ðB:2Þ mi ¼ ½ sin h; cos h; 0: II. (1 1 0) Surface 0 C11 Appendix B. Surface Green’s function Using these matrices, the Barnett–Lothe tensors can be written as [9] Z 1 p Sð/Þ ¼ N1 ðx; /Þ dx and p 0 ðB:3Þ Z 1 p N3 ðx; /Þ dx; Lð/Þ ¼ p 0 where N1 ðh; /Þ ¼ T1 ðh; /ÞRT ðh; /Þ and N3 ðh; /Þ ¼ Rðh; /ÞT1 ðh; /ÞRðh; /Þ Qðh; /Þ: ðB:4Þ 232 V.B. Shenoy, C.V. Ciobanu / Surface Science 554 (2004) 222–232 References [1] S. Dieluweit, H. Ibach, M. Geisen, T.L. Einstein, Phys. Rev. B 67 (2003) 121410(R). [2] G.S. Bales, A. Zangwill, Phys. Rev. B 41 (1990) 5500. [3] O. Pierre-Louis, M.R. D’Orsogna, T.L. Einstein, Phys. Rev. Lett. 82 (1999) 3661. [4] J. Tersoff, E. Pehlke, Phys. Rev. Lett. 68 (1992) 816. [5] V.I. Marchenko, Y.A. Parshin, Sov. Phys. JETP 52 (1980) 129. [6] C. Jayaprakash, C. Rottman, W.F. Saam, Phys. Rev. B 30 (1984) 6549. [7] J.P. Hirth, J. Lothe, Theory of Dislocations, second ed., John Wiley, 1982. [8] R. Peierls, Proc. Roy. Soc. 52 (1940) 23; F.R.N. Nabarro, Proc. Phys. Soc. Lond. 59 (1947) 256. [9] T.C.T. Ting, Q.J. Mech. Appl. Math. 45 (1991) 119. [10] G. Lu, N. Kioussis, V.V. Bulatov, E. Kaxiras, Mater. Sci. Eng. A 309 (2001) 142. [11] O.N. Mryasov, Y.N. Gornostyrev, A. van Schilfgaarde, A.J. Freeman, Acta. Mater. 50 (2002) 4545. [12] D.M. Barnett, J. Lothe, Phys. Norv. 7 (1973) 13. [13] H.J.W. Zandvliet, Rev. Mod. Phys. 72 (2000) 593. [14] M. Giesen, Prog. Surf. Sci. 68 (2001) 1. [15] L.E. Shilkrot, D. Srolovitz, Phys. Rev. B 53 (1996) 11120. [16] J. Wan, Y.L. Fang, D.W. Gong, S.G. Shen, X.Q. Fan, Model. Sim. Mater. Sci. Eng. 7 (1999) 189. [17] R.C. Cammarata, Prog. Surf. Sci. 46 (1994) 1. [18] V.B. Shenoy, C.V. Ciobanu, Phys. Rev.B. 67 (2003) 081402(R).