betaFeedback

WorksPhysics and Astronomy

Thermodynamic theory of dislocation/grain boundary interaction

Abstract

The thermodynamic theory of dislocation/grain boundary interaction, including dislocation pile-up against, absorption by, and transfer through the grain boundary, is developed for nonuniform plastic deformations in polycrystals. The case study is carried out on the boundary conditions affecting work hardening of a bicrystal subjected to plane constrained shear for three types of grain boundaries: (i) impermeable hard grain boundary, (ii) grain boundary that allows dislocation transfer without absorption, (iii) grain boundary that absorbs dislocations and allows them to pass later.

Introduction

The interaction between dislocations and grain boundaries plays an essential role in the mechanical response of polycrystalline materials to loading. Several outcomes of this interaction are possible. At low dislocation density the grain boundary blocks the movement of dislocations and forces them to pile-up against it, with the byproduct of linearly increased kinematic work hardening. As the dislocation density exceeds some critical value, a certain amount of dislocations is absorbed by the grain boundary, causing an increase in the surface dislocation density and the misorientation between adjacent grains. Dislocations can also be emitted at a later stage of straining in the neighboring grain or transferred through the grain boundary. In the presence of grain boundaries, non-uniform plastic deformation occurs, leading to non-redundant (geometrically necessary) dislocations, so the need for continuum models that include the gradient of plastic deformation becomes evident. Berdichevsky & Sedov [4] were the first to introduce the curl of plastic deformation (Nye-Bilby-Kröner’s tensor) into the continuum dislocation theory (CDT) and associate it with the density of non-redundant dislocations. They also proposed the thermodynamics of dislocations, using this density of non-redundant dislocations as a state variable in the free energy and giving the variational formulation of CDT (see the further developments in [5, 29]). The main drawback of this thermodynamic approach is the absence of configurational entropy and its dual quantity, the effective temperature of dislocations, introduced a few decades later by Langer et al. [22]. Together with the density of redundant dislocations, this configurational entropy enables ones to build up the meaningful thermodynamics of dislocations satisfying two universal laws for plastic flows discovered in [28, 25]. Recently, one of us extended CDT by introducing the density of redundant dislocations as well as the configurational entropy as additional state variables [26]. He has shown that the new theory, called thermodynamic dislocation theory (TDT), can predict both kinematic and isotropic work hardening. This TDT has been applied to model grain boundaries as hard obstacles that block dislocations, leading to dislocation pile-up, kinematic work hardening, and Bauschinger effect [32, 17]. Piao & Le [38] further applied the theory to the problem of interaction between screw dislocations and permeable grain boundary based on the experiments performed by [20]. There, an interfacial dissipation potential is proposed and additional boundary conditions for the grain boundary are derived to model screw dislocation/grain boundary interaction.

Alternative phenomenological models that include the interaction between dislocations and grain boundaries have been proposed in the context of strain gradient plasticity by Fleck et al. [11], Aifantis [2], Gao et al. [14], Gurtin [18], and a large number of investigations have been carried out within this approach. Fredriksson & Gudmundson [13] proposed an energetic model according to which the plastic work is stored as interfacial energy that depends on the jump of plastic slip at the grain boundary, while Fleck & Willis [12] argued that material hardening due to the grain boundary is mainly carried by energy dissipation. Note that both models allow for a discontinuity in the plastic slip at the grain boundary. Gurtin [19] introduced a Burgers’ tensor at the grain boundary, analogous to the Nye-Bilby-Kröner’s tensor for bulk crystals, to determine both the interfacial energy and dissipation. One merit of this tensorial measure is that it allows the theory to satisfy the slip transfer criteria. Van Beers et al. [44] proposed a similar theory that takes into account the criteria of slip transfer, but differs from the formulation of Gurtin [19]. Inter-grain interaction modules from two theories are compared by Gottschalk et al. [15]. We also mention the references [3, 45, 10, 46, 39, 40] which use the same approach. Similar to the CDT developed in [4, 5, 29], this standard strain gradient plasticity approach ignores completely configurational entropy, so none of the work done within this approach is consistent with thermodynamics of dislocations proposed in [22, 28, 25].

The aim of this work is to model the interaction of edge dislocations with the grain boundary in polycrystals within the framework of TDT. In contrast to the interaction between screw dislocations and grain boundaries studied in our previous paper [38], the processes of dislocation pile-up against, absorption by, and transfer through the grain boundary may occur in this situation. Using the variational formulation together with the proposed surface energy and dissipation potential, we will derive the interface boundary conditions for these processes. The case study is carried out on the boundary conditions affecting work hardening of a bicrystal subjected to plane constrained shear for three types of grain boundaries: (i) impermeable hard grain boundary, (ii) grain boundary that allows dislocation transfer without absorption, (iii) grain boundary that absorbs dislocations and allows them to pass later. Note that, according to various experimental observations reported e.g. in [42, 36], the grain boundary could be migrated during this dislocation absorption, but this possible migration is ignored in this continuum model. We show especially the effect of dislocation absorption on work hardening.

The paper is organized as follows. After this brief Introduction, we lay down the kinematics and thermodynamic framework of TDT in Section 2. In Section 3, we analyze the effects of dissipation potential and interfacial energy on the work hardening of materials and derive the conditions at the grain boundary for various processes during plastic deformation. In the subsequent Section 4, the application of the developed theory to the problem of a bicrystal subjected to plane constrained shear is illustrated and its numerical treatment is presented. Detailed numerical simulations are then performed and the results are discussed. We conclude the paper with a brief summary in Section 5.

Formulation for bulk crystals

Kinematics

In the small strain theory, we ignore the differences between the Lagrangian and Eulerian coordinates and assume that the total displacement gradient (distortion) ∇u\nabla\mathbf{u} can be decomposed additively into elastic and plastic counterparts

∇u=βe+β,\nabla\mathbf{u} = \boldsymbol{\beta}^{e} + \boldsymbol{\beta},

where u(x)\mathbf{u}(\mathbf{x}) is a displacement field, x=(x1,x2,x3)\mathbf{x} = (x_{1}, x_{2}, x_{3}) a position vector of a material point of the body, and ∇\nabla the nabla operator. The incompatible plastic distortion of the crystal having nn active slip systems takes the form

β(x)=∑α=1nβα(x)sα⊗mα,\boldsymbol{\beta}(\mathbf{x}) = \sum_{\alpha=1}^{n} \beta^{\alpha}(\mathbf{x})\mathbf{s}^{\alpha} \otimes\mathbf{m}^{\alpha},

with sα\mathbf{s}^{\alpha} and mα\mathbf{m}^{\alpha} being the pair of unit vectors denoting the slip direction and the normal to the slip planes of the corresponding α\alpha-th slip system. The plastic slip of the α\alpha-th slip system, βα(x)\beta^{\alpha}(\mathbf{x}), is assumed to be continuously differentiable function except at the interfaces or grain boundaries. Since sα\mathbf{s}^{\alpha} and mα\mathbf{m}^{\alpha} are mutually orthogonal, the trace of the plastic distortion always vanishes, tr⁡β=0\operatorname{tr}\boldsymbol{\beta}=0, which indicates volume-preserving plastic deformation due to the conservative motion of dislocations. The total strain ε\boldsymbol{\varepsilon} and the plastic strain εp\boldsymbol{\varepsilon}^{p} are the symmetric part of the displacement gradient and the plastic distortion, respectively, so that

ε=12(∇u+u∇),εp=12(β+βT).\boldsymbol{\varepsilon} = \frac{1}{2}(\nabla\mathbf{u} + \mathbf{u}\nabla), \qquad\boldsymbol{\varepsilon}^{p} = \frac{1}{2}(\boldsymbol{\beta} + \boldsymbol{\beta}^{T}).

Accordingly, the elastic strain equals

εe=ε−εp.\boldsymbol{\varepsilon}^{e} = \boldsymbol{\varepsilon} - \boldsymbol{\varepsilon}^{p}.

The net Burger’s vector B\mathbf{B} of non-redundant dislocations whose lines cross the area SS bounded by the closed curve ∂S\partial S is given by

B=∮∂Sβ⋅dx=∫Sα⋅n da,\mathbf{B} = \oint_{\partial S} \boldsymbol{\beta} \cdot\mathrm{d}\mathbf{x} = \int_{S} \boldsymbol{\alpha} \cdot\mathbf{n}\,\mathrm{d}a,

where α\boldsymbol{\alpha} is the Nye-Bilby-Kröner’s dislocation tensor [37, 7, 21],

α=−β×∇,(αij=ϵjklβil,k).\boldsymbol{\alpha} = -\boldsymbol{\beta} \times\nabla, \qquad(\alpha_{ij} = \epsilon_{jkl}\beta_{il,k}).

In particular, if there is only one active slip system such that s\mathbf{s} and m\mathbf{m} lie in the (x1,x2)(x_1,x_2)-plane, while the dislocation lines are parallel to the x3x_3-axis, then α\boldsymbol{\alpha} under the plane strain has only two nonzero components αi3=si(∂sβ)\alpha_{i3}=s_i(\partial_s\beta), where ∂s\partial_s denotes the derivative in the s\mathbf{s}-direction. Consequently, the scalar density of the non-redundant dislocations is given by

ρg=∣dB∣b da=1b∣∂sβ∣,\rho^{\mathrm{g}}=\frac{|\mathrm{d}\mathbf{B}|}{b\,\mathrm{d}a}=\frac{1}{b}|\partial_s\beta|,

with bb being the magnitude of Burger’s vector.

Thermodynamic framework

Suppose the material body in form of a slab of constant depth LL occupies the 3-D domain V\mathcal{V}, and its boundary ∂V\partial\mathcal{V} consists of two nonintersecting surfaces, ∂k\partial_k and ∂s\partial_s, on which the displacement field u(x)\mathbf{u}(\mathbf{x}) and the traction-free condition are specified, respectively. Assuming that no body force acts on the crystals, the energy functional reads

I=∫Vψ(εe,ρg,ρr,χ) d3x,(1)I=\int_{\mathcal{V}}\psi(\boldsymbol{\varepsilon}^{e},\rho^{\mathrm{g}},\rho^{\mathrm{r}},\chi)\,\mathrm{d}^{3}x, \tag*{(1)}

where d3x=dx1dx2dx3\mathrm{d}^{3}x=\mathrm{d}x_1\mathrm{d}x_2\mathrm{d}x_3 denotes the volume element and ψ\psi is the density of the Helmholtz free energy. The dislocation network is assumed to be two-dimensional, such that the dislocation lines are parallel to the x3x_3-axis and the Burgers’ vector of edge dislocations lies in the (x1,x2)(x_1,x_2)-plane. We let ρr\rho^{\mathrm{r}} denote the density of redundant dislocations whose resultant Burgers’ vector vanishes, and ρg\rho^{\mathrm{g}} the density of non-redundant dislocations (see subsection 2.1). The sum of the two types of dislocation densities is the total dislocation density ρ=ρr+ρg\rho=\rho^{\mathrm{r}}+\rho^{\mathrm{g}}. In the spirit of the dislocation mediated plasticity [22], which decomposes the thermodynamic system of dislocated crystal into two subsystems and traces it back to statistical aspects of the defects, χ\chi is the effective temperature characterizing the slow rearrangement of configuration of atoms during the motion of dislocations, while the ordinary temperature TT characterizes the fast vibration of atoms of the kinetic-vibrational subsystem. In some cases, such as thermal softening [23, 33, 30] or adiabatic shear banding [34], the ordinary temperature plays an important role. However, in the present study we set TT to be constant and drop it from the list of arguments of the free energy density ψ\psi. The state variables εe\boldsymbol{\varepsilon}^{e} and ρg\rho^{\mathrm{g}} are dependent variables expressed by u\mathbf{u} and β\beta, while ρr\rho^{\mathrm{r}} and χ\chi are independent state variables. We decompose ψ\psi into the part due to the elastic strain, ψe\psi_{\mathrm{e}}, and the remaining parts

ψ(εe,ρr,ρg,χ)=ψe(εe)+ψr(ρr)+ψm(ρg)+ψc(χ,ρ),\psi(\boldsymbol{\varepsilon}^{\mathrm{e}},\rho^{\mathrm{r}},\rho^{\mathrm{g}},\chi)=\psi_{\mathrm{e}}(\boldsymbol{\varepsilon}^{\mathrm{e}})+\psi_{\mathrm{r}}(\rho^{\mathrm{r}})+\psi_{\mathrm{m}}(\rho^{\mathrm{g}})+\psi_{\mathrm{c}}(\chi,\rho),

where ψr\psi_{\mathrm{r}} is the self-energy density of redundant dislocations, ψm\psi_{\mathrm{m}} the energy density of the non-redundant dislocations, while ψc\psi_{\mathrm{c}} the configurational heat [26]. Their expressions are given below

ψe(εe)=12λ(εkke)2+μεijeεije,ψr(ρr)=γDρr,ψm(ρg)=γDρssgln⁡(11−ρgρssg),ψc(χ,ρ)=−χˉ(−ρln⁡(a2ρ)+ρ).\begin{aligned} \psi_{\mathrm{e}}(\boldsymbol{\varepsilon}^{\mathrm{e}})&=\frac{1}{2}\lambda(\varepsilon_{kk}^{\mathrm{e}})^2+\mu\varepsilon_{ij}^{\mathrm{e}}\varepsilon_{ij}^{\mathrm{e}}, \qquad\psi_{\mathrm{r}}(\rho^{\mathrm{r}})=\gamma_{\mathrm{D}}\rho^{\mathrm{r}},\\ \psi_{\mathrm{m}}(\rho^{\mathrm{g}})&=\gamma_{\mathrm{D}}\rho_{\mathrm{ss}}^{\mathrm{g}}\ln\left(\frac{1}{1-\frac{\rho^{\mathrm{g}}}{\rho_{\mathrm{ss}}^{\mathrm{g}}}}\right),\\ \psi_{\mathrm{c}}(\chi,\rho)&=-\bar{\chi}(-\rho\ln(a^2\rho)+\rho). \end{aligned}

where λ\lambda and μ\mu are Lamé constants, and γD\gamma_{\mathrm{D}} the dislocation energy per unit length. Since −ρln⁡(a2ρ)+ρ-\rho\ln(a^2\rho)+\rho is the entropy per unit area of the two-dimensional dislocation network (see the detailed derivation in [9]), the “two-dimensional” temperature χˉ=χ/L\bar{\chi}=\chi/L is introduced to get the configurational heat having the unit of energy density [26]. Note that aa is a length scale of the order of atomic spacing. The defect energy ψm\psi_{\mathrm{m}} describing the interactions of non-redundant dislocations captures the kinematic hardening. The logarithmic term, originated from [6], ensures that the defect energy increases linearly for small dislocation density and tends to infinity as ρg\rho^{\mathrm{g}} approaches the saturated density of non-redundant dislocations ρssg\rho_{\mathrm{ss}}^{\mathrm{g}}. We assume that ρssg=k0ρss\rho_{\mathrm{ss}}^{\mathrm{g}}=k_0\rho_{\mathrm{ss}} is a fraction of ρss\rho_{\mathrm{ss}} by coefficient k0k_0, where ρss=(1/a2)e−γD/χˉ\rho_{\mathrm{ss}}=(1/a^2)e^{-\gamma_{\mathrm{D}}/\bar{\chi}} is the steady-state dislocation density determined by minimizing the free energy of the configurational subsystem [22]. Note that k0k_0 is characterized by the interaction between dislocations of the same sign, which is similar to the interaction between charges [16]. We will see in the subsequent Section the role of ρssg\rho_{\mathrm{ss}}^{\mathrm{g}} in the contribution of interfacial energy to the work hardening.

The applied loading can lead to nucleation, propagation and movement of dislocations, which result in plastic deformations in the crystal. In these processes, dislocations always experience resistance, which causes energy dissipation. The dissipation potential in the bulk has the form

Db(β˙,ρ˙,χ˙)=τiβ˙+12dρρ˙2+12dχχ˙2.(2)D_{\mathrm{b}}(\dot{\beta},\dot{\rho},\dot{\chi})=\tau_{\mathrm{i}}\dot{\beta}+\frac{1}{2}d_{\rho}\dot{\rho}^{2}+\frac{1}{2}d_{\chi}\dot{\chi}^{2}. \tag*{(2)}

The first term is the plastic power, where τi\tau_{\mathrm{i}} is the internal stress, while the last two terms represent the dissipation due to the multiplication of dislocations and the increase of the configurational temperature [26]. Langer et al. [22] developed the kinetics of dislocation depinning that dominantly controls the plastic slip rate. Averaging this kinetic equation over the representative volume element, we have established in [35] that

β˙=q(T,τi,ρr)t0=bt0ρ[fP(T,τi,ρr)−fP(T,−τi,ρr)]\dot{\beta}=\frac{q(T,\tau_i,\rho^{r})}{t_0}=\frac{b}{t_0}\sqrt{\rho}\left[f_{\mathrm{P}}(T,\tau_i,\rho^{r})-f_{\mathrm{P}}(T,-\tau_i,\rho^{r})\right]

where t0t_0 is a microscopic time scale of the order of the inverse Debye frequency, and

fP(T,τi,ρr)=exp⁡[−TPTexp⁡(−τiτT(ρr))].f_{\mathrm{P}}(T,\tau_i,\rho^{r})=\exp\left[-\frac{T_{\mathrm{P}}}{T}\exp\left(-\frac{\tau_i}{\tau_{\mathrm{T}}(\rho^{r})}\right)\right].

Note that TPT_{\mathrm{P}} is the activation temperature of the potential well of a pinning site and τT(ρr)=μTbρr\tau_{\mathrm{T}}(\rho^{r})=\mu_{\mathrm{T}}b\sqrt{\rho^{r}} is the Taylor stress with μT\mu_{\mathrm{T}} being proportional to the shear modulus μ\mu. It was shown that this kinetic equation for β˙\dot{\beta} can correctly describe the rate-reducing behavior of isotropic work hardening both when the crystal is loaded on one direction and in the load reversal process [38]. Provided there is only one slip system inclined at an angle ϕ\phi to the x1x_1-axis, then, using similar arguments as in [17], we lay down one governing equation for τ˙i\dot{\tau}_i

τ˙i=μq0t0(cos⁡2ϕ−q(T,τi,ρr)q0)(3)\dot{\tau}_i=\mu\frac{q_0}{t_0}\left(\cos2\phi-\frac{q(T,\tau_i,\rho^{r})}{q_0}\right) \tag*{(3)}

where q0/t0q_0/t_0 denotes the total shear strain rate. This equation plays the role of the kinematic constraint for the internal stress τi\tau_i. The remaining equations in the bulk can be derived from the following variational principle: the true displacement field u˘(x)\breve{u}(\mathbf{x}), the true plastic slips β˘(x)\breve{\beta}(\mathbf{x}), the true density of redundant dislocations ρ˘r(x)\breve{\rho}^{r}(\mathbf{x}), and the true configurational temperature χ˘(x)\breve{\chi}(\mathbf{x}) obey the variational equation

δI+∫V(∂Db∂β˙δβ+∂Db∂ρ˙δρ+∂Db∂χ˙δχ)d3x=0(4)\delta I+\int_{\mathcal{V}}\left(\frac{\partial D_b}{\partial\dot{\beta}}\delta\beta+\frac{\partial D_b}{\partial\dot{\rho}}\delta\rho+\frac{\partial D_b}{\partial\dot{\chi}}\delta\chi\right)\mathrm{d}^{3}x=0 \tag*{(4)}

for all variations of admissible fields u(x)\mathbf{u}(\mathbf{x}), β(x)\beta(\mathbf{x}), ρr(x)\rho^{r}(\mathbf{x}) and χ(x)\chi(\mathbf{x}). Focusing on the contribution of the grain boundary to work hardening, we assume that the whole external boundary ∂V\partial\mathcal{V} admit dislocation to reach it. In this case the plastic slip can be varied arbitrarily at ∂V\partial\mathcal{V}. The variational equation (4) allows one to derive the remaining governing equations. From the variation of II with respect to the displacement field u(x)\mathbf{u}(\mathbf{x}), we obtain the quasi-static equilibrium equation and the boundary condition [31]

σij,j=0in V,\sigma_{ij,j}=0 \quad\text{in } \mathcal{V},
σijnj=0on ∂s,\sigma_{ij}n_j=0 \quad\text{on } \partial_s,

with σ=∂ψe/∂εe\boldsymbol{\sigma}=\partial\psi_e/\partial\boldsymbol{\varepsilon}^e being the Cauchy stress tensor, and n\mathbf{n} the outward unit normal vector to the external boundary. In Eq. (4) the vanishing variation with respect to the plastic slip β(x)\beta(\mathbf{x}) yields the balance of microforces acting on dislocations [26] and the natural boundary condition at the entire external boundary as

τ−τb−τi=0in V,\tau-\tau_b-\tau_i=0 \quad\text{in } \mathcal{V},
∂ψm∂ρσ=γDon ∂V,\frac{\partial\psi_m}{\partial\rho^\sigma}=\gamma_D \quad\text{on } \partial\mathcal{V},

where τ=siσijmj\tau=s_i\sigma_{ij}m_j is the resolved shear stress acting on the slip system, while the back stress is

τb=−1b(∂ψm∂ρσsign⁡β,s),s=−1b2∂2ψm∂(ρσ)2β,ss.\tau_b=-\frac{1}{b}\left(\frac{\partial\psi_m}{\partial\rho^\sigma}\operatorname{sign}\beta_{,s}\right)_{,s}=-\frac{1}{b^2}\frac{\partial^2\psi_m}{\partial(\rho^\sigma)^2}\beta_{,ss}.

Note that (⋅),s=si∂i(⋅)(\cdot)_{,s}=s_i\partial_i(\cdot) is the short-hand notation for ∂s(⋅)\partial_s(\cdot). Following the suggestions by [26] for functions of ρp\rho^p and χ\chi, the remaining equations obtained from Eq. (4) for χ˙\dot{\chi} and ρ˙\dot{\rho} can be cast into the form

χ˙=KχϵDτiq(T,τi,ρr)t0(1−χχ0),ρ˙=Kρτia2ν(T,ρr,q0)2q(T,τi,ρr)t0(1−ρρss(χ)).(5)\begin{aligned} \dot{\chi}&=\mathcal{K}_{\chi}\epsilon_D\tau_i\frac{q(T,\tau_i,\rho^r)}{t_0}\left(1-\frac{\chi}{\chi_0}\right),\\ \dot{\rho}&=\mathcal{K}_{\rho}\frac{\tau_i}{a^2\nu(T,\rho^r,q_0)^2}\frac{q(T,\tau_i,\rho^r)}{t_0}\left(1-\frac{\rho}{\rho_{\mathrm{ss}}(\chi)}\right). \tag*{(5)} \end{aligned}

Here, χ0\chi_0 is a constant denoting the steady-state configurational temperature, while the steady-state dislocation density at a given effective temperature is ρss(χ)=(1/a2)e−γD/χˉ\rho_{\mathrm{ss}}(\chi)=(1/a^2)e^{-\gamma_D/\bar{\chi}}. Kχ\mathcal{K}_{\chi} is a factor inversely proportional to the effective specific heat and Kρ\mathcal{K}_{\rho} is an energy-conversion coefficient [24]. The function ν(T,ρr,q0)\nu(T,\rho^r,q_0) is defined as

ν(T,ρr,q0)=ln⁡(TPT)−ln⁡[ln⁡(bρrq0)].\nu(T,\rho^r,q_0)=\ln\left(\frac{T_P}{T}\right)-\ln\left[\ln\left(\frac{b\sqrt{\rho^r}}{q_0}\right)\right].

Interface boundary conditions

As a prototype, we study a simple problem of plane deformation of a bicrystal whose single grain boundary, which is a plane perpendicular to the yy-axis, lies at y=ygy=y_g (see Figure 1). The extension to the case of polycrystals with multiple grain boundaries can be done without complications. Since the tensorial index notation is no longer required, the coordinates (x1,x2,x3)(x_1,x_2,x_3) are changed to (x,y,z)(x,y,z) for simplicity. For the sake of definiteness, we assume that only edge dislocations move to and interact with the grain boundary. If the plastic slip β\beta depends only on yy due to the sample geometry and specific boundary conditions, then the non-redundant dislocation density is proportional to β,y\beta_{,y}, ρg=χ∣β,y∣/b\rho^g=\chi|\beta_{,y}|/b, where χ\chi depends on the slip system, as will be discussed in Section 4.

Plane constrained shear of bicrystal

Figure 1. (Color online) Plane constrained shear of bicrystal

Interfacial energy and dissipation

Dislocations are long-lived and well-defined defects and have various interactions with the grain boundary, e.g., dislocation pile-up, absorption, and transmission. Within the continuum approach, it should be possible to mimic these interactions by the interfacial energy and dissipation that reflect the underlying physics at the nano/submicron scale. The low-angle grain boundary itself can be regarded as an array of dislocations [8] whose interfacial energy depends on the surface dislocation density [41, 43]. It also acts as an additional barrier to the movement of dislocations and can later absorb or release them. Therefore, in addition to bulk dissipation, interfacial dissipation must also be taken into account. We propose the densities of interfacial energy and dissipation potential as follows

ψs(β∣yg±0)=ψs(∣[ ⁣[β] ⁣]Γ∣)(6)\psi_s(\beta|_{y_g\pm0})=\psi_s(|[\![\beta]\!]_\Gamma|) \tag*{(6)}

and

Ds(β˙∣yg±0)=ζy∣⟨⟨β˙⟩⟩Γ∣b+12ζa([ ⁣[β˙] ⁣]Γ)2b2.(7)D_s(\dot{\beta}|_{y_g\pm0})=\zeta_y\frac{|\langle\langle\dot{\beta}\rangle\rangle_\Gamma|}{b}+\frac{1}{2}\zeta_a\frac{([\![\dot{\beta}]\!]_\Gamma)^2}{b^2}. \tag*{(7)}

Here, the vertical line followed by yg±0y_g\pm0 indicates the limits of the preceding expression as yy approaches the grain boundary position ygy_g from above and below, while ⟨⟨⋅⟩⟩Γ\langle\langle\cdot\rangle\rangle_\Gamma and [ ⁣[⋅] ⁣]Γ[\![\cdot]\!]_\Gamma denote the mean-value and the jump of a variable at the interface Γ\Gamma, respectively. Thus, according to these definitions

⟨⟨β˙⟩⟩Γ=12(β˙∣yg+0+β˙∣yg−0),[ ⁣[β˙] ⁣]Γ=(β˙∣yg+0−β˙∣yg−0).\langle\langle\dot{\beta}\rangle\rangle_\Gamma=\frac{1}{2}\left(\dot{\beta}|_{y_g+0}+\dot{\beta}|_{y_g-0}\right),\qquad[\![\dot{\beta}]\!]_\Gamma=\left(\dot{\beta}|_{y_g+0}-\dot{\beta}|_{y_g-0}\right).

Eq. (6) requires that only the surface dislocation density ∣⟦β⟧Γ∣/b\lvert\llbracket\beta\rrbracket_{\Gamma}\rvert/b, regarded as the state variable, can be the argument of ψs\psi_{\mathrm{s}}. The first term on the right-hand side of Eq. (7) is the dissipation by the rate of the mean plastic slip at the interface, where ζy\zeta_{\mathrm{y}} plays a role of the yield surface tension. The second term in (7), quadratic in ∣β˙∣Γ\lvert\dot{\beta}\rvert_{\Gamma}, describes rate-dependent dissipation due to the acoustic emission of waves generated during the absorption of dislocations by the grain boundary.

The boundary conditions involving interfacial energy and dissipation can be derived from the following variational equation

δ(I+∫Γψs(∣⟦β⟧Γ∣) da)+∫V(∂Db∂β˙δβ˙+∂Db∂ρ˙δρ˙+∂Db∂χ˙δχ˙) d3x+∫Γ(∂Ds∂∣β˙∣Γδ∣β˙∣Γ+∂Ds∂⟨β˙⟩Γδ⟨β˙⟩Γ) da=0,(8)\delta\left(I+\int_{\Gamma}\psi_{\mathrm{s}}\left(\lvert\llbracket\beta\rrbracket_{\Gamma}\rvert\right)\,\mathrm{d}a\right)+\int_{\mathcal{V}}\left(\frac{\partial D_{\mathrm{b}}}{\partial\dot{\beta}}\delta\dot{\beta}+\frac{\partial D_{\mathrm{b}}}{\partial\dot{\rho}}\delta\dot{\rho}+\frac{\partial D_{\mathrm{b}}}{\partial\dot{\chi}}\delta\dot{\chi}\right)\,\mathrm{d}^{3}x+\int_{\Gamma}\left(\frac{\partial D_{\mathrm{s}}}{\partial\lvert\dot{\beta}\rvert_{\Gamma}}\delta\lvert\dot{\beta}\rvert_{\Gamma}+\frac{\partial D_{\mathrm{s}}}{\partial\langle\dot{\beta}\rangle_{\Gamma}}\delta\langle\dot{\beta}\rangle_{\Gamma}\right)\,\mathrm{d}a=0, \tag*{(8)}

which extends Eq. (4) to the case involving the grain boundary, where da\mathrm{d}a denotes the area element, while II and DbD_{\mathrm{b}} remain exactly the same as in (1) and (2), respectively. Edge dislocations may either penetrate the grain boundary or be absorbed by it through dissociation of the dislocations. The latter process requires the continuity of displacement field but causes a jump in plastic slip ⟦β⟧Γ\llbracket\beta\rrbracket_{\Gamma} at the interface. This is revealed by the volume integral containing the variation of the non-redundant dislocation density δρg\delta\rho^{\mathrm{g}} during the integration by parts, permitting the discontinuity to enter into the interface as

∫V(∂ψm∂ρg+∂ψχ∂ρg+∂Db∂ρ˙g)δρg d3x=−∫Vχb(∂ψm∂ρg−γD)sign⁡β,yδβ d3x+∫∂sχb(∂ψm∂ρg−γD)sign⁡β,yδβ da−∫Γ[χb(∂ψm∂ρg−γD)sign⁡β,yδβ]Γ da.(9)\int_{\mathcal{V}}\left(\frac{\partial\psi_{\mathrm{m}}}{\partial\rho^{\mathrm{g}}}+\frac{\partial\psi_{\chi}}{\partial\rho^{\mathrm{g}}}+\frac{\partial D_{\mathrm{b}}}{\partial\dot{\rho}^{\mathrm{g}}}\right)\delta\rho^{\mathrm{g}}\,\mathrm{d}^{3}x=-\int_{\mathcal{V}}\frac{\chi}{b}\left(\frac{\partial\psi_{\mathrm{m}}}{\partial\rho^{\mathrm{g}}}-\gamma_{\mathrm{D}}\right)\operatorname{sign}\beta_{,y}\delta\beta\,\mathrm{d}^{3}x+\int_{\partial_{\mathrm{s}}}\frac{\chi}{b}\left(\frac{\partial\psi_{\mathrm{m}}}{\partial\rho^{\mathrm{g}}}-\gamma_{\mathrm{D}}\right)\operatorname{sign}\beta_{,y}\delta\beta\,\mathrm{d}a-\int_{\Gamma}\left[\frac{\chi}{b}\left(\frac{\partial\psi_{\mathrm{m}}}{\partial\rho^{\mathrm{g}}}-\gamma_{\mathrm{D}}\right)\operatorname{sign}\beta_{,y}\delta\beta\right]_{\Gamma}\,\mathrm{d}a. \tag*{(9)}

The governing equation with respect to ρg\rho^{\mathrm{g}} from Eq. (4), given as [26, 27]

γD+χln⁡(a2ρ)/L+dρρ˙=0,\gamma_{\mathrm{D}}+\chi\ln\left(a^{2}\rho\right)/L+d_{\rho}\dot{\rho}=0,

is utilized for γD\gamma_{\mathrm{D}} in the derivation of Eq. (9). Expanding the interface integral term on the right hand side of Eq. (9) by means of the jump identity [1]

⟦pq⟧Γ=⟦p⟧Γ⟨ ⁣⟨q⟩ ⁣⟩Γ+⟦q⟧Γ⟨ ⁣⟨p⟩ ⁣⟩Γ\llbracket p q\rrbracket_{\Gamma}=\llbracket p\rrbracket_{\Gamma}\langle\!\langle q\rangle\!\rangle_{\Gamma}+\llbracket q\rrbracket_{\Gamma}\langle\!\langle p\rangle\!\rangle_{\Gamma}

and substituting it into Eq. (8), we transform the latter to

\int_{\Gamma}\left\{-\left\langle\left\langle\frac{\varkappa}{b}\left(\frac{\partial\psi_m}{\partial\rho^g}-\gamma_D\right)\operatorname{sign}\beta_{,y}\right\rangle\right\rangle_{\Gamma}\delta\llbracket\beta\rrbracket_{\Gamma}-\left\llbracket\frac{\varkappa}{b}\left(\frac{\partial\psi_m}{\partial\rho^g}-\gamma_D\right)\operatorname{sign}\beta_{,y}\right\rrbracket_{\Gamma}\delta\langle\langle\beta\rangle\rangle_{\Gamma}+\frac{\partial\psi_s}{\partial\lvert\beta\rvert_{\Gamma}}\delta\lvert\beta\rvert_{\Gamma}+\zeta_y\frac{1}{b}\operatorname{sign}\langle\langle\dot{\beta}\rangle\rangle_{\Gamma}\delta\langle\langle\beta\rangle\rangle_{\Gamma}+\zeta_a\frac{1}{b^2}\llbracket\dot{\beta}\rrbracket_{\Gamma}\delta\llbracket\beta\rrbracket_{\Gamma}\right\}\,da=0. \tag*{(10)}

To derive consequences from (10), we need to analyze the variations of plastic slip on both sides of the grain boundary, which may be subject to constraints depending on the accumulated dislocation densities. We therefore consider two different processes.

Pile-up process: When dislocations move toward the grain boundary under the Peach-Koehler force, they first pile up against it. Since the dislocations cannot initially penetrate the grain boundary acting as an obstacle, the plastic slips on both sides of the grain boundary are subject to homogeneous Dirichlet boundary condition which fulfills Eq. (10) due to δβ∣yg−0=δβ∣yg+0=0\delta\beta\lvert_{y_g-0}=\delta\beta\lvert_{y_g+0}=0. So we have during the pile-up process

β˙(yg±0,t)=0.(11)\dot{\beta}(y_g\pm0,t)=0. \tag*{(11)}

The time derivative in Eq. (11) ensures the continuity of plastic slip at the interface with respect to time in the reversal loading [38].

Absorption and transmission processes: As more dislocations of the same sign pile up against the interface, the back stress near the pile-up sites is increased. When the accumulated non-redundant dislocation density and with it the back stress in the vicinity of the grain boundary exceeds a threshold, the latter can no longer block dislocations. When the grain boundary starts to absorb dislocations, the jump in plastic slip as well as the misorientation between two adjacent grains will increase. The plastic slip at the grain boundary as well as its jump ⟦β⟧Γ\llbracket\beta\rrbracket_{\Gamma} may change during the process of dislocation absorption, and consequently, δ⟨⟨β⟩⟩Γ\delta\langle\langle\beta\rangle\rangle_{\Gamma} and δ⟦β⟧Γ\delta\llbracket\beta\rrbracket_{\Gamma} can be chosen independently and arbitrarily. Eq. (10) then implies

ϰ(∂ψm∂ρg−γD)∣yg+0+ϰ(∂ψm∂ρg−γD)∣yg−0=ζy,(12)\varkappa\left(\frac{\partial\psi_m}{\partial\rho^g}-\gamma_D\right)\bigg|_{y_g+0}+\varkappa\left(\frac{\partial\psi_m}{\partial\rho^g}-\gamma_D\right)\bigg|_{y_g-0}=\zeta_y, \tag*{(12)}

and

\llbracket\dot{\beta}\rrbracket_{\Gamma}=\frac{b^2}{\zeta_a}\left(\left\llbracket\frac{\varkappa}{2b}\frac{\partial\psi_m}{\partial\rho^g}\right\rrbracket_{\Gamma}-\frac{\partial\psi_s}{\partial\lvert\beta\rvert_{\Gamma}}\right). \tag*{(13)}

provided sign⁡β,y∣yg±0=±1\operatorname{sign}_{\beta,y}|_{y_g\pm0}=\pm1, sign⁡⟨⟨β˙⟩⟩Γ=1\operatorname{sign}\langle\langle\dot{\beta}\rangle\rangle_{\Gamma}=1 and sign⁡[[β˙]]Γ=1\operatorname{sign}[[\dot{\beta}]]_{\Gamma}=1.

For simplicity let us assume the symmetry such that κl=κu=κ\kappa_l=\kappa_u=\kappa. Let ρy\rho_y be the root of the equation ∂ψm/∂ρg=ζy/2κ+γD\partial\psi_m/\partial\rho^g=\zeta_y/2\kappa+\gamma_D which exists and is unique due to the monotonicity of function ∂ψm/∂ρg\partial\psi_m/\partial\rho^g. It turns out that this dislocation density plays the role of the first threshold. If the dislocation density on one side of the grain boundary exceeds this threshold value, (12) can only be satisfied if

ρg∣yg+0=κ∣β,y∣/b∣yg+0=ρy+ρj,ρg∣yg−0=κ∣β,y∣/b∣yg−0=ρy−ρj.(14)\begin{aligned} \rho^g|_{y_g+0}&=\kappa|\beta_{,y}|/b|_{y_g+0}=\rho_y+\rho_j,\\ \rho^g|_{y_g-0}&=\kappa|\beta_{,y}|/b|_{y_g-0}=\rho_y-\rho_j. \tag*{(14)} \end{aligned}

Indeed, taking into account that ρj\rho_j is much less than ρy\rho_y, we can expand function ∂ψm/∂ρg\partial\psi_m/\partial\rho^g into the Taylor series near ρy\rho_y and neglect terms of order higher than ρj\rho_j. Then it is easy to see that (14) satisfies (12).

But there is another threshold that does not permit dislocations to reach the grain boundary even if the dislocation density on one side exceeds ρy\rho_y. The reason lies in the second term on the right-hand side of (13). Since ψs\psi_s is a function of ∣[[β]]Γ∣|[[\beta]]_{\Gamma}|, its derivative is equal to

∂ψs∂∣[[β]]Γ∣=∂ψs∂[[β]]Γsign⁡[[β]]Γ.\frac{\partial\psi_s}{\partial|[[\beta]]_{\Gamma}|} = \frac{\partial\psi_s}{\partial[[\beta]]_{\Gamma}} \operatorname{sign}[[\beta]]_{\Gamma}.

Because the largest absolute value of this derivative is achieved at [[β]]Γ=ϑ0[[\beta]]_{\Gamma}=\vartheta_0, with ϑ0\vartheta_0 being the initial misalignment, as long as the first term in the bracket on the right-hand side is less than this value, [[β]]Γ[[\beta]]_{\Gamma} cannot increase. To simplify (13) we take into account (14) and the smallness of ρj\rho_j. Expanding ∂ψm/∂ρg\partial\psi_m/\partial\rho^g into the Taylor series near ρy\rho_y we transform (13) to

[[β˙]]Γ=b2ca(κb∂2ψm∂(ρg)2(ρy)ρj−∂ψs∂∣[[β]]Γ∣sign⁡[[β]]Γ).(15)[[\dot{\beta}]]_{\Gamma} = \frac{b^2}{c_a} \left( \frac{\kappa}{b} \frac{\partial^2\psi_m}{\partial(\rho^g)^2}(\rho_y)\rho_j - \frac{\partial\psi_s}{\partial|[[\beta]]_{\Gamma}|} \operatorname{sign}[[\beta]]_{\Gamma} \right). \tag*{(15)}

For the density of interfacial energy we will consider the modified Read-Shockley’s interfacial energy [41],

ψs=γD4π(1−ν)bϑln⁡eϑmϑ+δ,(16)\psi_s= \frac{\gamma_D}{4\pi(1-\nu)b} \vartheta\ln\frac{e\vartheta_m}{\vartheta+\delta}, \tag*{(16)}

where ϑ=∣[[β]]Γ∣\vartheta=|[[\beta]]_{\Gamma}|. A small number δ\delta is added to the denominator of the logarithm to make the derivative of ψs\psi_s finite at ϑ=0\vartheta=0. Finally, we need an equation for ρy\rho_y which is similar to (3) for τi\tau_i. As such we propose

ρy=ρy0+κa∣[[β]]Γ∣.(17)\rho_y=\rho_{y0}+\kappa_a|[[\beta]]_{\Gamma}|. \tag*{(17)}

Application

Plane constrained shear

Suppose that a bicrystal slab is subjected to plane constrained shear by a given displacement uy(t)=γ˙tyu_y(t)=\dot{\gamma}ty specified at the upper and lower surfaces (see Figure 1) and that the system is driven at a constant shear rate γ˙=q0/t0\dot{\gamma}=q_0/t_0. Let the green colored plane Γ\Gamma be the plane of the grain boundary that is parallel to the (x,y)(x,y)-plane and is located at y=ygy=y_g. This grain boundary divides the bicrystal occupying V\mathcal{V} into two perfectly bonded single crystals occupying V−\mathcal{V}^{-} and V+\mathcal{V}^{+}, i.e. V=V−∪V+∪Γ\mathcal{V}=\mathcal{V}^{-}\cup\mathcal{V}^{+}\cup\Gamma. Let the cross section of this bicrystal perpendicular to the zz-axis be a rectangle of width cc and height hh, having the same shape and size over the entire length LL. We assume h≪c≪Lh\ll c\ll L to neglect the end effect near x=0x=0 and x=cx=c. Edge dislocations can occur during the plastic deformation, and we assume that only one slip system is activated in each single crystal, colored blue and red, respectively. The slip directions are perpendicular to the zz axis and inclined at angles φu\varphi_u and φl\varphi_l to the plane of the grain boundary. Given the geometry of the specimen as well as boundary conditions, we assume that the displacements (ux,uy)(u_x,u_y) as well as the plastic slip β\beta depends only on yy: ux=ux(y)u_x=u_x(y), uy=uy(y)u_y=u_y(y), β=β(y)\beta=\beta(y). The displacement uzu_z vanishes.

When the plastic deformation occurs, the edge dislocations move on the active slip system with the slip direction s=(cos⁡ϕ,sin⁡ϕ,0)T\mathbf{s}=(\cos\phi,\sin\phi,0)^{T} and the normal to the slip plane m=(−sin⁡ϕ,cos⁡ϕ,0)T\mathbf{m}=(-\sin\phi,\cos\phi,0)^{\mathrm{T}}. The plastic distortion tensor is given by

β=β(y,t)(−sin⁡ϕcos⁡ϕcos⁡2ϕ0−sin⁡2ϕsin⁡ϕcos⁡ϕ0000)\boldsymbol{\beta}=\beta(y,t)\begin{pmatrix}-\sin\phi\cos\phi& \cos^{2}\phi& 0\\-\sin^{2}\phi& \sin\phi\cos\phi& 0\\0 & 0 & 0\end{pmatrix}

The plastic and elastic strain tensor becomes

εp=12β(y,t)(−sin⁡2ϕcos⁡2ϕ0cos⁡2ϕsin⁡2ϕ0000),\boldsymbol{\varepsilon}^{\mathrm{p}}=\frac{1}{2}\beta(y,t)\begin{pmatrix}-\sin2\phi& \cos2\phi& 0\\\cos2\phi& \sin2\phi& 0\\0 & 0 & 0\end{pmatrix},
εe=12(βsin⁡2ϕux,y−βcos⁡2ϕ0ux,y−βcos⁡2ϕ2uy,y−βsin⁡2ϕ0000),\boldsymbol{\varepsilon}^{\mathrm{e}}=\frac{1}{2}\begin{pmatrix}\beta\sin2\phi& u_{x,y}-\beta\cos2\phi& 0\\u_{x,y}-\beta\cos2\phi& 2u_{y,y}-\beta\sin2\phi& 0\\0 & 0 & 0\end{pmatrix},

where the displacement field u=(ux,uy,0)\mathbf{u}=(u_x,u_y,0) is yet unknown. The non-zero components of Nye-Bilby-Kröner’s tensor are αxz=β,ysin⁡ϕcos⁡ϕ\alpha_{xz}=\beta_{,y}\sin\phi\cos\phi and αyz=β,ysin⁡2ϕ\alpha_{yz}=\beta_{,y}\sin^{2}\phi. Therefore the density of the non-redundant dislocation per unit area perpendicular to the zz-axis equals

ρg=1b∣α⋅ez∣=κ∣β,y∣b,where κ=∣sin⁡ϕ∣.\rho^{g}=\frac{1}{b}|\boldsymbol{\alpha}\cdot\mathbf{e}_z|=\frac{\kappa|\beta_{,y}|}{b},\qquad\text{where }\kappa=|\sin\phi|.

The total energy functional of a crystal becomes

I=cL∫0h[ψe(εe)+γDρr+ψm(ρg)+ψc]dy+cLψi∣y=yg.(18)I=cL\int_{0}^{h}\left[\psi_{\mathrm{e}}(\boldsymbol{\varepsilon}^{\mathrm{e}})+\gamma_{\mathrm{D}}\rho^{r}+\psi_{\mathrm{m}}(\rho^{g})+\psi_{\mathrm{c}}\right]\mathrm{d}y+cL\psi_{\mathrm{i}}\big|_{y=y_g}. \tag*{(18)}

where ψe\psi_{\mathrm{e}} is the energy density due to the elastic strain

ψe=12λuy,y2+12μ(ux,y−βcos⁡2ϕ)2+14μβ2sin⁡22ϕ+μ(uy,y−12βsin⁡2ϕ)2.\psi_{\mathrm{e}}=\frac{1}{2}\lambda u_{y,y}^{2}+\frac{1}{2}\mu(u_{x,y}-\beta\cos2\phi)^{2}+\frac{1}{4}\mu\beta^{2}\sin^{2}2\phi+\mu\left(u_{y,y}-\frac{1}{2}\beta\sin2\phi\right)^{2}.

Note that the angle ϕ(y)\phi(y) is piecewise constant:

ϕ(y)={ϕlfor 0<y<yg,ϕufor yg<y<h.\phi(y)=\begin{cases}\phi_l & \text{for }0<y<y_g,\\\phi_u & \text{for }y_g<y<h.\end{cases}

Taking the variation of functional II with respect to uxu_x and uyu_y at fixed β(y)\beta(y) we obtain the equilibrium equations which, after integration, lead to

ux,y=γ+(β−⟨β⟩)cos⁡2ϕ,u_{x,y}=\gamma+(\beta-\langle\beta\rangle)\cos2\phi,
uy,y=κ(β−⟨β⟩)sin⁡2ϕ.(19)u_{y,y}=\kappa(\beta-\langle\beta\rangle)\sin2\phi. \tag*{(19)}

with κ=μλ+2μ\kappa= \frac{\mu}{\lambda+2\mu} and ⟨β⟩=1h∫0hβdy\langle\beta\rangle= \frac{1}{h}\int_{0}^{h}\beta dy. Inserting Eq. (19) into Eq. (18) we reduce the energy functional to

I=cL\int_{0}^{h}\left[\frac{1}{2}\mu\kappa\langle\beta\rangle^{2}\sin^{2}2\phi+\frac{1}{2}\mu\left(\langle\beta\rangle\cos2\phi-\gamma\right)^{2}+\frac{1}{2}\mu(1-\kappa)\beta^{2}\sin^{2}2\phi+\gamma_{\mathrm{D}}\rho^{\mathrm{r}}+\gamma_{\mathrm{D}}k_{0}\rho_{\mathrm{ss}}\ln\left(\frac{1}{1-\frac{\rho^{\mathrm{g}}}{k_{0}\rho_{\mathrm{ss}}}\right)-\chi\left(-\rho\ln\left(a^{2}\rho\right)+\rho\right)/L\right]dy+cL\psi_{i}\bigg|_{y=y_{\mathrm{g}}}.}

Varying this functional with respect to β\beta and substituting into (8), we obtain the equilibrium equation

τ−τb−τi=0,(20)\tau-\tau_{\mathrm{b}}-\tau_{\mathrm{i}}=0, \tag*{(20)}

with the resolved and back shear stresses being

τ=−μ(κ⟨β⟩sin⁡22ϕ+(⟨β⟩cos⁡2ϕ−γ)cos⁡2ϕ+(1−κ)βsin⁡22ϕ),\tau=-\mu\left(\kappa\langle\beta\rangle\sin^{2}2\phi+\left(\langle\beta\rangle\cos2\phi-\gamma\right)\cos2\phi+(1-\kappa)\beta\sin^{2}2\phi\right),

and

τb=−C1(1−C2∣β,y∣)2β,yy,C1=γDk0ρssb2sin⁡2ϕ,C2=1k0ρssb∣sin⁡ϕ∣.\tau_{\mathrm{b}}=-\frac{C_{1}}{\left(1-C_{2}\lvert\beta_{,y}\rvert\right)^{2}}\beta_{,yy},\qquad C_{1}=\frac{\gamma_{\mathrm{D}}}{k_{0}\rho_{\mathrm{ss}}b^{2}}\sin^{2}\phi,\qquad C_{2}=\frac{1}{k_{0}\rho_{\mathrm{ss}}b}\lvert\sin\phi\rvert.

At the top and bottom surfaces of the sample, the boundary conditions ∂ψm/∂ρg=γD\partial\psi_{\mathrm{m}}/\partial\rho^{\mathrm{g}}=\gamma_{\mathrm{D}} must be fulfilled. It is obvious that these imply the Neuman conditions β,y(0,t)=β,y(h,t)=0\beta_{,y}(0,t)=\beta_{,y}(h,t)=0. Besides, at y=ygy=y_{\mathrm{g}}, the boundary conditions (13) (or (15), alternatively), (14), and (17) must be fulfilled.

Rescaled boundary-value problem

To facilitate numerical integration, we introduce the following rescaled variables and quantities

y~=yb,ρ~=a2ρ,ρ~r=a2ρr,ρ~g=b2ρg=κ∣β,y~∣,χ~=χˉγD,θ=TTP,\tilde{y}=\frac{y}{b},\qquad\tilde{\rho}=a^{2}\rho,\qquad\tilde{\rho}^{\mathrm{r}}=a^{2}\rho^{\mathrm{r}},\qquad\tilde{\rho}^{\mathrm{g}}=b^{2}\rho^{\mathrm{g}}=\kappa\lvert\beta_{,\tilde{y}}\rvert,\qquad\tilde{\chi}=\frac{\bar{\chi}}{\gamma_{\mathrm{D}}},\qquad\theta=\frac{T}{T_{\mathrm{P}}},
τ~=τμ,τ~i=τiμ,τ~b=τbμ,\tilde{\tau}=\frac{\tau}{\mu},\qquad\tilde{\tau}_{\mathrm{i}}=\frac{\tau_{\mathrm{i}}}{\mu},\qquad\tilde{\tau}_{\mathrm{b}}=\frac{\tau_{\mathrm{b}}}{\mu},

where the variable y~\tilde{y} changes from zero to h~=h/b\tilde{h}=h/b. The plastic slip rate is rewritten in the form

q(T,τi,ρr)t0=baq~(θ,τ~i,ρ~r)t0.\frac{q(T,\tau_{\mathrm{i}},\rho^{\mathrm{r}})}{t_{0}}=\frac{b}{a}\frac{\tilde{q}(\theta,\tilde{\tau}_{\mathrm{i}},\tilde{\rho}^{\mathrm{r}})}{t_{0}}.

in which

q~(θ,τ~i,ρ~r)=ρ~r(f~P(θ,τ~i,ρ~r)−f~P(θ,−τ~i,ρ~r)).\tilde{q}(\theta,\tilde{\tau}_i,\tilde{\rho}^{r})=\sqrt{\tilde{\rho}^{r}}\left(\tilde{f}_{P}(\theta,\tilde{\tau}_i,\tilde{\rho}^{r})-\tilde{f}_{P}(\theta,-\tilde{\tau}_i,\tilde{\rho}^{r})\right).

We set μ~T=(b/a)μT=μr\tilde{\mu}_{T}=(b/a)\mu_{T}=\mu_{r} and assume that rr is independent of temperature and strain rate. Then

f~P(θ,τ~i,ρ~r)=exp⁡(−1θexp⁡(−τ~irρ~r)).\tilde{f}_{P}(\theta,\tilde{\tau}_i,\tilde{\rho}^{r})=\exp\left(-\frac{1}{\theta}\exp\left(-\frac{\tilde{\tau}_i}{r\sqrt{\tilde{\rho}^{r}}}\right)\right).

We define q~0=(a/b)q0\tilde{q}_{0}=(a/b)q_{0} such that q/q0=q~/q~0q/q_{0}=\tilde{q}/\tilde{q}_{0}, and function ν~\tilde{\nu} becomes

ν~(θ,ρ~r,q~0)=ln⁡(1θ)−ln⁡(ln⁡(ρ~rq~0)).\tilde{\nu}(\theta,\tilde{\rho}^{r},\tilde{q}_{0})=\ln\left(\frac{1}{\theta}\right)-\ln\left(\ln\left(\frac{\sqrt{\tilde{\rho}^{r}}}{\tilde{q}_{0}}\right)\right).

The dimensionless steady-state quantities are

ρ~ss(χ~)=exp⁡(−1χ~),χ~0=χ0eD.\tilde{\rho}_{ss}(\tilde{\chi})=\exp\left(-\frac{1}{\tilde{\chi}}\right),\qquad\tilde{\chi}_{0}=\frac{\chi_{0}}{e_{D}}.

Using q~\tilde{q} instead of qq as the dimensionless measure of plastic strain rate means that we are effectively rescaling t0t_{0} by a factor b/ab/a. We assume that (a/b)t0=10−12 s(a/b)t_{0}=10^{-12}\,\mathrm{s}. Since the shear rate γ˙=q0/t0\dot{\gamma}=q_{0}/t_{0} is constant, we can replace the time derivative by the derivative with respect to γ\gamma so that t0∂/∂t→q0∂/∂γt_{0}\partial/\partial t\to q_{0}\partial/\partial\gamma. Using the introduced dimensionless variables the governing equations in the bulk, Eq. (3), Eq. (5) and Eq. (20), are transformed to the following system of partial differential equations

τ~−τ~b−τ~i=0,∂τ~i∂γ=cos⁡2ϕ−q~(θ,τ~i,ρ~r)q~0,∂χ~∂γ=Kχτ~iq~(θ,τ~i,ρ~r)q~0(1−χ~χ~0),∂ρ~∂γ=Kρτ~iq~(θ,τ~i,ρ~r)ν~(θ,ρ~r,q~0)2q~0(1−ρ~ρ~ss(χ~)),(21)\begin{aligned} \tilde{\tau}-\tilde{\tau}_{b}-\tilde{\tau}_{i}&=0,\\ \frac{\partial\tilde{\tau}_{i}}{\partial\gamma}&=\cos2\phi-\frac{\tilde{q}(\theta,\tilde{\tau}_{i},\tilde{\rho}^{r})}{\tilde{q}_{0}},\\ \frac{\partial\tilde{\chi}}{\partial\gamma}&=K_{\chi}\tilde{\tau}_{i}\frac{\tilde{q}(\theta,\tilde{\tau}_{i},\tilde{\rho}^{r})}{\tilde{q}_{0}}\left(1-\frac{\tilde{\chi}}{\tilde{\chi}_{0}}\right),\\ \frac{\partial\tilde{\rho}}{\partial\gamma}&=K_{\rho}\tilde{\tau}_{i}\frac{\tilde{q}(\theta,\tilde{\tau}_{i},\tilde{\rho}^{r})}{\tilde{\nu}(\theta,\tilde{\rho}^{r},\tilde{q}_{0})2\tilde{q}_{0}}\left(1-\frac{\tilde{\rho}}{\tilde{\rho}_{ss}(\tilde{\chi})}\right), \tag*{(21)} \end{aligned}

with the rescaled resolved shear stress and back stress being

τ~=−(κ⟨β⟩sin⁡22ϕ+(⟨β⟩cos⁡2ϕ−γ)cos⁡2ϕ+(1−κ)βsin⁡22ϕ),\tilde{\tau}=-\left(\kappa\langle\beta\rangle\sin^{2}2\phi+\left(\langle\beta\rangle\cos2\phi-\gamma\right)\cos2\phi+(1-\kappa)\beta\sin^{2}2\phi\right),
τ~b=−k1(1−k2∣β,y~∣)2β,y~y~.\tilde{\tau}_{b}=-\frac{k_{1}}{\left(1-k_{2}\lvert\beta_{,\tilde{y}}\rvert\right)^{2}}\beta_{,\tilde{y}\tilde{y}}.

where

k1=γDμb2ksin⁡2ϕ,k2=k∣sin⁡ϕ∣,k=a2k0ρ~ssb2.k_1=\frac{\gamma_{\mathrm{D}}}{\mu b^2}k\sin^2\phi,\qquad k_2=k|\sin\phi|,\qquad k=\frac{a^2}{k_0\tilde{\rho}_{ss}b^2}.

During the absorption and transmission process, the density of non-redundant dislocation ρg\rho^{\mathrm{g}} near the grain boundary hardly reaches the saturated density ρssg\rho^{\mathrm{g}}_{ss}. Therefore, using the Taylor expansion, we can approximate the defect energy ψm\psi_m as follows

ψm=γDψ~m,ψ~m=ρssg(ρgρssg+12(ρgρssg)2).\psi_m=\gamma_{\mathrm{D}}\tilde{\psi}_m,\qquad\tilde{\psi}_m=\rho^{\mathrm{g}}_{ss}\left(\frac{\rho^{\mathrm{g}}}{\rho^{\mathrm{g}}_{ss}}+\frac{1}{2}\left(\frac{\rho^{\mathrm{g}}}{\rho^{\mathrm{g}}_{ss}}\right)^2\right).

By this approximation, the equilibrium equation for microforces (21)$_1$, a stiff second-order quasi-linear differential equation, can be transformed into an elliptic differential equation

Aβ′′+Bβ+C=0,(22)A\beta''+B\beta+C=0, \tag*{(22)}

which is numerically more stable than the first one when the jump condition is implemented. The prime denotes the derivative of a function with respect to y~\tilde{y}, and

A=γDμb2a2b21k0ρ~sssin⁡2ϕ,B=(1−κ)sin⁡22ϕ,A=\frac{\gamma_{\mathrm{D}}}{\mu b^2}\frac{a^2}{b^2}\frac{1}{k_0\tilde{\rho}_{ss}}\sin^2\phi,\qquad B=(1-\kappa)\sin^2 2\phi,
C=−(κ⟨β⟩sin⁡22ϕ+(⟨β⟩cos⁡2ϕ−γ)cos⁡2ϕ+τ~i).C=-\left(\kappa\langle\beta\rangle\sin^2 2\phi+\left(\langle\beta\rangle\cos2\phi-\gamma\right)\cos2\phi+\tilde{\tau}_i\right).

At the external boundaries Eq. (21)$_(1)$ is subjected to the boundary condition

β,y~(0,f~)=0,β,y~(h~,f~)=0.(23)\beta_{,\tilde{y}}(0,\tilde{f})=0,\qquad\beta_{,\tilde{y}}(\tilde{h},\tilde{f})=0. \tag*{(23)}

From (13) we get the rescaled interface condition at the grain boundary as

\frac{\partial\llbracket\beta\rrbracket_{\Gamma}}{\partial\gamma} =\frac{1}{\tilde{\zeta}_a}\left(\frac{\kappa}{2}\left\llbracket\frac{\partial\tilde{\psi}_m}{\partial\tilde{\rho}^{\mathrm{g}}}\right\rrbracket_{\Gamma} -\frac{\operatorname{sign}\llbracket\beta\rrbracket_{\Gamma}}{4\pi(1-\nu_0)} \log\frac{\vartheta_m}{|\llbracket\beta\rrbracket_{\Gamma}|+\delta}\right),
⟦β,y~⟧Γ=2κρ~y.(24)\llbracket\beta_{,\tilde{y}}\rrbracket_{\Gamma}=\frac{2}{\kappa}\tilde{\rho}_y. \tag*{(24)}

where ζ~a=ζat0/(q0bγD)\tilde{\zeta}_a=\zeta_a t_0/(q_0b\gamma_{\mathrm{D}}) and ρ~y=b2ρy\tilde{\rho}_y=b^2\rho_y. The dimensionless form of Eq. (17) reads

ρ~y=ρ~y0+κ~a∣⟦β⟧Γ∣,\tilde{\rho}_y=\tilde{\rho}_{y0}+\tilde{\kappa}_a|\llbracket\beta\rrbracket_{\Gamma}|,

where κ~a=b2κa\tilde{\kappa}_{\mathrm{a}}=b^2\kappa_{\mathrm{a}}.

We also need to pose the initial conditions for β\beta, τ~i\tilde{\tau}_{\mathrm{i}}, ρ~\tilde{\rho} and χ~\tilde{\chi}. For the internal shear stress, we assume that the specimen is not plastically pre-deformed, i.e. τ~i(y~,0)=0\tilde{\tau}_{\mathrm{i}}(\tilde{y},0)=0. As for the plastic slip, two scenarios can be studied: (i) vanishing initial misalignment, (ii) nonzero initial misalignment. The numerical simulations will be performed for the first case where β(y~,0)=0\beta(\tilde{y},0)=0. The rescaled dislocation density ρ~(y~,0)\tilde{\rho}(\tilde{y},0) and the rescaled disorder temperature χ~(y~,0)\tilde{\chi}(\tilde{y},0) are assigned reasonable initial values based on various numerical simulations. It should be noted that, from the physical point of view, these initial values depend on the sample preparation.

Numerical implementation

We discretize the boundary value problem formulated above by a finite difference method called the line method. The latter replaces the spatial derivatives by finite differences, which allow us to transform partial differential equations into a system of ordinary differential equations with respect to γ\gamma. We illustrate this method for the case y~g=h~/2\tilde{y}_{\mathrm{g}}=\tilde{h}/2. For the case y~g≠h~/2\tilde{y}_{\mathrm{g}}\ne\tilde{h}/2 the discretization can be done in a similar way. Let the interval 0<y~<h~0<\tilde{y}<\tilde{h} be divided into 2n2n equal subintervals. Then the step in y~\tilde{y} along the grid becomes Δy~=h~/2n\Delta\tilde{y}=\tilde{h}/2n. Because of the jump in β\beta at the interface, we distinguish the left and right nodes located at the same point y~g=nΔy~\tilde{y}_{\mathrm{g}}=n\Delta\tilde{y}. The coordinates of the nodes on the left interval 0<y~<h~/20<\tilde{y}<\tilde{h}/2 are

y~i=iΔy~,i=0,…,n,\tilde{y}_{i}=i\Delta\tilde{y},\qquad i=0,\ldots,n,

while those on the right interval h~/2<y~<h~\tilde{h}/2<\tilde{y}<\tilde{h} are

y~i=(i−1)Δy~,i=n+1,…,2n+1.\tilde{y}_{i}=(i-1)\Delta\tilde{y},\qquad i=n+1,\ldots,2n+1.

Thus, altogether we have 2(n+1)2(n+1) nodes. The spatial derivatives of the plastic slip, β,y~\beta_{,\tilde{y}} and β,y~y~\beta_{,\tilde{y}\tilde{y}}, for all internal nodes not lying at the external boundaries or the interface, are approximated as

∂β(y~i,γ)∂y~=βi+1−βi−12Δy~,∂2β(y~i,γ)∂y~2=βi+1−2βi+βi−1Δy~2,\frac{\partial\beta(\tilde{y}_{i},\gamma)}{\partial\tilde{y}}=\frac{\beta_{i+1}-\beta_{i-1}}{2\Delta\tilde{y}},\qquad \frac{\partial^{2}\beta(\tilde{y}_{i},\gamma)}{\partial\tilde{y}^{2}}=\frac{\beta_{i+1}-2\beta_{i}+\beta_{i-1}}{\Delta\tilde{y}^{2}},

where βi=β(y~i,γ)\beta_{i}=\beta(\tilde{y}_{i},\gamma). We decompose the integral ⟨β⟩\langle\beta\rangle in the resolved shear stress into two integrals that are calculated using the trapezoidal rule. Thus

⟨β⟩=12n(12β0+β1+⋯+12βn+12βn+1+βn+2+⋯+12β2n+1).\langle\beta\rangle=\frac{1}{2n}\left(\frac{1}{2}\beta_{0}+\beta_{1}+\cdots+\frac{1}{2}\beta_{n}+\frac{1}{2}\beta_{n+1}+\beta_{n+2}+\cdots+\frac{1}{2}\beta_{2n+1}\right).

After these replacements Eq. (21) becomes an ordinary differential equation in all internal nodes.

The first derivatives at the external nodes are replaced by

∂β(y~0,γ)∂y~=β1−β0Δy~,∂β(y~2n+1,γ)∂y~=β2n+1−β2nΔy~.\frac{\partial\beta(\widetilde{y}_0,\gamma)}{\partial\widetilde{y}}=\frac{\beta_1-\beta_0}{\Delta\widetilde{y}},\qquad\frac{\partial\beta(\widetilde{y}_{2n+1},\gamma)}{\partial\widetilde{y}}=\frac{\beta_{2n+1}-\beta_{2n}}{\Delta\widetilde{y}}.

Thus, the boundary conditions (23) become

β0=β1,β2n+1=β2n.\beta_0=\beta_1,\qquad\beta_{2n+1}=\beta_{2n}.

The first left and right derivatives at the interface are replaced by

∂β(y~n,γ)∂y~=βn−βn−1Δy~,∂β(y~n+1,γ)∂y~=βn+2−βn+1Δy~.\frac{\partial\beta(\widetilde{y}_n,\gamma)}{\partial\widetilde{y}}=\frac{\beta_n-\beta_{n-1}}{\Delta\widetilde{y}},\qquad\frac{\partial\beta(\widetilde{y}_{n+1},\gamma)}{\partial\widetilde{y}}=\frac{\beta_{n+2}-\beta_{n+1}}{\Delta\widetilde{y}}.

while the jump in β\beta is

⟦β⟧Γ=βn+1−βn.\llbracket\beta\rrbracket_{\Gamma}=\beta_{n+1}-\beta_n.

These transform Eqs. (24) into two differential-algebraic equations. Since the three remaining equations of (21) do not contain derivatives, they are satisfied node by node. Besides, as the unknown functions are continuous at the interface, we have

τ~in=τ~in+1,ρ~n=ρ~n+1,χ~n=χ~n+1.\widetilde{\tau}_{\mathrm{in}}=\widetilde{\tau}_{\mathrm{in}+1},\qquad\widetilde{\rho}_n=\widetilde{\rho}_{n+1},\qquad\widetilde{\chi}_n=\widetilde{\chi}_{n+1}.

It is easy to see that the system of partial differential equations and boundary conditions becomes a system of 4(2n+1)4(2n+1) ordinary differential-algebraic equations (DAE) for 4(2n+1)4(2n+1) unknowns, so the problem is well-posed. For the numerical discretization we choose n=200n=200, while for the integration of DAEs with respect to the time-like variable γ\gamma a step size Δγ=10−4\Delta\gamma=10^{-4} is chosen. The DAE-system is then integrated by the Matlab integrator ode15s.

Numerical results

In the numerical simulations, we keep the temperature and shear rate constant at T=298 KT=298\,\mathrm{K} and q~0=10−13\widetilde{q}_0=10^{-13}. The parameters for the thermodynamic dislocation theory for copper are chosen as presented in [22]:

r=0.0323,θ=0.0073,Kχ=350,Kρ=96.1,χ~0=0.25.r=0.0323,\qquad\theta=0.0073,\qquad\mathcal{K}_{\chi}=350,\qquad\mathcal{K}_{\rho}=96.1,\qquad\widetilde{\chi}_0=0.25.

We choose the following parameters for the copper bicrystal

h=5.1 μm,b=0.255 nm,a=10b,μ=50 GPa,ν0=0.34,ϕ=30∘.h=5.1\,\mu\mathrm{m},\qquad b=0.255\,\mathrm{nm},\qquad a=10b,\qquad\mu=50\,\mathrm{GPa},\qquad\nu_0=0.34,\qquad\phi=30^\circ.

The active slip systems of two grains in the bicrystal are constrained to symmetry. The initial conditions for dislocation density and configuration temperature are ρ~(0)=6.25×10−5\tilde{\rho}(0)=6.25 \times10^{-5}, χ~(0)=0.23\tilde{\chi}(0)=0.23 and we assume γD=μb2\gamma_D=\mu b^2. The critical density is set as ρ~y0=1.3×10−5\tilde{\rho}_{y0}=1.3 \times10^{-5} and the other two interfacial dissipation parameters are given as ζ~a=1.5\tilde{\zeta}_a=1.5 and κ~a=5×10−4\tilde{\kappa}_a=5 \times10^{-4}. The parameters for the defect energy and interfacial energy are chosen as follows: k0=0.2k_0=0.2, ϑm=0.192\vartheta_m=0.192, δ=1×10−7\delta=1 \times10^{-7}. Unless some parameters are changed for the parameter study, they are kept as default values.

The saturation density of the non-redundant dislocations ρss\rho_{\mathrm{ss}} as well as the defect energy ψm\psi_m is characterized by the factor k0k_0, so that its value affects the distribution of plastic slip and work hardening. Figure 2(a) shows the distributions of plastic slip in the pile-up process for three different values of k0k_0. When k0k_0 is small, the plastic slip decreases gradually near the grain boundary, while it decreases more steeply to zero for a higher value of k0k_0. Since the slope of plastic slip is proportional to the accumulated dislocation density, these distributions imply that a smaller k0k_0 leads to a larger dislocation accumulation zone and a lower accumulated density of non-redundant dislocations near the grain boundary. This is because a small k0k_0 makes ρss\rho_{\mathrm{ss}} small, so a lower density of non-redundant dislocations is required for the backstress to have an effect on the balance of micro-forces. The hardening behavior in the pile-up process is shown in Figure 2(b). It can be seen that the hardening rate is larger for smaller k0k_0. The reason is that the wider dislocation accumulation zone and the slower gradient increase of plastic slip resulting from the small k0k_0 lead to a higher resolved shear stress and consequently a larger hardening rate. A smaller value of k0k_0 also leads to a longer length of the accumulation phase. This is because the accumulated ρg\rho^{g} is relatively low for small k0k_0 and therefore it takes longer to reach the required critical density ρy\rho_y.

Distribution of plastic slips and corresponding hardening behaviors

Figure 2. (Color online) (a) Distribution of plastic slips at γ=8∗10−3\gamma=8 \ast10^{-3} for k0=0.1k_0=0.1, k0=0.2k_0=0.2, and k0=0.5k_0=0.5. (b) The corresponding hardening behaviors in the pile-up process.

At different shear strains γ\gamma, the distribution of plastic slip in the processes of dislocation pile-up is shown in Figure 3(a). The Neumann boundary conditions β,y~(0)=β,y~(h~)=0\beta_{,\tilde{y}}(0)=\beta_{,\tilde{y}}(\tilde{h})=0 cause the curve of plastic slip to remain flat on two sides, while a groove is formed at the position of the grain boundary due to the Dirichlet interface condition. A dislocation pile-up zone is located in the center and two dislocation-free zones are located at the boundary layers. In this process, the homogeneous Dirichlet interface condition is applied, and the dislocations cannot penetrate into the grain boundary. Therefore, the groove sinks with increasing strain, but the tip remains attached to the bottom (Figure 3(a)). Figure 3(b) shows the evolution of the misalignment at the grain boundary, which is obtained from Eq. (13). When the non-redundant dislocation density ρg\rho^{g} reaches the critical value ρy0\rho_{y0} at γ=0.011\gamma=0.011, the evolution equation is activated. The grain boundary starts absorbing dislocations and the misorientation increases only when the strain exceeds the second threshold value at γ=0.0151\gamma=0.0151, up to which the accumulation process continues. ues. The dislocation absorption process stops at γ=0.0403\gamma= 0.0403 when the slope of the misorientation changes.

Plastic slips during the pile-up process and evolution of the misorientation at the grain boundary

Figure 3. (Color online) (a) Plastic slips during the pile-up process, (b) Evolution of the misorientation at the grain boundary.

The distribution of plastic slip during the dislocation absorption process is shown in Figure 4(a). In this process, the grain boundary is no longer impenetrable but absorbs dislocations, increasing the misorientation and causing non-redundant dislocations to accumulate. These are two factors that cause the distribution of plastic slip to take a different form than in the pile-up process, as shown in Figure 4(a). Here, it can be observed that the plastic slip on two sides of the grain boundary converges not to a single point, but to two points apart, and the distance between them increases, which means that the misorientation of the grain boundary increases. To see the effect of the additionally accumulated non-redundant dislocations, we omit the jump from the plastic slip distribution at γ=0.02\gamma= 0.02 and γ=0.03\gamma= 0.03 (solid lines) and compare it with that at the beginning of the absorption process (dashed lines) in Figure 4(b). It can be seen that the tip leaves the bottom, and as the shear stress increases, the discrepancy between the solid and dashed lines also increases, indicating that the groove does not stop sinking.

Plastic slip distributions for the absorption process and omitting the jump

Figure 4. (Color online) (a) Plastic slips for absorption process, (b) Plastic slip omitting the jump.

In the pile-up process, the rate of accumulated dislocation density is the highest among the different processes because all dislocations are blocked by the grain boundary. When positive and negative dislocations enter the grain boundary and increase the misorientation in the absorption process, the ability of the grain boundary to block dislocations from transfer is enhanced, and no additional dislocations are accumulated after this process. This behavior can be seen in Figure 5(a). In the pile-up process, the positive and negative non-redundant dislocation densities are symmetrically distributed on both sides of the grain boundary, while in the absorption process, a jump in the non-redundant density occurs near the grain boundary, as shown in Figure 5(b). Under the interfacial condition Eq. (14), the same amount of dislocation density is shifted from one side to the other.

Evolution of the mean density of the non-redundant dislocations and the non-redundant dislocation density at two sides of grain boundary

Figure 5. (Color online) (a) Evolution of the mean density of the non-redundant dislocations, (b) Evolution of the non-redundant dislocation density at two sides of grain boundary.

In Figure 6(a), the stress-strain curve of the present model (red curve) is shown, where P1 and P2 indicate the first and second thresholds and P3 the end of the absorption process. For comparison we plot also the stress-strain curves obtained by two other models. The blue curve represents the hardening behavior of bicrystals with impenetrable grain boundary, which blocks all non-redundant dislocations and therefore hardens the most. The black curve describes the hardening behavior of bicrystals without absorption process. When ρg\rho^{g} reaches the critical density ρy0\rho_{y0}, the pile-up process ends at P1 and the traversal process follows. These two curves represent the limits above and below which the hardening with absorption can vary due to different interfacial dissipation. Figure 6(b) shows the partition of work hardening into isotropic and kinematic hardening. Isotropic work hardening results from the internal stress required by the dislocations to overcome the resistance due to pinning. Kinematic work hardening is caused by back stresses describing the interaction between non-redundant dislocations. In the non-uniform plastic deformation of small size crystals, kinematic work hardening tends to contribute strongly, while in the present study the contribution is small because bicrystals have only one grain boundary where non-redundant dislocations accumulate. It is worth mentioning that for crystals with multiple grain boundaries, the kinematic work hardening rate is higher [38].

Rescaled stress-strain curves and partition of work hardening

Figure 6. (Color online) Rescaled stress-strain curves: (a) with impenetrable grain boundary (blue), with dislocation absorption process included (red), without dislocation absorption process (black), (b) Partition of work hardening.

Among the interfacial dissipation parameters, ρy0\rho_{y0} determines the first threshold and ζ~a\tilde{\zeta}_a the second, so they determine the onset of the dislocation absorption process. The size of the jump in the plastic slip as well as the length of the absorption process are controlled by ζ~a\tilde{\zeta}_a, as shown in Figure 7(a). Note that the jump exhibits a linear dependence on the shear strain. The kinematic strain hardening in the absorption process (Figure 7(b)) is influenced by the additional accumulated non-redundant dislocations originating from the increased misorientation, so that κ~a\tilde{\kappa}_a is the key controlling parameter.

Two-panel plots showing the influence of $\tilde{\zeta}_a$ and $\tilde{\kappa}_a$

Figure 7. (Color online) (a) Influence of ζ~a\tilde{\zeta}_a to the evolution of jump in plastic slip, (b) Influence of κ~a\tilde{\kappa}_a to the hardening behavior.

Finally, we show the distribution of plastic slip at γ=0.01\gamma= 0.01 with κa=2×10−4\kappa_a = 2 \times10^{-4} for ϕ=30∘\phi= 30^\circ, ϕ=35∘\phi= 35^\circ and ϕ=55∘\phi= 55^\circ in Figure 8(a). It can be seen that for 0<ϕ<45∘0 < \phi< 45^\circ and 135∘<ϕ<180∘135^\circ< \phi< 180^\circ the plastic slip is positively distributed, while for 45∘<ϕ<135∘45^\circ< \phi< 135^\circ it is the reverse. The dashed curve is symmetrical to the black curve about the horizontal coordinate axis. We see that the blue and dashed curves coincide in the dislocation-free zone. This is because the resolved shear stress τ\tau and the residual stress τi\tau_i for ϕ=35∘\phi= 35^\circ

Two-panel plots showing distributions of plastic slips and corresponding hardening behaviors

Figure 8. (Color online) (a) Distribution of plastic slips in pile-up process for ϕ=30∘\phi= 30^\circ, ϕ=35∘\phi= 35^\circ and ϕ=55∘\phi= 55^\circ, (b) The corresponding hardening behaviors.

are identical in magnitude but opposite in sign to those for ϕ=55∘\phi= 55^\circ. The distributions of plastic slip are different near the grain boundary because the accumulation of non-redundant dislocation density is inconsistent due to the different slip systems. Fig. 8(b) shows the strain hardening behavior for three different ϕ\phi, where the asterisks indicate the beginning and end of the absorption processes. It can be seen that the slip system affects not only the hardening rate but also the length of the absorption process lγl_\gamma, as indicated in the bar graph.

Conclusion

In the framework of thermodynamic dislocation theory for nonuniform plastic deformation, we have developed the interface conditions for various interactions of dislocations with the grain boundary, e.g., dislocation accumulation, absorption, and transfer. Interfacial dissipation in terms of rates for the mean and jump of plastic slip has been proposed. An evolution equation for grain boundary misorientation has been developed. The proposed interface dissipation yields two thresholds for the dislocation absorption process. The dislocation absorption leads to asymmetry in the distribution of plastic slip and non-redundant dislocation density on two sides of the grain boundary. It turns out that due to the additional accumulated density of non-redundant dislocations, kinematic work hardening also develops during the process of dislocation absorption. The extent of misorientation and the length of the dislocation absorption process is determined not only by the interfacial dissipation, but also by the slip system of the grains.

References

  1. [1]Abeyaratne, R., Knowles, J. K. (1990). On the driving traction acting on a surface of strain discontinuity in a continuum. J. Mech. Phys. Solids 38, 345–360.
  2. [2]Aifantis, K. E. (1999). Strain gradient interpretation of size effects. Int. J. Fract. 95, 299–314.
  3. [3]Aifantis, K. E., Willis, J. R. (2005). The role of interfaces in enhancing the yield strength of composites and polycrystals. J. Mech. Phys. Solids 53, 1047–1070.DOI
  4. [4]Berdichevsky, V. L., Sedov, L. I. (1967). Dynamic theory of continuously distributed dislocations. Its relation to plasticity theory. J. Appl. Math. Mech. (PMM) 31, 989–1006.DOI
  5. [5]Berdichevsky, V. L. (2006a). Continuum theory of dislocations revisited. Contin. Mech. Therm. 18, 195-222.DOI
  6. [6]Berdichevsky, V. L. (2006b). On thermodynamics of crystal plasticity. Scripta Mater. 54, 711-716.DOI
  7. [7]Bilby, B. A. (1955). Types of dislocation source. In Report of Bristol conference on defects in crystalline solids (pp. 124-133). The Physical Soc. Bristol 1954, London.
  8. [8]Burgers, J. M. (1939). Some considerations on the fields of stress connected with dislocations in a regular crystal lattice I. Proc. Kon. Ned. Akad. Wet. 42, 293.
  9. [9]Chowdhury, S. R., Roy, D., Reddy, J. N., Srinivasa, A. (2016). Fluctuation relation based continuum model for thermoviscoplasticity in metals. J. Mech. Phys. Solids 96, 353–368.arxiv.org/abs/1603.09036
  10. [10]Ekh, M., Bargmann, Grymer, M. (2011). Influence of grain boundary conditions on modeling of size-dependence in polycrystals. Acta Mech. 218, 103–113.DOI
  11. [11]Fleck, N. A., Muller, G. M., Ashby, M. F., Hutchinson, J. W. (1994). A mathematical basis for strain-gradient plasticity theory - Part I: Scalar plastic multiplier. J. Mech. Phys. Solids 57, 161-177.
  12. [12]Fleck, N. A., Willis, J. R. (2009). Strain gradient plasticity: Theory and experiment. Acta Metall. Mater. 42, 475-487.
  13. [13]Fredriksson, P., Gudmundson, P. (2007). Modelling of the interface between a thin film and a substrate within a strain gradient plasticity framework. J. Mech. Phys. Solids 55, 939-955.DOI
  14. [14]Gao, H., Huang, Y., Nix, W. D., Hutchinson, J. W. (1999). Mechanism-based strain gradient plasticity – I. J. Mech. Phys. Solids 47, 1239–1263.DOI
  15. [15]Gottschalk, D., McBride, A., Reddy, B. D., Javili, A., Wriggers, P., Hirschberger, C. B. (2016). Computational and theoretical aspects of a grain-boundary model that accounts for grain misorientation and grain-boundary orientation. Comput. Mater. Sci. 111, 443–459.DOI
  16. [16]Groma, I., Csikor, F. F., Zaiser, M. (2003). Spatial correlations and higher-order gradient terms in a continuum description of dislocation dynamics. Acta Mater. 51, 1271–1281.DOI
  17. [17]Günther, F., Le, K. C. (2021). Plane constrained shear of single crystals. Arch. Appl. Mech. 91, 2109–2126.DOI
  18. [18]Gurtin, M. E. (2003). On a framework for small-deformation viscoplasticity: free energy, microforces, strain gradients. Int. J. Plasticity 19, 47–90DOI
  19. [19]Gurtin, M. E. (2008). A theory of grain boundaries that accounts automatically for grain misorientation and grain-boundary orientation. J. Mech. Phys. Solids 56, 640–662.DOI
  20. [20]Kondo, S., Mitsuma, T., Shibata, N., and Ikuhara, Y., 2016. Direct observation of individual dislocation interaction processes with grain boundaries. Sci. Adv. 2, 1501926.pubmed.ncbi.nlm.nih.gov/27847862
  21. [21]Kröner, E. (1955). Der fundamentale Zusammenhang zwischen Versetzungsdichte und Spannungsfunktionen. Z. Phys. 142, 463–475.DOI
  22. [22]Langer, J.S., Bouchbinder, E., Lookman, T. (2010). Thermodynamic theory of dislocation-mediated plasticity. Acta Mater. 53, 3718–3732.arxiv.org/abs/0908.3913
  23. [23]Langer, J. S. (2016). Thermal effects in dislocation theory. Phys. Rev. E 94, 063004.DOI
  24. [24]Langer, J. S. (2017). Thermodynamic theory of dislocation-enabled plasticity. Phys. Rev. E 96, 053005.DOI
  25. [25]Langer, J.S., Le, K.C. (2020). Scaling confirmation of the thermodynamic dislocation theory. Proc. Natl. Acad. Sci. USA 117, 29431-29434.DOI
  26. [26]Le, K.C. (2018). Thermodynamic dislocation theory for non-uniform plastic deformations. J. Mech. Phys. Solids 111, 157–169.arxiv.org/abs/1705.08794
  27. [27]Le, K.C. (2019). Thermodynamic dislocation theory: Finite deformations. Int. J. Eng. Sci. 139, 1–10.arxiv.org/abs/1902.06761
  28. [28]Le, K.C. (2020). Two universal laws for plastic flows and the consistent thermodynamic dislocation theory. Mech. Res. Commun. 109, 103597.arxiv.org/abs/2002.03771
  29. [29]Le, K.C., Günther, C. (2014). Nonlinear continuum dislocation theory revisited. Int. J. Plasticity 53, 164–178.DOI
  30. [30]Le, K.C., Piao, Y. (2019). Thermal softening during high-temperature torsional deformation of aluminum bars. Int. J. Eng. Sci. 137, 1–7.arxiv.org/abs/1811.11500
  31. [31]Le, K.C., Sembiring, P., Tran, T.N. (2016). Continuum dislocation theory accounting for redundant dislocations and Taylor hardening. Int. J. Eng. Sci. 106, 155–167.arxiv.org/abs/1506.03583
  32. [32]Le, K. C., Tran, T. M. (2018). Thermodynamic dislocation theory: Bauschinger effect. Phys. Rev. E 97, 043002.DOI
  33. [33]Le, K.C., Tran, T.M., Langer, J.S. (2017). Thermodynamic dislocation theory of of high-temperature deformation in aluminum and steel. Phys. Rev. E 96, 013004.DOI
  34. [34]Le, K.C., Tran, T.M., Langer, J.S. (2018). Thermodynamic dislocation theory of adiabatic shear banding in steel. Scripta Mater. 149, 62–65.arxiv.org/abs/1710.05566
  35. [35]Le, K.C., Le, T.H., Tran, T.M. (2020). Averaging in dislocation mediated plasticity. Int. J. Eng. Sci. 149, 103230.DOI
  36. [36]Molodov, D. A., Gorkaya, T., Gottstein, G. (2011). Dynamics of grain boundaries under applied mechanical stress. J. Mater. Sci. 46, 4318–4326.DOI
  37. [37]Nye, J.F. (1953) Some geometrical relations in dislocated crystals. Acta Metall. 1, 153–162.DOI
  38. [38]Piao, Y., Le, K.C. (2020). Dislocation impediment by the grain boundaries in polycrystals. Acta Mech. 232, 3193–3213.arxiv.org/abs/2008.09056
  39. [39]Peng, X., Huang, G. (2015). Modeling dislocation absorption by surfaces within the framework of strain gradient crystal plasticity. Int. J. Solids Struct. 72, 98–107.DOI
  40. [40]Pouriayevali, H., Xu, BX. (2017). Decomposition of dislocation densities at grain boundary in a finite-deformation gradient crystal-plasticity framework. Int. J. Plasticity 96, 36-55.DOI
  41. [41]Read, W.T., Shockley, W. (1950). Dislocation models of crystal grain boundaries. Phys. Rev. 78, 275–289.DOI
  42. [42]Rupert, T. J., Gianola, D. S., Gan, Y., Hemker, K. J. (2009). Experimental observations of stress-driven grain boundary migration. Science 326, 1686-1690.DOI
  43. [43]Vitek, V. (1987). On the difference between the misorientation dependences of the energies of tilt and twist boundaries. Scripta Metall. 21, 711–714.DOI
  44. [44]Van Beers, P.R.M., McShane, G.J., Kouznetsova, V.G., Geers, M.G.D. (2013). Grain boundary interface mechanics in strain gradient crystal plasticity. J. Mech. Phys. Solids 61, 2659–2679.
  45. [45]Voyiadjis, G., Deliktas, B. (2009). Formulation of strain gradient plasticity with interface energy in a consistent thermodynamic framework. Int. J. Plasticity 25, 1997-2024.DOI
  46. [46]Voyiadjis, G., Faghihi, D., Zhang, Y. (2014). A theory for grain boundaries with strain-gradient plasticity. Int. J. Solids Struct 51, 1872–1889.DOI

Paper details

vlnc:W_5pbkj7hm
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents