Elsevier

Mechanics of Materials

Volume 136, September 2019, 103074
Mechanics of Materials

Ab initio aided strain gradient elasticity theory in prediction of nanocomponent fracture

https://doi.org/10.1016/j.mechmat.2019.103074Get rights and content

Highlights

  • Strain gradient elasticity theory (SGET) was combined with ab-initio and molecular statics simulations.

  • This hybrid approach proved to give reasonable predictions of fracture problems at micro and nano scales.

  • Ab-initio calculations of the phonon-dispersion relations were fitted by the dispersion relations predicted by SGET.

  • Ab-initio calculations of the displacement field near the screw dislocation was fitted by the SGET solution of this field.

  • Ab-initio calculations were used to find the deformation mode of atoms when a single bond breaks at the crack tip.

Abstract

The aim of the paper is to address fracture problems in nanoscale-sized cracked components using a simplified form of the strain gradient elasticity theory aided by ab initio calculations. Quantification of the material length scale parameter l1 of the simplified form of the strain gradient elasticity theory plays a key role in the analysis. The parameter l1 is identified for silicon and tungsten single crystals using first principles calculations. Specifically, the parameter l1 is extracted from phonon-dispersions generated by ab-initio calculations and, for comparison, by adjusting the analytical strain gradient elasticity theory solution for the displacement field near the screw dislocation with the ab-initio calculations of this field. The obtained results are further used in the strain gradient elasticity modeling of crack stability in nano-panels made of silicon and tungsten single crystals, where due to size effects and nonlocal material point interactions the classical linear fracture mechanics breaks down. The cusp-like crack tip opening profiles determined by the gradient elasticity theory and a hybrid atomistic approach at the moment of nano-panels fracture revealed a very good mutual agreement.

Introduction

In the analysis of brittle fracture of nanostructured solids (e.g. nanocomponents of advanced devices, with the thickness ranging from few nanometers to few hundred nanometers) two major questions arise in an attempt to predict the fracture toughness: (i) whether continuum fracture mechanics works at this scale and (ii) whether the stress intensity factor is sufficient to predict fracture toughness. The two questions are closely interconnected when considering the basic hypothesis of fracture mechanics that the K-dominant region must be geometrically much larger than the fracture process zone which incorporates inelastic deformation near the crack-tip. It was shown that beyond a certain limit, the size of the K-dominance zone starts to decrease with decreasing size (width) of the nanocomponent (Shimada et al., 2015). When the size of the K-dominance zone becomes smaller than 2–3 nm it starts to be comparable to the fracture process zone (0.4–0.6 nm) that embraces a highly concentrated discrete motion of atoms near the crack-tip in atomistic models. The fracture process starts to be dominated by far-stress field terms and, therefore, the critical stress intensity factor can no more represent the total fracture driving force. Indeed, it becomes size dependent and inadequate to characterize fracture toughness of a nanostructure. The Griffith energy release rate (ERR) exhibits similar behavior. This critical case corresponds to the specimen width of approximately 10 nm below which the classical continuum approach breaks down. Such a failure comes up because, in the critically small specimens, the actual strain energy distribution near the crack-tip is no longer described by the classical continuum approach. In this respect cohesive zone models with a traction-separation law determined from atomistic simulations combined with the assumption that the material behavior near the crack tip is controlled by the K-field may fail beyond a certain critical size of a nanocomponent.

There are attempts in literature to capture the size dependency of the fracture toughness Kc using a two parameter model based on Williams's expansion, however the size dependency of Kc and/or critical ERR contradicts the constant critical ERR obtained using atomistic simulations (Sun and Qian, 2009). By analyzing the stability of a crack in the silicon nano-panel (Kotoul and Skalka, 2017) we proved that the continuum assumption can be extended beyond the limit of the classical fracture mechanics using the strain gradient elasticity theory (SGET) proposed by Mindlin (1964) which removes the presence of singular strain field at the crack tip consistently with atomistic simulations based on either the density functional theory (DFT) or molecular statics/dynamics (MS/D). This approach allows addressing deformation and fracture problems at micron and nano scales in an effective and computationally robust manner thus helping to bridge the gap between classical continuum theories and atomic-lattice theories.

The most simplified form of SGET employs only one internal length scale parameter in addition to the two classical Lamé constants. In this context it should be noted that even in case of linear, isotropic, gradient-dependent elastic material, the most general form of the strain energy depends (except of the two Lamé constants) on five additional gradient coefficients which appear in Mindlin's strain gradient theory (Mindlin, 1964). Besides the difficulties to find these coefficients, the resulting equations are quite formidable and hard to solve. Thus, a physical simplification could be useful. As already suggested in Aifantis (1984), the higher-order elastic parameters might be assumed proportional to the conventional elastic stiffness coefficients by the internal length material parameter. Persuasive arguments for such assumption are offered by Lazar and Maugin (2005). One of the arguments follows the approach suggested by Feynman in linear gravity (Feynman, 1995). The second one is based on the suggestion that the elastic energy should exhibit the same symmetry in gradient terms like the stress-strain symmetry of the elastic energy.

Thus, the crucial point is to identify the above mentioned internal length scale parameter. In the present paper, two approaches are employed for its identification by fitting the solutions for (i) phonon dispersion curves and (ii) displacement fields near the screw dislocation as obtained by DFT approaches. Since the simplified form of SGET should particularly work for isotropic materials, the tungsten single crystal was selected for determination of its length scale parameter and the critical ERR by combined SGET/DFT and DFT/MS approaches, respectively. Moreover, experimental values of the critical ERR are also known for this material. The silicon single crystal with a rather small anisotropy and widely experimentally explored fracture properties was selected as reference material. Both the critical ERR and the critical crack tip opening displacement (CTOD) for the Si crystal were already calculated using the atomistic approaches (Shimada et al., 2015; Perez and Gumbsch, 2000) and, therefore, these time-consuming computations could be omitted here. Let us note that both selected crystals are convenient for the present study also because they are, at least at low temperatures, intrinsically brittle (Pokluda et al., 2015).

Due to higher-order terms the classical linear relationship between the frequency and the wave vector becomes nonlinear for longer wave-vectors thus manifesting dispersive effects, the first phenomenon suitable for a determination of the length scale parameter. The acoustic part of the dynamical matrices is simplified by requiring that the stress-strain symmetry of the elastic energy should also be valid for the gradient terms, which leads to Laplacian version of SGET. Thus, the sixth-order tensor of higher order elastic constants of the SGET is replaced by a fourth-order tensor of classical elasticity multiplied by the second power of the yet unknown material length scale parameter. So instead of calculating six independent components of the sixth-order tensor by numerical fitting of the acoustic part from ab-initio calculations (based on the density functional theory (DFT)) one needs to obtain just a single material length scale parameter. Apparently, it would be sufficient to consider the dynamical matrix along a single high-symmetry direction, e.g. [1,0,0]. However, also other high-symmetry directions such as [1,1,0] or [1,1,1] should be considered to determine whether the material length scale parameter changes or not. Additionally, as shown by Mindlin (1964), one also needs to consider higher order terms in the kinetic energy which lead to acceleration gradients in the equation of motion.

In the second method for a determination of the length scale parameter, the analytical SGET-solution for the displacement field near the screw dislocation in the gradient elasticity is compared with the DFT calculations of the developing displacement field near the screw dislocation in cubic crystals of W and Si. We started with the screw dislocation with the Burgers vector ½ a [111] in tungsten (expressed in terms of the lattice constant a) and consecutively, the screw dislocation ½ a 〈110〉 was analyzed in silicon.

The stability of cracks in the tungsten nanocrystal was then studied using molecular statics (MS) simulations and both the critical CTOD and ERR were determined as the critical fracture parameters. For the silicon crystal these data were taken over from the literature (Shimada et al., 2015; Perez and Gumbsch, 2000). Simultaneously, the critical fracture parameters and the identified length scale parameters were used in simulations of the crack stability in W and Si nanocrystals in terms of the SGET.

Section snippets

Basic equations of strain gradient elasticity

A detailed presentation of the Form II of Mindlin's theory of SGET can be found, e.g., in Gourgiotis and Georgiadis (2009), Mindlin and Eshel (1968), Grentzelou and Georgiadis (2008), Georgiadis and Grentzelou (2006) and Aravas and Giannakopoulos (2009). The strain energy density w is a function of the linear strain tensor εij=1/2(jui+iuj) and of its gradient ∂kɛij,w=w(εij,kεij). The monopolar stress tensor τij and the so-called dipolar (or double) stress tensor mkij are then defined asτij=w

Determination of the length scale parameters for tungsten and silicon

We have used two approaches to determine length scale parameters as briefly mentioned in Introduction. The first one is based on fitting acoustic phonon dispersions generated by ab-initio calculations along high-symmetry directions. Similar procedure was also used in the work (Maranganti and Sharma, 2007), where however more higher order coefficients were considered, but on the other hand, the acceleration gradients were not incorporated.

Considering a plane-wave solution of the formui=ui0expi(ωt

Analysis of nano-cracks in silicon and tungsten single crystals

It was well documented experimentally and by atomistic simulations (Gumbsch, 2001, Gumbsch, 2003) that the brittle fracture process in both silicon and tungsten single crystals exhibits an anisotropy with respect to the direction of crack propagation on a specific cleavage plane due to lattice trapping nanopanels effects. This anisotropy is easier to observe in silicon which can be produced as a virtually dislocation-free single crystal and a crack can propagate without a dislocation activity

Summary

The material length scale parameters l1 of the simplified form of the SGET was determined by fitting the ab-initio DFT calculations of the phonon-dispersion relations for crystals of silicon and tungsten and, for comparison, by adjusting the analytical SGET solution for the displacement field near the screw dislocation with the DFT calculations of this field. Both adjusting methods led to the values l1 = 0.28 nm for the Si-crystal and l1 = 0.22 nm for W crystal. These values well corresponded

Acknowledgements

The authors gratefully acknowledge a financial support of the Czech Science Foundation under the Project No. 17-18566S. Computational resources were provided by the Ministry of Education, Youths and Sports of the Czech Republic under the Project IT4Innovations National Supercomputer Center (Project No. LM2015070) within the program Projects of Large Research, Development and Innovations Infrastructures. Fruitful discussions with Dr. Jan Fikar from the Institute of Physics of Materials, Academy

References (41)

Cited by (13)

  • Interaction of a semi-infinite crack with a screw dislocation within Mindlin's first strain-gradient elasticity

    2022, Engineering Fracture Mechanics
    Citation Excerpt :

    This study shows a coincidence between the nonsingular stress field predicted by such continuum model at the crack tip and the results obtained from the simulations. Also, Kotoul et al. [53] have shown that the nonsingular strain field at crack tips predicted by Mindlin’s strain-gradient theory is in agreement with the results of atomistic simulations based on the density functional theory and molecular statics/dynamics. Accordingly, they emphasized on the effectiveness and robustness of this theory for addressing deformation and fracture problems at micron and nano scales.

  • Determination of length scale parameters of strain gradient continuum theory for crystalline solids using a computational quantum mechanical model based on density functional theory

    2022, Current Applied Physics
    Citation Excerpt :

    In recent decades, many researchers have utilized ab initio methods like DFT for the calculation of the elastic constants of solids [19–26]. However, a very limited number of studies [17,27–32] have been conducted for the calculation of length scale parameters of enhanced continuum theories using such methods. In the present study, the characteristic lengths of an enhanced continuum theory, i.e., Mindlin's strain gradient continuum theory, have been determined for a number of main-group or transition metal solids that have face-centered cubic (fcc) or body-centered cubic (bcc) structures by using ab initio DFT.

  • From continuum to quantum mechanics study on the fracture of nanoscale notched brittle materials

    2021, International Journal of Mechanical Sciences
    Citation Excerpt :

    Nevertheless, as the bond-breaking processes involve complex electronic rearrangements [42,43], first-principles calculations based on quantum mechanics becomes a good candidate to correctly illustrate the large changes in bond angles and coordination that affect the interatomic forces. Up to now, the fracture strength of silicon, which is defined as the maximum stress to render the materials elastically unstable, has been carefully investigated using first-principles calculations [44-48]. However, the fracture strength of silicon is not inherent while exhibits remarkable size and geometry dependence [47,49-51].

  • Geometric modelling of elastic and elastic-plastic solids by separation of deformation energy and Prandtl operators

    2020, International Journal of Solids and Structures
    Citation Excerpt :

    Finite volume methods (Keilegavlen and Nordbotten, 2017; Sokolova et al., 2019) and boundary-type methods (Markous, 2019; Liu et al., 2019) can alternatively be applied to study the deformation of solids. At present, the finite element methods are the most mature and commonly used tool to analyse solids subjected to mechanical or thermomechanical loadings (Liu et al., 2015; Fielder et al., 2017; Rajasekaran and Khaniki, 2017; Barboura and Li, 2018; Šeruga and Nagode, 2019; Kotoul et al., 2019; Xu and Cheng, 2019; Xiao and Yu, 2019; Pereira et al., 2019; Numanoglu and Civalek, 2019; Falope et al., 2019). All these methods are based on continuum formulations of governing equations, i.e. by partial differential equations, complemented by boundary conditions defining boundary value problems.

View all citing articles on Scopus
View full text