A scheme for the passage from atomic to continuum theory for thin films, nanotubes and nanorods

https://doi.org/10.1016/S0022-5096(99)00091-5Get rights and content

Abstract

We propose a scheme for the direct passage from atomic level to continuum level. The scheme is applicable to geometries, like films, rods and tubes, in which one or more dimensions are large relative to atomic scale but other dimensions may be of atomic scale. The atomic theory is assumed to be governed by a variational principle resting on the Born–Oppenheimer approximation. The atomic level energy is further assumed to satisfy certain decay properties when evaluated for disjoint sets of atoms. The scheme is based on two hypotheses: (1) distortions are limited, and (2) there are many atoms in certain directions. The scheme produces in a natural way the variables of the continuum theory. In the case of a film, the continuum theory that emerges is a Cosserat membrane theory with (ν−1) Cosserat vectors, ν being the number of atomic layers in the film. The arguments presented are not mathematically rigorous. One difficulty is that it is not clear under which circumstances our decay hypothesis on interatomic interactions is consistent with quantum mechanics or density functional theory.

Introduction

The ability to reproducibly synthesize structures having an atomic scale dimension, and the subsequent investigation of the unusual properties and possible applications of these nanoscale structures, has given rise to the field of nanotechnology. Properties that are known to be structure-sensitive on the macroscale, such as strength, plasticity, magnetic hysteresis, superconductivity, and properties altered by phase transformation are expected to exhibit unusual behavior on scales that are smaller than the typical microstructural feature that is responsible for the sensitivity.

When all of the dimensions of a nanoscale object are small, it is possible to use purely atomic scale computational methods to understand its behavior. But when at least one dimension of the object is much larger than atomic scale, purely atomistic methods become time consuming, and they also suffer from the drawback of offering only one-configuration one-computation; that is, they sometimes do not give insight into the variety of possible behaviors of the structure.

Lacking a suitable alternative, the current practice is to transport macroscale continuum theories down to the nanoscale. For example, the motion of nanoscale cantilevers is often analyzed using macroscale beam theory (e.g., Wong et al., 1997). The moment–curvature relation in these beam theories is given as the product EI of Young’s modulus and moment of inertia. For a nanotube or nanorod, it is not clear what one should use for the moment of inertia I of the cross-section. One could imagine that it represents a moment of inertia of a kind of cross-sectional envelope of the atoms. But cuts at different locations along the rod give very different sets of atoms. Furthermore, the origins of the term EI — the comparison of a solution in Euler beam theory with a similar solution of linear elasticity — indicate that the presence of this product is associated with a linear variation of stress across the cross-section, a situation that is difficult to imagine on the nanoscale. More than that, even the concept of stress (or, in electromagnetism, the concept of the electromagnetic field in a polarizable or magnetizable medium) is essentially macroscopic and may not be a reasonable concept to use at the nanoscale.

The existence of a nanoscale dimension is in the authors’ opinion not a reason by itself to reject continuum theory. Roughly, continuum theories should arise when there are one or more dimensions much larger than atomic scale and there is some macroscopic homogeneity. In statics, the problem of deriving continuum theory is essentially a problem of partitioning the configurational variables into two sets: one set is treated by atomic methods and the other set furnishes the variables of the continuum theory. Along the way, a successful method of partitioning achieves simplification. The opportunity for simplification arises from the fact that for an extended film large classes of nanoscale configurations have nearly the same energy, as compared to a typical change of energy of the whole film.

We propose the following method for deriving continuum theory from atomic theory. Consider, for example, a film having ν atomic layers and lateral dimensions k×k, where k is much larger than atomic scale. First, to limit the number of possible states, we assume that the distortion of the film is bounded; that is, we assume that the affine deformation experienced by every triple of atoms has a bounded gradient. This immediately implies that there exist some well-defined macroscale averages. Then the method is described as follows:

  • 1.

    Calculate all averages of atomic positions that are well-defined in the limit k→∞. In mathematical terms, for each fixed k rescale the deformation to a fixed domain, and calculate all weak limits of the sequence of rescaled deformations. Use these averages as the variables of continuum theory.

  • 2.

    Find a shrinking neighborhood Nk of these averages having the property that, if a sequence yk of deformations lies in the neighborhood for each k, then they converge in the sense found above to these averages.

  • 3.

    Minimize the atomic degrees of freedom in the neighborhood for each k.

  • 4.

    Calculate the energy per unit area, pass to the limit k→∞ and extract a continuum theory.

In this paper, we outline this procedure explicitly in the case of thin films. The argument at this stage is formal, but we believe that it can be made rigorous. We concentrate on the physical basis of the scheme in this paper. The scheme has several desirable features. First, it is compatible with the basic physics of interatomic forces, governed by quantum mechanics or simplified versions such as density functional theory. In particular, full many-body interatomic potentials as arising from minimization over electronic states are treated, and the positions of individual atoms are not assigned a priori, just certain averages. Second, it yields a specific procedure for calculating a continuum level energy in terms of the atomic scale relaxation under homogeneous conditions, with certain averages fixed. Essentially, it gives a quantitative interpretation of the notion of “local state” in continuum mechanics. Finally, unlike classical rules for passing from atomic to continuum levels, this method in a natural way produces the variables of the continuum theory.

To develop this procedure, there remains a lot of work to do. We only carry out the argument at a formal level for the thin film case, and indicate briefly how it works for nanotubes and rods. Second, as presented and expected, the scheme only delivers the membrane energy. The bending energy occurs at higher order in 1/k. One could think of applying a similar procedure to the difference between the original energy and the membrane energy, multiplied by a suitable power of k (k3 is expected classically), but we have no results in this direction. Third, the method is explicitly static. Lattice vibrations could be added by superimposing the formulas of statistical mechanics on the atomic scale relaxation as discussed by many authors, but full-scale dynamic motions are not treated by this method. Fourth, the continuum theory that emerges needs to be further simplified to be widely applicable. For example, the linearized theory needs to be explored, and any possible separation between the effects of geometry and atomic forces (such as embodied in expressions like EI) should be exposed.

Section snippets

Kinematics

Consider a single crystal film having ν monolayers. For simplicity we picture the nuclei as occupying a reference configuration on a Bravais latticeL=Le1,e2,e3=μiei:μiareintegersand a film of thickness h=ν−1|e3| and planar section Sk=0,k×0,k, so that the nuclei in the reference configuration of the film occupy the positions,LSk×0,hThe theory is intended to apply to films that have been released from the substrate, or else to films attached to substrates that are also very thin. In other words,

Weak neighborhoods

Having identified u and the ν−1 functions bi as the basic functions of the continuum theory, we now face the task of designing a suitable shrinking neighborhood of these functions, which we denote by Ñεu,b1,…,bν−1. The subscript ε refers to the a mesoscopic length scale ε=εk. All deformations inside the neighborhood should be allowed to compete in the atomic scale relaxation. Thus, the problem of energy minimization is reduced to a problem of, first, minimizing over deformations in the

Interpretation of the weak neighborhood

Since y and therefore ỹ are entirely determined by the nuclear positions, any condition on these deformations is interpretable in terms of nuclear positions. In particular, it is interesting to interpret the weak neighborhood (17) in this way. We begin with Eq. (13), and rewrite this condition in terms of y by using Eq. (6):yx1,x2,x3−kuk−1x1,k−1x2≤kℓ1ε.Hence, the actual atomic positions are uniformly close to a scaled family of larger and larger surfaces of the form kuk−1x1,k−1x2, related by

Passage to continuum theory

The atomic level theory shall rest on the assumption that the energy is expressible as a function of the nuclear positions. This is the fundamental setting delivered by quantum mechanics in the Born–Oppenheimer approximation after minimization over electronic states. (More precisely, the energy depends on the nuclear positions and their charges, i.e. the atomic species involved, but for the sake of simplicity we focus on a single species here.) In the present framework, in original and rescaled

Comments on the implied atomic scale calculation

Now we discuss briefly the atomic scale calculation implied by the right-hand side of Eq. (37). We work in original variables, this being the natural setting for atomic scale. Technically this calculation takes place near a deformed triangle but the shape of the set likely does not matter with suitable decay properties of E. First we note that if E is rotation and translation invariant, i.e.,ERyx+c;xLM×0,h=Eyx;xLM×0,hforallR∈SO3,cR3,then ϕ is similarly invariant:ϕRA,Rb1,…,Rbν−1A,b1,…,bν−1

Nanotubes

We make some brief remarks about the application of the scheme to other geometries. It can be seen from the above that the exact nature of the independent variables of the continuum theory will depend on the atomic structure of the object from which the condition of limited distortion is imposed. For definiteness we focus on a particular geometry. The most well-studied nanotube is certainly the carbon nanotube (Iijima, 1991), whose structure has been carefully elucidated by Amelinckx et al. (

Discussion

We have given a scheme for the direct passage from atomic to continuum theory applicable to cases in which one or more dimensions of the body are large. Of course, the scheme could be applied to the case in which all dimensions are large, but in this paper we discussed only the lower dimensional cases which have more interesting kinematics. The scheme is very much influenced by geometry, and the small parameter that leads to simplification is taken to be the ratio of length scales.

One could

Acknowledgements

This work was supported by AFOSR/MURI F49620-98-1-0433. The research also benefitted from the support of ARO DA/DAAG55-98-1-0335, NSF DMS-9505077 and ONR/DARPA N00014-95-1-1145 and -91-J-4034.

References (14)

There are more references available in the full text version of this article.

Cited by (99)

  • Mechanical properties of helically twisted carbyne fibers

    2020, International Journal of Mechanical Sciences
  • A concurrent multi-scale technique in modeling heterogeneous FCC nano-crystalline structures

    2015, Mechanics of Materials
    Citation Excerpt :

    Each time, the data is transferred to the coarse scale after being obtained from the fine scale and vice versa throughout the whole field. This sequential approach seems effective for systems in which the coarse scale variations may be considered homogeneous and quasi-static when standing in a small-scale view (Friesecke and James, 2000; Lu and Kaxiras, 2005). The concurrent method on the other hand, is a way to do well from the fact that in many cases not the entire domain demands equal resolution.

  • Foundational issues, analysis and geometry in continuum mechanics: introduction

    2023, Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences
View all citing articles on Scopus
View full text