Elsevier

Applied Numerical Mathematics

Volume 104, June 2016, Pages 204-217
Applied Numerical Mathematics

A fitted finite element method for the numerical approximation of void electro-stress migration

https://doi.org/10.1016/j.apnum.2015.08.008Get rights and content

Abstract

Microelectronic circuits usually contain small voids or cracks, and if those defects are large enough to sever the line, they cause an open circuit. A fully practical finite element method for the temporal analysis of the migration of voids in the presence of surface diffusion, electric loading and elastic stress is presented. We simulate a bulk–interface coupled system, with a moving interface governed by a fourth-order geometric evolution equation and a bulk where the electric potential and the displacement field are computed. The method presented here follows a fitted approach, since the interface grid is part of the boundary of the bulk grid. A detailed analysis, in terms of experimental order of convergence (when the exact solution to the free boundary problem is known) and coupling operations (e.g., smoothing/remeshing of the grids, intersection between elements of the two grids), is carried out. A comparison with a previously introduced unfitted approach (where the two grids are totally independent) is also performed, along with several numerical simulations in order to test the accuracy of the methods.

Introduction

Microelectronic circuits contain thin lines of aluminium alloy, that make electric contact between neighbouring devices possible. These lines are passivated with a layer of oxide at large temperatures, and during the cooling process large stresses are induced. As the dimensions of microchips are reduced further and further, and since interconnects always contain small voids or cracks, it is of great interest to investigate the physical mechanisms that impede such a reduction, due to mechanical failures in the lines induced by the motion of the cracks. The problem analysed in this paper involves the evolution over time of voids in a conducting metal line where three different contributions to the drift of the voids are present: the surface tension, the electric field and the elastic energy. This phenomenon is known as electro-stress migration; for further details see, e.g., [36], [14], [3], and the references therein.

As the height of interconnect lines is much smaller than the dimensions of the base, voids generally fully penetrate the conducting material. Hence it is common to consider a two dimensional model for void electro-stress migration, and this is the approach that we are going to pursue in this paper. In addition, for ease of exposition, we assume that the interconnect line is given by a rectangular solid. The electric field is induced in the line by prescribing the voltage on its vertical boundaries, while the displacement field is induced by prescribing the stresses on its four boundaries.

In this paper, based on our previous work in [31], we introduce a novel front-tracking, fitted finite element method for the approximation of void electro-stress migration. The main difference to the approximation presented in [31] is that here we consider the fitted approach, which means that the interface mesh is always part of the boundary of the bulk grid. Moreover, we also include the effect of stress-migration into the model. As an aside we note that our method inherits the good interface mesh properties from the approximation in [31]. In particular, the vertices on the discrete interface equidistribute asymptotically so that no reparameterisation of the discrete interface is necessary in practice.

The paper is organised as follows. In Section 2 we give a mathematical description of the problem of void electro-stress migration that we are interested in. We also give a brief overview of the different numerical methods applicable to this problem. In addition, we highlight the differences between the fitted approach presented in this paper and the unfitted approach previously introduced by the authors in [31]. Section 3 contains a detailed description of our proposed finite element approximation. In Section 4 we discuss possible solution methods of the algebraic system of equations arising at each time level. In addition, we present details on the bulk mesh smoothing strategy. Finally, in Section 5 we perform a convergence experiment for a test case in which the exact solution is known, and we present various other examples of the application of our numerical method.

Section snippets

Problem formulation

For the formulation of the governing equations we closely follow the presentation in [31], see also [3]. Let Ω=(L1,L1)×(L2,L2), where L1,L2>0, be the domain that contains the conductor. We denote the boundary of Ω with ∂Ω. At any time t[0,T], let Γ(t)Ω be the boundary of the void Ω(t) inside the conductor Ω. Then Γ(t)=Ω(t) and Ω+(t):=ΩΩ(t) denotes the conducting region (see Fig. 1). Now the evolution of the interface Γ(t), which represents the void boundary, is given byV=α1ϰss+α2ϕss+α

Finite element approximation

We begin with the finite element approximation for quantities defined over the bulk mesh. Let 0=t0<t1<<tM1<tM=T be a partitioning of [0,T] into possibly variable time steps τm:=tm+1tm, m=0,,M1. We set τ:=maxm=0,,M1τm. Let Tm be a partitioning of Ω+m, a polygonal approximation of Ω+(tm), into open disjoint triangles. Let Γm be the inner boundary of Ω+m, so that Ω+m=ΓmΩ. We can now define the standard finite element space of piecewise linear functions:Sm:={χC(Ω+m):χ|o is linear oTm},

Solution method

Due to the special structure of the system (17a), (17b), (17c), (17d), the equations for Φm+1, Um+1 and (Xm+1,κm+1) decouple. In practice, we can find the unique solution to (17a), (17b), (17c), (17d) as follows. First we find Φm+1Sgm such thatΘmΦm+1=0, where ΘmRK×K is the standard stiffness matrix for the Laplacian on Ω+m, i.e.[Θm]kl:=Ω+mψkmψlmdx, where {ψkm}k=1K are the basis functions of the unconstrained finite element space Sm. In the above we have ignored the effect of the

Numerical experiments

We implement our finite element approximation within the framework of the C++-based software DUNE, see [11], [10], and we employ the Alberta grid manager, see [34]. Unless stated otherwise, we use uniform time steps τm=τ, m=0,,M1, for all the numerical experiments in this section. As we will compare our numerical results to the phase field computations in [9], [3], we also fix α1=116π2. When stress-migration is considered, we assume that (6) holds and put α3=18π throughout this section. On

Conclusion

We have presented a fitted front-tracking method for the approximation of electro-stress migration using parametric finite elements. The main properties of our method are that we can prove an equidistribution property for the vertices on the discrete interface, and that in the absence of external forces the scheme can be shown to be unconditionally stable. We note that in practice, in all our computations, the method is also stable in the presence of forces due to the applied electric field and

References (36)

  • A. Schmidt

    Computation of three dimensional dendrites with finite elements

    J. Comput. Phys.

    (1996)
  • L. Xia et al.

    A finite element analysis of the motion and evolution of voids due to strain and electromigration induced surface diffusion

    J. Mech. Phys. Solids

    (1997)
  • J. Barrett et al.

    Finite element approximation of a phase field model for surface diffusion of voids in a stressed solid

    Math. Comput.

    (2006)
  • J.W. Barrett et al.

    A phase field model for the electromigration of intergranular voids

    Interfaces Free Bound.

    (2007)
  • J.W. Barrett et al.

    Finite element approximation of one-sided Stefan problems with anisotropic, approximately crystalline, Gibbs–Thomson law

    Adv. Differ. Equ.

    (2013)
  • J.W. Barrett et al.

    Finite element approximation of a phase field model for void electromigration

    SIAM J. Numer. Anal.

    (2004)
  • P. Bastian et al.

    A generic grid interface for parallel and adaptive scientific computing, part II: implementation and tests in DUNE

    Computing

    (2008)
  • P. Bastian et al.

    A generic grid interface for parallel and adaptive scientific computing, part I: abstract framework

    Computing

    (2008)
  • Cited by (0)

    View full text