Surface Effects in Solid Mechanics PDF

Advanced Structured Materials
Holm Altenbach
Nikita F. Morozov Editors
Surface Effects
in Solid Mechanics
Models, Simulations and Applications
Advanced Structured Materials
Volume 30
Series Editors
Andreas Öchsner
Lucas F. M. da Silva
Holm Altenbach
For further volumes:

http://www.springer.com/series/8611
Holm Altenbach Nikita F. Morozov
•
Editors
Surface Effects in Solid

Mechanics
Models, Simulations and Applications
123
Editors
Holm Altenbach Nikita F. Morozov
Lehrstuhl für Technische Mechanik St. Petersburg State University
Otto-von-Guericke-Universität Magdeburg St. Petersburg
Magdeburg Russia
Germany
ISSN 1869-8433 ISSN 1869-8441 (electronic)

ISBN 978-3-642-35782-4 ISBN 978-3-642-35783-1 (eBook)
DOI 10.1007/978-3-642-35783-1
Springer Heidelberg New York Dordrecht London
Library of Congress Control Number: 2013931510
Springer-Verlag Berlin Heidelberg 2013

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or
information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed. Exempted from this legal reservation are brief
excerpts in connection with reviews or scholarly analysis or material supplied specifically for the
purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the
work. Duplication of this publication or parts thereof is permitted only under the provisions of
the Copyright Law of the Publisher’s location, in its current version, and permission for use must always
be obtained from Springer. Permissions for use may be obtained through RightsLink at the Copyright
Clearance Center. Violations are liable to prosecution under the respective Copyright Law.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt
from the relevant protective laws and regulations and therefore free for general use.
While the advice and information in this book are believed to be true and accurate at the date of
publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for
any errors or omissions that may be made. The publisher makes no warranty, express or implied, with
respect to the material contained herein.
Printed on acid-free paper
Springer is part of Springer Science?Business Media (www.springer.com)

Preface
The collection of papers ‘‘Surface Effects in Solid Mechanics: Models, Simulations

and Applications’’ is devoted to the modeling of non-traditional effects in material
behavior which are related to surface phenomena. In the classical mechanics as
usual the properties of the bulk material are the focus of researchers. The problem
solution is based on the following elements:
• geometrical relations presenting, for example, the displacements, strains, strain
rates, etc.,
• equilibrium equation or equations of motion,
• constitutive equations (stress–strain relations among others), and
• evolution equations describing the evolution of inner processes like hardening or
damage.
The problem statement is finished, if the boundary and initial conditions are fixed.
After that using analytical, semi-analytical, or numerical methods problems of the
engineering practice can be solved.
With the miniaturizing of structures it is not enough to present only the bulk
properties of the materials. It is well known from the physics of materials or
material science that classical properties like the Young modulus changes dra-
matically if the size of the structure or specimen attains the nanometer range. For
an adequate description of such behavior several approaches are suggested. All of
them are based on the surface properties of the interface behavior or the lattice
properties, etc., which should be taken into account.
Various approaches are discussed within this book. Most of the contributions
were selected from lectures presented during the mini-symposium ‘‘Surface Effects
in Nano-Mechanics’’ at the European Solid Mechanics Conference, which was
held in Graz (Austria) in July 2012. The book contains 14 papers, which are
presented in alphabetical order. In the first paper some mathematical aspects of
initial-boundary and boundary-value problems for elastic bodies including surface
stresses are present. The mechanical properties of materials considering surface
effects are presented in the second paper. Graphene, which is a monolayer of
carbon atoms packed into a two-dimensional honeycomb lattice, in the simplest
v
vi Preface
case can be modeled as an elastic material. The relevant properties are presented
by Berinskii and Borodich. A comparison of atomistic and surface enhanced
continuum approaches at finite temperature is given in the next paper. After that, in
the next paper electro-elastic coupling is introduced and some special effects are
investigated. Plane problems and their solution based on the Goursat–Kolosov
complex potentials and Muskhelishvili’s technique are introduced in two papers
from Grekov’s group in St. Petersburg. The paper by Kutelova et al. is devoted to
some experimental observations related to surface effects. This contribution is
close to applications. In the paper of Nasedkin and Eremeyev the problem of
natural oscillations of piezoelectric bodies of nanosizes taking into account surface
stresses and electric charges is discussed. The stability and structural transition in
crystal lattices are studied in the paper of Podolskaya et al. An excellent survey on
mathematical modeling of phenomena caused by surface stresses in solids is given
by Povstenko. The buckling of a supported annular plate with a non-Euclidean
metric and made of graphene is studied with an atomistic and a continuum
mechanics approach in the paper of Schwartzbart and Steindl. The paper of
Ustinov et al. is devoted to peculiarities in describing surface and interface effects
in elasticity. Finally, the kinetics of chemical reaction fronts is presented in the last
paper.
Magdeburg, St. Petersburg, August 2012 Holm Altenbach

Nikita F. Morozov
Contents
Mathematical Study of Boundary-Value Problems of Linear

Elasticity with Surface Stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Holm Altenbach, Victor A. Eremeyev and Leonid P. Lebedev
On the Influence of Residual Surface Stresses on the Properties

of Structures at the Nanoscale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Holm Altenbach, Victor A. Eremeyev and Nikita F. Morozov
On the Isotropic Elastic Properties of Graphene Crystal Lattice . . . . . 33

Igor E. Berinskii and Feodor M. Borodich
A Comparison of Atomistic and Surface Enhanced

Continuum Approaches at Finite Temperature. . . . . . . . . . . . . . . . . . 43
Denis Davydov, Ali Javili, Paul Steinmann and Andrew McBride
Surface Mechanics and Full-Field Measurements: Investigation

of the Electro-Elastic Coupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Cécile Flammier, Frédéric Kanoufi, Sorin Munteanu, Jean Paul Roger,
Gilles Tessier and Fabien Amiot
Effect of a Type of Loading on Stresses at a Planar Boundary

of a Nanomaterial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Mikhail A. Grekov and Yulia I. Vikulina
Surface Stress in an Elastic Plane with a Nearly Circular Hole. . . . . . 81

Mikhail A. Grekov and Anna A. Yazovskaya
vii
viii Contents
Glass Spheres: Functionalization, Surface Modification

and Mechanical Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Zinaida Kutelova, Hendrik Mainka, Katja Mader,
Werner Hintz and Jürgen Tomas
Spectral Properties of Piezoelectric Bodies with Surface Effects . . . . . 105

Andrey V. Nasedkin and Victor A. Eremeyev
Stability and Structural Transitions in Crystal Lattices . . . . . . . . . . . 123

Ekaterina Podolskaya, Artem Panchenko and Anton Krivtsov
Mathematical Modeling of Phenomena Caused by Surface Stresses

in Solids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Yuriy Povstenko
Buckling of a Supported Annular Plate

with a Non-Euclidean Metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
Michael Schwarzbart and Alois Steindl
On the Modeling of Surface and Interface Elastic Effects

in Case of Eigenstrains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
Konstantin B. Ustinov, Robert V. Goldstein and Valentin A. Gorodtsov
On Kinetics of Chemical Reaction Fronts in Elastic Solids . . . . . . . . . 181

Elena N. Vilchevskaya and Alexander B. Freidin
Contributors
Holm Altenbach Lehrstuhl für Technische Mechanik, Institut für Mechanik,

Fakultät für Mechanik, Otto-von-Guericke-Universtät Magdeburg, Universität-
splatz 2, 39106 Magdeburg, Germany, e-mail: holm.altenbach@ovgu.de
Fabien Amiot FEMTO-ST, CNRS UMR 6174/UFC/ENSMM/UTBM, 24 chemin
de l’épitathe, 25000 Besançon, France, e-mail: fabien.amiot@femto-st.fr
Igor E. Berinskii Institute for Problems in Mechanical Engineering, Russian
Academy of Sciences, Bolshoy pr., V.O. 61, 199178 St. Petersburg, Russia, e-mail:
iberinsk@gmail.com
Feodor M. Borodich School of Engineering, Cardiff University, Cardiff CF24
3AA, UK, e-mail: borodichfm@cardiff.ac.uk
Denis Davydov Chair of Applied Mechanics, University of Erlangen-Nuremberg,
Egerlandstr. 5, 91058 Erlangen, Germany, e-mail: Denis.Davydov@ltm.uni-
erlangen.de
Victor A. Eremeyev Lehrstuhl für Technische Mechanik, Institut für Mechanik,
Fakultät für Mechanik, Otto-von-Guericke-Universtät Magdeburg, Universität-
splatz 2, 39106 Magdeburg, Germany; South Scientific Center of RASci and South
Federal University, Milchakova St. 8a, 344090 Rostov on Don, Russia, e-mail:
eremeyev.victor@gmail.com
Cécile Flammier FEMTO-ST, CNRS UMR 6174/UFC/ENSMM/UTBM, 24
chemin de l’épitathe, 25000 Besançon, France, e-mail: cecile.flammier@femto-st.fr
Alexander B. Freidin Institute for Problems in Mechanical Engineering, Russian
alexander.freidin@gmail.com
Robert V. Goldstein A. Ishlinsky Institute for Problems in Mechanics RAS,
Prospect Vernadskogo 101, 119526 Moscow, Russia, e-mail: goldst@ipmnet.ru
ix
x Contributors
Valentin A. Gorodtsov A. Ishlinsky Institute for Problems in Mechanics RAS,

Prospect Vernadskogo 101, 119526 Moscow, Russia, e-mail: ustinov@ipmnet.ru;
goldst@ipmnet.ru; gorod@ipmnet.ru
Mikhail A. Grekov Faculty of Applied Mathematics and Control Processes,
Saint-Petersburg State University, Universitetski pr. 35, 198504 St. Petersburg,
Russia, e-mail: magrekov@mail.ru
Werner Hintz Department of Process Engineering, Mechanical Process Engi-
neering, Otto-von-Guericke-University Magdeburg, Universitätsplatz 2, 39106
Magdeburg, Germany, e-mail: werner.hintz@ovgu.de
Ali Javili Chair of Applied Mechanics, University of Erlangen-Nuremberg,
Egerlandstr. 5, 91058 Erlangen, Germany, e-mail: Ali.Javili@ltm.uni-erlangen.de
Frédéric Kanoufi PECSA, ESPCI ParisTech/CNRS UMR 7195, 10 rue
Vauquelin, 75005 Paris, France, e-mail: frederic.kanoufi@espci.fr
Anton M. Krivtsov Institute for Problems in Mechanical Engineering, Russian
Academy of Sciences, Bolshoy pr., V.O. 61, 199178 St. Petersburg, Russia;
St. Petersburg State Polytechnical University, 29 Politekhnicheskaya st., 195251
St. Petersburg, Russia, e-mail: akrivtsov@bk.ru
Zinaida Kutelova Department of Process Engineering, Mechanical Process
Engineering, Otto-von-Guericke-University Magdeburg, Universitätsplatz 2,
39106 Magdeburg, Germany, e-mail: zinaida.kutelova@ovgu.de
Leonid P. Lebedev Universidad Nacional de Colombia, Cr. 45, # 2685, Bogotá
D.C., Colombia, e-mail: llebedev@unal.edu.co
Katja Mader Department of Process Engineering, Mechanical Process Engi-
Magdeburg, Germany, e-mail: katja.mader@ovgu.de
Hendrik Mainka Department of Process Engineering, Mechanical Process
Engineering, Otto-von-Guericke-University Magdeburg, Universitätsplatz 2,
39106 Magdeburg, Germany, e-mail: hendrik.mainka@ovgu.de
Andrew McBride CERECAM, University of Cape Town, Rondebosch 7701,
South Africa, e-mail: andrew.mcbride@uct.ac.za
Nikita F. Morozov St. Petersburg State University, Bibliotechnaya sq. 2, 198904
St. Petersburg, Russia, e-mail: morozov@NM1016.spb.edu
Sorin Muntenau PECSA, ESPCI ParisTech/CNRS UMR 7195, 10 rue Vauqu-
elin, 75005 Paris, France, e-mail: sorin.munteanu@espci.fr
Andrey V. Nasedkin Southern Federal University, Miltchakova str., 8a, 344090
Rostov on Don, Russia, e-mail: nasedkin@math.sfedu.ru
Contributors xi
Artem Panchenko St. Petersburg State Polytechnical University, 29 Polite-

khnicheskaya st., 195251
St. Petersburg, Russia, e-mail: ArtemQT@yandex.ru
Ekaterina Podolskaya Institute for Problems in Mechanical Engineering, Russian
Academy of Sciences, Bolshoy pr., V.O. 61, 199178 St. Petersburg, Russia;
St. Petersburg State Polytechnical University, 29 Politekhnicheskaya st., 195251
St. Petersburg, Russia, e-mail: katepodolskaya@gmail.com
Yuriy Povstenko Institute of Mathematics and Computer Science, Jan Długosz
University Cze˛stochowa, Armii Krajowej 13/15, 42-200 Czestochowa, Poland,
e-mail: j.povstenko@ajd.czest.pl
Jean Paul Roger Institut Langevin, ESPCI ParisTech/CNRS UMR 7587, 10 rue
Vauquelin, 75005 Paris, France, e-mail: jean-paul.roger@espci.fr
Michael Schwarzbart Department of Applied and Numerical Mechanics, Uni-
versity of Applied Sciences Wiener Neustadt, Johannes Gutenberg-Straße 3, 2700
Wiener Neustadt, Austria, e-mail: michael.schwarzbart@fhwn.ac.at
Alois Steindl Institute of Mechanics and Mechatronics, Vienna University of
Technology, Wiedner Hauptstrasse 8-10 325/2, 1040 Vienna, Austria, e-mail:
alois.steindl@tuwien.ac.at
Paul Steinmann Chair of Applied Mechanics, University of Erlangen-Nuremberg,
Egerlandstr. 5, 91058 Erlangen, Germany, e-mail: Paul.Steinmann@ltm.uni-
erlangen.de
Gilles Tessier Institut Langevin, ESPCI ParisTech/CNRS UMR 7587, 10 rue
Vauquelin, 75005 Paris, France, e-mail: tessier@optique.espci.fr
Jürgen Tomas Department of Process Engineering, Mechanical Process Engi-
Magdeburg, Germany, e-mail: juergen.tomas@ovgu.de
Yulia I. Vikulina Faculty of Applied Mathematics and Control Processes,
St. Petersburg State University, Universitetski pr. 35, 198504 St. Petersburg,
Russia, e-mail: julia.vikulina@gmail.com
Elena N. Vilchevskaya Institute for Problems in Mechanical Engineering, Russian
vilchevska@gmail.com
Anna A. Yazovskaya Faculty of Applied Mathematics and Control Processes,
Saint-Petersburg State University, Universitetski pr. 35, 198504 St. Petersburg,
Russia, e-mail: yazovskaya_aa@mail.ru
Mathematical Study of Boundary-Value
Problems of Linear Elasticity with Surface
Stresses
Holm Altenbach, Victor A. Eremeyev and Leonid P. Lebedev
Abstract Following [1, 2] a mathematical investigation of initial-boundary and

boundary-value problems of statics, dynamics and natural oscillations for elastic bod-
ies including surface stresses is presented. The weak setup of the problems based on
mechanical variational principles is given with introducing of corresponding energy
spaces. Theorems of uniqueness and existence of the weak solution in energy spaces
of static and dynamic problems are formulated and proved. The studies are performed
applying the functional analysis techniques. Solutions of the problems under con-
sideration are more smooth on the boundary surface than solutions of corresponding
problems of the classical linear elasticity. The weak setup of the eigen-value prob-
lems is based on the Rayleigh variational principle. Certain spectral properties are
established for the problems under consideration. In particular, bounds for the eigen-
frequencies of an elastic body with surface stresses are presented. These bounds
demonstrate increases in both the rigidity of the body and of the eigenfrequencies
over those of the body with surface stresses neglected. The considered weak state-
ments of the initial and boundary problems constitute the mathematical foundation
for some numerical methods, in particular, for the finite element method.
H. Altenbach (B) · V. A. Eremeyev

Institut für Mechanik, Fakultät für Maschinenbau, Otto-von-Guericke-Universität Magdeburg,
Universitätsplatz 2, 39106 Magdeburg, Germany
e-mail: holm.altenbach@ovgu.de
V. A. Eremeyev
South Scientific Center of RASci and South Federal University, Rostov on Don, Russia
e-mail: victor.eremeyev@ovgu.de; eremeyev.victor@gmail.com
L. P. Lebedev
Universidad Nacional de Colombia, Cr. 45, # 2685,Bogotá D.C., Colombia
e-mail: llebedev@unal.edu.co
H. Altenbach and N. F. Morozov (eds.), Surface Effects in Solid Mechanics, 1

Advanced Structured Materials 30, DOI: 10.1007/978-3-642-35783-1_1,
© Springer-Verlag Berlin Heidelberg 2013
2 H. Altenbach et al.
1 Introduction
For last years, the interest to the model of elastic solids with surface stresses proposed
by Gurtin and Murdoch [9] is rapidly growing with respect to nanomechanics, see
the review papers [7, 20]. The model [9] can predict the size effect observed in the
case of nanosized materials [19]. From the mechanical point of view the Gurtin–
Murdoch model describes an elastic body with an elastic membrane glued to the
body surface. The model was generalized by Steigmann and Ogden [18] taking into
account the bending stiffness of the elastic film glued to the surface. Generalizations
of the surface elasticity are summarized in [15] where, for example, the surface
couple stresses are considered. This means that the attached to the body elastic film
possesses properties of the two-dimensional Cosserat continuum. Another Cosserat-
type model of interfaces in solids is discussed in [16]. Mathematical studies of the
boundary-value problems of linear elasticity with surface stresses are provided in
[1, 2, 17].
The paper is organized as follows. In Sect. 2 we recall the basic equations of
linear elasticity with surface stresses. Here we also formulate the Lagrange varia-
tional principle. Section 3 presents the proof of the theorems on existence of weak
solutions of statics, eigenoscillations and dynamics of solids with surface elasticity.
This section is based on the results of [1]. Following [2], in Sect. 4 we analyze the
spectrum of the oscillation boundary-value problems in details. We obtain upper and
lower bounds for the eigenfrequencies for a body with surface stresses in terms of two
corresponding problems for the same body with the free and clamped boundaries.
Radial eigenoscillations of an elastic sphere with surface stresses are considered in
Sect. 5.
2 Basic Relations of Linear Elasticity with Surface Stresses
First we consider the problems with mixed boundary conditions. So
∇ · σ + ρf = ρ ü, x ∈ V,
u|Ωu = 0, n · σ |Ωt = t0 , n · σ |Ω S = t, t ≡ t0 + t S , (1)
x ∈ ∂ V ≡ Ω = Ωu ∪ Ωt ∪ Ω S ,
where σ is the stress tensor, ∇ is the 3D gradient operator (3D nabla operator), ρ
denotes the body density, f is the density of the volume forces, n is the external unit
normal to Ω, and the overdot denotes the partial derivative with respect to time t.
The surface stress vector t S is defined by
∂
tS = ∇S · τ , ∇S = ∇ − n , (2)
∂z
Mathematical Study of Boundary-Value Problems 3
Fig. 1 Elastic body with t t0 tS

surface stresses
V
z
x t0
ΩS
Ωt i3
Ωu n
0 i2
i1
where τ is the surface stress tensor on Ω S and z is the coordinate along the normal
to Ω S , see Fig. 1.
Next we consider the problem when the static conditions are given on the whole
boundary
n · σ |Ω = t, x ∈ Ω. (3)
For simplicity, we restrict ourselves to an isotropic material. The constitutive

equation for the material is given by the Hooke law
1
σ = 2με + λItr ε with ε = ε(u) ≡ ∇u + (∇u)T . (4)
2
For the surface stresses we assume the following constitutive equation
1
τ = 2μ S ε̃ε + λ S Atr ε̃ε with ε̃ε = ε̃ε (v) ≡ ∇ S v · A + A · (∇ S v)T , (5)
2
where v is the displacement of the film point x of Ω S . Here I and A ≡ I − n ⊗ n
are the three- and two-dimensional unit tensors, respectively, λ and μ are Lamé’s
coefficients of the bulk material whereas λ S and μ S are the elastic characteristics of
the surface film Ω S (they are the surface analogues of Lamé’s moduli), ε is the small
strain tensor, u is the displacement vector, and ε̃ε is the surface strain tensor. We use
the non-separation condition
u|Ω S = v.
In equilibrium, the dynamic equation (1) changes to
∇ · σ + ρf = 0. (6)
Thus the equilibrium boundary value problem for an elastic body with surface stresses
consists of Eq. (6) given in V and the boundary conditions
u|Ωu = 0, n · σ |Ωt = t0 , (n · σ − ∇ S · τ )|Ω S = t0 , (7)
where σ and τ satisfy relations (4) and (5), respectively. In Eq. (5) we set
1
ε̃ε = ε̃ε (u) ≡ ∇ S u · A + A · (∇ S u)T Ω .
2 S
If Ω S = Ω then parts Ωu and Ωt are absent. In what follows for simplicity we

consider two cases: Ω = Ω S and Ω = Ωu ∪ Ω S .
For the elastic body with surface stresses the Lagrange variational principle can
be suggested:
Theorem 1.1 A stationary point u of J (u) = E(u) − A(u) on the set of admissible
sufficiently smooth displacements, that is they satisfy the condition uΩ = 0, Ωu =
u
∅, is a solution of the equilibrium equations (6) for the elastic body in the volume V
together with the boundary conditions (7), and vice versa. Here

E(u) = W (ε) d V + U (ε̃ε ) dΩ, A(u) = ρf · u d V + t0 · u dΩ,
V ΩS V ΩS (8)
1 2 1
W (ε) ≡ λtr ε + με : ε, U (ε̃ε ) ≡ λ S tr 2ε̃ε + μ S ε̃ε : ε̃ε ,
2 2
where W, U are the strain energy of the isotropic elastic body and the surface strain
energy, respectively, sign: means the scalar product of two second-order tensors,
that is α : β = tr (α · β T ).
Proof The stationary condition for J leads to the variational equation

δ J ≡ (σ : δε − ρf · δu) d V + (τ : δε̃ε − t0 · δu) dΩ = 0 (9)
V ΩS
for any sufficiently smooth δu such that δu|Ωu = 0. Applying the Gauss–Ostrograd-
sky theorem we get

τ : δε̃ε dΩ = τ : (∇ S δu) dΩ = −
T
(∇ S ·τ )· δu dΩ + ν ·τ · δu ds,
ΩS ΩS ΩS Γ
where ν is the external unit normal to the boundary contour Γ of Ω S , ν lies in the
tangent plane to Ω, that is ν · n = 0. On Ωu admissible displacement δu = 0 so by
continuity, δu = 0 on Γ as well. If Ω S = Ω the contour integral is absent. Thus the
following formula for the first variation of J holds

δ J = − (∇ · σ + ρf) · δu d V + (n · σ − ∇ S · τ − t0 ) · δu dΩ. (10)
V ΩS
From Eq. (9) we derive Eqs. (6) and (7); when Ω S = Ω, condition (7) changes to
(3).
To prove the second part of the theorem we suppose u to be a solution of the
problem (6), (7). Dot-multiplying (6) by δu and integrating the result over V we then
repeat the above transformations in the reverse order. This brings us to the necessary
variational equation δ J = 0.
Suppose W and U to be positive definite functions of their arguments, that means
there exist positive constants c1 , c2 such that
W (ε) ≥ c1 ε : ε, U (ε̃ε ) ≥ c2ε̃ε : ε̃ε , ∀ε, ε̃ε , c1 , c2 > 0 (11)
For an isotropic homogeneous material this leads to the inequalities, see [1],
3λ + 2μ > 0, μ > 0, λ S + μ S > 0, μ S > 0. (12)
We should note that if U can be negative for some deformations then it can be
shown that E is not bounded from below.
The admissible range of surface material parameters is also studied in [11].

Let u◦ be a solution of the problem (6), (7). Then J (u) can be represented in the
form J (u) = J (u − u◦ ) − J (u◦ ). From this and the positivity assumptions (11) it
follows:
Theorem 1.2 Let the assumptions (11) hold and Ωu = ∅. On the set of admissible
smooth displacements u satisfying u|Ωu = 0, a stationary point of J (u) is a point
of minimum of J (u).
3 Existence Theorems of Weak Solutions
3.1 Statics
Let us introduce the energy space in which we will seek a weak solution. We start
(2)
with the set C0 of vector functions u that take values in IR 3 such that each of its
Cartesian components belongs to C (2) (V̄ ) and u|Ωu = 0. On C(2)
0 we introduce the
inner product

u, v e = [λtr ε(u)tr ε(v) + 2με(u) : ε(v)] d V
V

+ λ S tr ε̃ε (u)tr ε̃ε (v) + 2μ S ε̃ε (u) : ε̃ε (v) dΩ (13)
ΩS
and the corresponding energy norm u2e = u, u e that is the double strain energy
(2)
of the body 2E(u). Suppose the conditions (11) are fulfilled. Then on C0 the form
(2)
u, v e possesses all the properties of the inner product. However, C0 with this inner
product is not complete.
(2)
Definition 1.1 The completion of C0 with respect to the norm ue is called the
energy space E.
By Korn’s inequality on C(2)

0 , it is easily seen that
uW 1,2 (V ) ≤ C1 ue
with some constant C1 that does not depend on u ∈ E. To prove this inequality, we
should suppose some regularity of the boundary of V , for example, it should satisfy
the cone property. So we can use the properties of the Sobolev space W 1,2 (V ) for
the elements of space E. Note that the Sobolev norm for the vector functions on the
domain V is defined by the formula

u2W 1,2 (V ) = |u|2 + ∇u : ∇uT d V.
V
Let Ω S consists of a finite number of sufficiently smooth surfaces with smooth

boundaries. Using the techniques of the shell theory [4] it is easy to prove an analogue
of Korn’s inequality on Ω S ,
uW 1,2 (Ω S ) ≤ C2 ue
with some constant C2 that does not depend on u ∈ E. Here

u2W 1,2 (Ω = |u|2 + ∇ S u : ∇ S uT dΩ.
S)
ΩS
The element u of E is an element of W 1,2 (V )3 and so by the trace theorem,

it belongs to W 1/2,2 (Ω S )3 . However, the presence of the member given on
Ω S in the norm of E implies that u ∈ E possesses higher regularity, namely
u ∈ W 1,2 (Ω S )3 .
Definition 1.2 A weak solution of the problem (6), (7) is u ∈ E that satisfies the
equation

δJ ≡ (σ : δε − ρf · δu) d V + (τ : δε̃ε − t0 · δu) dΩ = 0
V ΩS
for any δu ∈ E.
Let us formulate:
Theorem 1.3 Let f ∈ L 6/5 (V )3 and t0 ∈ L p (Ω S )3 for some p > 1. There exists a
weak solution of the problem (6), (7) by Definition 1.2 that is unique.
Proof By the properties of the Sobolev space W 1,2 , functional A(δu) is linear and
continuous in E; it can be proved in a similar manner to the proof of this fact for
the work functional in linear elasticity, cf. [3, 8, 14]. By the Riesz representation
theorem for a linear continuous functional in a Hilbert space we have
A(δu) = u0 , δu e ,
where u0 is a uniquely defined element of E. So Eq. (9) reduces to the equation
u, δu e = u0 , δu e ,
which holds for any δu ∈ E. This equation has the unique solution u = u0 that
completes the proof.
3.1.1 Existence Theorem: Free Boundary
Using the techniques for the equilibrium problem for a free body under load in
the classic linear elasticity, we can prove the following theorem for a body free of
geometrical constraints. Now Ω S = Ω.
Theorem 1.4 Let f ∈ L 6/5 (V )3 and t0 ∈ L p (Ω S )3 for some p > 1 and the external
forces be self-balanced, that is

f dV + t0 dΩ = 0, x × f dV + x × t0 dΩ = 0.
V Ω V Ω
There exists a weak solution of the problem (6), (7) by Definition 1.2 that is unique
up to the infinitesimal rigid body motions a + x × b, where a, b are constant vectors
and x is the position vector of a body point.
3.1.2 On the Finite Element Method
To solve the equilibrium problems of solids with surface stresses the method of
finite elements is applied, see [10, 12, 13]. We briefly discuss some features of the
conforming version of the method [5]. In this version we generate the sequence of
the finite-dimensional subspaces Eh of the space E and find the approximate FEM
solutions uh ∈ Eh . By Céa’s lemma we have the bound for the error
u − uh e ≤ C inf u − uh e ,
uh ∈ Eh
where u ∈ E is the weak solution, and C is a constant. This estimate depends on the
error of the approximation in the volume and on the approximation error on surface
ΩS .
3.2 Eigenoscillations
Let us consider an important case of dynamical problems: the eigenoscillation prob-

lem for an elastic body with surface stresses. In this problem we seek solutions to
the homogeneous dynamic equation in displacements with f = 0, t0 = 0 in the
following form:
u = u(x, t) = w(x)eiωt .
Substituting this into (1), (7) expressed in displacements and canceling the factor
eiωt , we get

∇ · σ = −ρω2 w in V, wΩ = 0, (n · σ − ∇ S · τ )Ω = 0. (14)
u S
In the last equations σ and τ are given by the relations (4) and (5) with ε = ε(w)
and ε̃ε = ε̃ε (w).
In this case we have:
Theorem 1.5 For the oscillation eigenmodes w1 and w2 corresponding to distinct
eigenfrequencies ω1 and ω2 , respectively, the relation

ρw1 · w2 d V = 0 (15)
V
holds true. Moreover,

w1 , w2 e = 0. (16)
Equation (15) is called the orthogonality relation, and (16)—the generalized orthog-
onality relation for w1 and w2 .
Proof Let eigenmodes w1 and w2 correspond to different eigenfrequencies, so that

∇ · σ 1 = −ρω12 w1 inV, w1 Ω = 0, n · σ 1 Ω = ∇ S · τ 1 ,
u S

∇ · σ 2 = −ρω22 w2 inV, w2 Ω = 0, n · σ 2 Ω = ∇ S · τ 2 .
u S
Applying the Gauss–Ostrogradsky theorem we get


− σ 1 : ε2 d V − τ 1 : ε̃ε 2 dΩ + ω12 ρw1 · w2 d V = 0, (17)
V ΩS V

− σ 2 : ε1 d V − τ 2 : ε̃ε 2 dΩ + ω22 ρw2 · w1 d V = 0. (18)
V ΩS V

Subtracting these two equations, we get (ω12 − ω22 ) V ρw1 · w2 d V = 0.
Theorem 1.6 The eigenvalue problem under consideration has a discrete spectrum
which contains only eigenfrequencies ωk that possess the following properties
• All ωk are positive, ωk ≥ ω > 0.
• The set {ωk } is infinite and does not contain a finite accumulation point.
• To each ωk there corresponds no more than a finite number of linearly independent
eigensolutions which are assumed to be orthonormalized.
• The set of all these eigenmodes {wk } is a complete orthonormal system in the
energy space E. Besides the set {ωk wk } is a complete orthonormal system in
L 2 (V )3 with the scalar product

(w, v) = ρw · v d V.
V
3.3 Dynamics
In the theory under consideration the initial boundary value problem is given by the
equations
∇ · σ + ρf = ρ ü in V, (19)
u|Ωu = 0, (n · σ − ∇ S · τ )|Ω S = 0,
u|t=0 = u0 , u̇|t=0 = v0 .
Here we consider the weak setup of the problem as well. It differs from the static
statement as the displacements depend on time t. First we get

(σ : δε − ρf · δu) d V + τ : δε̃ε dΩ = − ρ ü · δu d V. (20)
V ΩS V
Now we integrate this with respect to t over [0, T ] and then integrate by parts in the
term containing ü. Using the additional condition δu|t=T = 0 we get the following
equation that is used for the weak formulation of the problem:
T T T
u, δu e dt = ρf · δu d V dt + ˙ d V dt +
ρ u̇ · δu ρv0 · δu d V.
0 0 V 0 V V
(21)
We introduce the following definitions:
Definition 1.3 The energy space E(0, T ) is the completion of the set of all vector
functions u(r, t) ∈ (C (2) (V × [0, T ]))3 that vanish on Ωu with respect to the norm
induced by the energy scalar product
T T
u, δu E(0,T ) = u, δu e dt + ˙ d V dt.
ρ u̇ · δu
0 0 V
D0T is the subspace of E(0, T ) that is the result of the completion of the subset
(C (2) (V × [0, T ]))3 of the vector functions that are zero on Ωu and take zero value
at t = T .
The set of the elements of E(0, T ) can be considered as a subspace of (W (1,2) (V ×
[0, T ]))3 with some additional properties on the surface Ω S .
Definition 1.4 u ∈ E(0, T ) is called a generalized solution of the dynamic problem
of the body with surface stresses if it satisfies Eq. (21) with any δu ∈ D0T and the
initial condition u|t=0 = u0 in the sense of L 2 (V )3 , that is

|u(r, 0) − u0 (r)|2 d V = 0.
V
Using these definitions we prove the following theorem:

Theorem 1.7 Suppose that
• u0 ∈ E and u0 = 0 on Ωu ,
• v0 ∈ (L 2 (V ))3 , and f ∈ (L 2 (V × [0, T ]))3 .
There exists (in the sense of Definition 1.4), a generalized solution to the dynamic
problem for the elastic body with surface stresses, and it is unique.
4 Spectrum and Stiffness of an Elastic Body

with Surface Stresses
Let us consider three boundary-value problems. The main problem is the following.
• Problem Pss :

∇ · σ = −ρω2 w, x ∈V ; wΩ = 0, (n · σ − ∇ S · τ )Ω = 0. (22)
u S
The spectrum of Problem Pss will be compared with the spectra of two following
problems:
• Problem P f :
∇ · σ = −ρω2 w, x ∈ V ; w|Ωu = 0, n · σ |Ω S = 0; (23)

and
• Problem P0 :
∇ · σ = −ρω2 w, x ∈ V ; w|Ω = 0. (24)
The weak setup of the eigenvalue problems consist of the following equations.
• For Problem Pss :

wk , v E = ωk2 ρwk · v d V ∀v ∈ E.
V
• For Problem P f :

wk , v H ≡ [λtr ε(wk )tr ε(wk ) + 2με(wk ) : ε(v)] d V = ωk2 ρwk · v d V
V V
(25)
∀v ∈ H.
• For Problem P0 :

wk , v H = ωk2 ρwk · v d V ∀v ∈ H0 . (26)
V
Let us note that we may consider E as a subset of H, and H0 as a subset of E
H0 ⊂ E ⊂ H.
4.1 Rayleigh Variational Principle
On the orthogonality of the eigensolutions established in Theorem 1.5 the Rayleigh

variational principle can be grounded.
Theorem 1.8 Eigenmodes are stationary points of the energy functional

E(w) = W (ε(w)) d V + U (ε̃ε (w)) dΩ
V ΩS
on the set of displacements satisfying the boundary conditions w|Ωu = 0 and subject
to the constraint
1
ρw · w d V = 1. (27)
2 V
Conversely, all the stationary points of E(w) on the above set of the displacements
are eigenmodes of the body that correspond to its eigenfrequencies.
We will change the form of Rayleigh’s principle to the one which is frequently
used in applications. In the new formulation, the integral restrictions on the set of w
are not stipulated.
Theorem 1.9 On the set of admissible vector-functions satisfying the condition
w|Ωu = 0, the oscillation eigenmodes are stationary points of the functional

E(w) 1
R(w) = with K (w) = ρw · w d V.
K (w) 2 V
R(w) is called Rayleigh’s quotient. Conversely, a stationary point of R(w) is an

eigenmode that corresponds to an eigenfrequency; the value of R(w) on an eigen-
mode is a squared eigenfrequency:
R(w) = ω2 . (28)
Property (28) allows us to estimate the eigenvalues using some approximation for
the eigenmodes.
4.2 The Least Eigenfrequency
Let R(w) be Rayleigh’s quotient for the body with surface stresses:

w2E
R(w) = , K (w) = ρw · w d V.
K (w) V
The squared least eigenfrequency ωmin is determined as the infimum of R(w):
ωmin
2
= inf R(w).
w∈E
For Problems P f and P0 , Rayleigh’s quotient coincides with the quotient in linear
elasticity, i.e.,
w2H
R0 (w) = .
K (w)
f
For Problem P f , the squared least eigenfrequency, (ωmin )2 , is the infimum of R0 (w)
over H. For Problem P0 , the squared least eigenfrequency, (ωmin ◦ )2 , is the infimum
of R0 (w) over H0 .
The properties of the spaces E, H, H0 and the functionals R, R0 allow us to prove
the following
Theorem 1.10 The least eigenfrequency of a bounded elastic body with surface
stresses (Problem Pss ) is not less than the least eigenfrequency for the same body with
free boundary Ω S (Problem P f ), and it is not greater than the least eigenfrequency
for the same body with fixed boundary (Problem P0 ):
f ◦
ωmin ≤ ωmin ≤ ωmin . (29)
Proof The proof follows from the following inequality chains:
f
(ωmin )2 = inf R0 (w) ≤ inf R0 (w) ≤ inf R(w) = (ωmin )2 ,
w∈H w∈E⊂H w∈E
◦
(ωmin )2 = inf R(w) ≤ inf R(w) = inf R0 (w) = (ωmin )2 .
w∈E w∈H0 ⊂E w∈H0

From the proof of Theorem 1.10 there follow two results.
f
Corollary 1.1 The equality ωmin = ωmin holds if and only if U (ε̃ε (wmin )) = 0 on
ΩS .
By positive definiteness, U = 0 if and only if ε̃ε (wmin ) = 0 on Ω S . Hence the
displacement wmin of Ω S describes an infinitesimal isometric deformation of Ω S . In
particular, ε̃ε = 0 if wmin describes a rigid body motion.
◦
Corollary 1.2 The equality ωmin = ωmin holds if and only if wmin = 0 on Ω S ,
which is when wmin ∈ H0 .
Cases when ε̃ε (wmin ) = 0 or wmin = 0 on Ω S are rare. Hence, in general we can
expect the strict inequalities
f ◦
ωmin < ωmin < ωmin .
The least eigenfrequency ωmin depends on λ S and μ S . An increase in the surface

elastic moduli implies an increase in the least eigenfrequency of Problem Pss . Indeed,
let us consider two bodies of equal shape and equal internal moduli λ and μ, but with
different values of λ S and μ S . Denote the surface moduli of the bodies by λ(1)
S , μS
(1)
and λ(2) (2) (1)

S , μ S , respectively. Denote the least eigenfrequencies of the bodies by ωmin
(2)
and ωmin , respectively.
Theorem 1.11 Let
(1) (2) (1) (1) (2) (2)
0 < μS ≤ μS , 0 < λS + μS ≤ λS + μS . (30)
Then
(1) (2)
ωmin ≤ ωmin . (31)
Proof The proof follows immediately from the inequality R(1) (w) ≤ R(2) (w),
where R(α) (w), α = 1, 2, are Rayleigh’s quotients for the bodies. Indeed, from
(30) it follows that U(1) ≤ U(2) , where
1 (α) 2 (α)
U(α) = λ tr ε̃ε + μ S ε̃ε : ε̃ε , α = 1, 2.
2 S
The corresponding energy spaces E(1) and E(2) for the problems coincide up to the
form of the energy norms, which are equivalent, and so the infimum is taken over
the same set of elements E(1) = E(2) = E. Hence

(1) 2 (2) 2
ωmin = inf R(1) (w) ≤ inf R(2) (w) = ωmin .
w∈E w∈E

The least eigenfrequency depends continuously on λ S and μ S :
Theorem 1.12 For any number ε > 0, there exists a number δ > 0 such that
(1) (2)
|ωmin − ωmin | ≤ ε,
whenever |μ(1) (2) (1) (2)

S − μ S | ≤ δ and |λ S − λ S | ≤ δ.
Corollary 1.3 The least eigenfrequency of a bounded elastic body with surface
stresses tends to the least eigenfrequency for the same body with free boundary Ω S :
f
ωmin → ωmin as λ S → 0 and μ S → 0.
4.3 Higher Eigenfrequencies: Courant’s Minimax Principle
1. Denote by H(k) the subspace of H spanned by k − 1 arbitrary chosen elements

(k)
v1 , v2 , . . . , vk−1 of H, k > 1. The space H⊥ is its “orthogonal” complement in
H:
(k)
H⊥ = {w ∈ H | w, v1 L = w, v2 L = · · · = w, vk−1 L = 0} ,
where
u, v L = ρu · v d V.
V
(k) (k) we denote the subset of elements of H(k)

H⊥ is a closed subspace of H. By H ⊥ ⊥
with the constraint w, w L = 1, i.e.,
(k) = {w ∈ H(k) | w, w L = 1}.

H⊥ ⊥
2. Define
dss [v1 , v2 , . . . , vk−1 ] = inf

w2E ,
(k)
H E
⊥
d f [v1 , v2 , . . . , vk−1 ] = inf w2H ,

(k)
H ⊥
d0 [v1 , v2 , . . . , vk−1 ] = inf

w2H .
(k)
H H0
⊥
3. Repetition of the proof of Courant’s principle [6, pp. 406–407], shows that
by taking the supremum of these quantities over all possible combinations
v1 , v2 , . . . , vk−1 in H, we obtain the following eigenfrequencies.
For Problem Pss : ωk2 = sup dss [v1 , v2 , . . . , vk−1 ].
v1 ,...,vk−1
f2
For Problem P f : ωk = sup d f [v1 , v2 , . . . , vk−1 ].
v1 ,...,vk−1
For Problem P0 : ωk◦ 2 = sup d0 [v1 , v2 , . . . , vk−1 ].
v1 ,...,vk−1
Finally we obtain
Theorem 1.13 Let ωk be eigenfrequencies of a bounded elastic body with surface
stresses enumerated in increasing order as ω0 ≤ ω1 ≤ ω2 , . . ., and let ωk and ωk◦
f
be correspondingly ordered eigenfrequencies of the elastic body with free boundary

Ω S and with fixed boundary, respectively. Then
ωk ≤ ωk ≤ ωk◦ , k = 1, 2, 3, . . . .
f
(32)
Proof First we prove the left-hand inequality of (32). Since wH ≤ wE , we
have
d f [v1 , v2 , . . . , vk−1 ] ≤ inf w2E ≤ dss [v1 , v2 , . . . , vk−1 ].
(k)
H ⊥
Thus, for the greatest values of d f and dss we obtain
sup d f [v1 , v2 , . . . , vk−1 ] ≤ sup dss [v1 , v2 , . . . , vk−1 ].

v1 ,...,vk−1 v1 ,...,vk−1
f2
We conclude that ωk ≤ ωk2 . The right-hand inequality of (32) is proved in a similar
manner.
Proof When w ∈ H0 , we see that wE = wH . So we have
d0 [v1 , v2 , . . . , vk−1 ] = inf

w2H = inf

w2E .
(k)
H H0 (k)
H H0
⊥ ⊥
But
inf

w2E ≥ inf w2E .

(k)
H H0 (k)
H
⊥ ⊥
So
d0 [v1 , v2 , . . . , vk−1 ] ≥ dss [v1 , v2 , . . . , vk−1 ].
Thus
ωk2 = sup dss [v1 , v2 , . . . , vk−1 ] ≤ sup d0 [v1 , v2 , . . . , vk−1 ] = ωk◦ 2 .

v1 ,...,vk−1 v1 ,...,vk−1

As in the case of the least eigenfrequency higher eigenfrequencies depend con-
tinuously on elastic moduli.
(1)
Theorem 1.14 Let ωk be eigenfrequencies of a bounded elastic body with moduli
(1) (1) (1) (1) (1)
λ, μ and surface elastic moduli λ S , μ S , ordered as ω0 ≤ ω1 ≤ ω2 , . . .. Let
(2)
ωk be the ordered eigenfrequencies for the elastic body with moduli λ, μ but with
(2) (2)
surface moduli λ S , μ S . Let
(1) (2) (1) (1) (2) (2)

μS ≤ μS , λS + μS ≤ λS + μS .
Then
(1) (2)
ωk ≤ ωk for k = 1, 2, 3, . . . . (33)
5 Radial Oscillations of an Elastic Sphere with Surface Stresses
To illustrate the spectral properties of Problems Pss , P f , and P0 , we consider the

oscillations of an elastic sphere. The displacement field is
w = w(r )er , w(0) = 0. (34)
For the radially symmetric problem, the point r = 0 corresponds to Ωu . Substituting

w(r ) = r f (r ) into Eq. (22)1 , we reduce the equation to
4 ρω2
f + f + η2 f = 0, where η2 = . (35)
r λ + 2μ
The solution of Eq. (35) satisfying (34) is
ηr cos ηr − sin ηr
f (r ) = . (36)
r3
Then the radial component of the stress tensor σ is
w(r )
σr (r ) = (2μ + λ)w (r ) + 2λ .
r
Let us consider the surface stresses in the case of radial deformation. The boundary
portion Ω S is r = a, where a is the radius of the sphere. In spherical coordinates
one obtains
1 ∂ 1 ∂
∇S = eθ + eφ
a ∂θ sin θ ∂φ
where φ, θ are the angular coordinate variables and eθ , eφ are their corresponding
basis vectors. So
w(a) w(a)
ε̃ε = (eθ eθ + eφ eφ ) ≡ A.
a a
We obtain the surface stress tensor and its divergence:
w(a) 2τ
τ = τ A, τ = 2(μ S + λ S ) , ∇ S · τ = − n.
a a
The boundary conditions for the fixed boundary, the free boundary, and the bound-
ary with surface stresses, reduce to
w(a) = 0, (37)
σr (a) = 0, (38)
2τ
σr (a) = − , (39)
a
respectively. Using Eq. (36), we transform Eqs. (37)–(39) to the following transcen-
dental equations w.r.t. η:
ηa cos(ηa) − sin(ηa) = 0, (40)

(λ + 2μ)η a sin(ηa) + 4 μ [ηa cos(ηa) − sin(ηa)] = 0,
2 2
(41)
(λ + 2μ)η a sin(ηa) + 4 μ [ηa cos(ηa) − sin(ηa)]
2 2
− α (cos (ηa) ηa − sin (ηa)) = 0, (42)
where α = 4(λ S + μ S )/μa is a dimensionless parameter. We denote the solutions

of Eqs. (40)–(42) by ηk◦ , ηk , and ηk , respectively.
f
For Problem Pss , the dependencies of ηk on α are given in Fig. 2. Here the dashed
and stroke-dashed lines correspond to ηk and ηk◦ for Problems P f and P0 , respec-
f
tively. The first seven eigenfrequencies are presented in Table 1, assuming λ = μ

and α = 1.
7
ηk ηk
a a
10
3 20 6
2 5
5 15
1 4
0 10
0 10 20 30 40 α 0 10 20 30 40 α
Fig. 2 Dependencies of the normalized eigenfrequencies ηk on α for k = 1, 2, . . . , 7
Table 1 Normalized eigenfrequencies of an elastic sphere for problems P f , Pss , and P0

k 1 2 3 4 5 6 7
f
ηk /a 2.563 6.059 9.279 12.459 15.622 18.778 21.930
ηk /a 2.744 6.117 9.317 12.486 15.644 18.796 21.945
ηk◦ /a 4.493 7.725 10.904 14.066 17.221 20.371 23.519
6 Conclusion
For an elastic body with surface stresses the theorems of existence and uniqueness of
the solutions of both the static and dynamic problems in energy spaces are formulated
and proven. Some properties of the spectrum of the problems are derived. Solutions
of the problems under consideration are more smooth on the boundary surface than
solutions of corresponding problems of the classical linear elasticity. We established
lower and upper bounds for the eigenfrequencies of an elastic body with surface
stresses. These bounds cannot be improved. For the kth eigenfrequency, the lower
bound is the kth eigenfrequency of the same body with free boundary, while the upper
bound is the kth eigenfrequency of the same body with fixed boundary. The increase
in the eigenfrequencies for the elastic body with surface stresses, in comparison with
the same body with free boundary, can be interpreted as the increase in the stiffness.
Acknowledgments The second author was supported by the DFG grant No. AL 341/33-1 and by
the RFBR with the grant No. 12-01-00038.
References
1. Altenbach, H., Eremeyev, V.A., Lebedev, L.P.: On the existence of solution in the linear elas-
ticity with surface stresses. ZAMM 90(7), 535–536 (2010)
2. Altenbach, H., Eremeyev, V.A., Lebedev, L.P.: On the spectrum and stiffness of an elastic body
with surface stresses. ZAMM 91(9), 699–710 (2011)
3. Ciarlet, P.G.: Mathematical Elasticity. Vol. I: Three-Dimensional Elasticity. North-Holland,
Amsterdam (1988)
4. Ciarlet, P.G.: Mathematical Elasticity. Vol. III: Theory of Shells. North-Holland, Amsterdam
(2000)
5. Ciarlet, P.G.: The Finite Element Method for Elliptic Problems. SIAM, Philadelphia (2002)
6. Courant, R., Hilbert, D.: Methods of Mathematical Physics, vol. I. Wiley, Singapore (1989)
7. Duan, H.L., Wang, J., Karihaloo, B.L.: Theory of elasticity at the nanoscale. Adv. Appl. Mech.
42, 1–68 (2008)
8. Fichera, G.: Existence theorems in elasticity. In: S. Flügge (ed.) Handbuch der Physik, vol.
VIa/2, pp. 347–389. Springer, Berlin (1972)
9. Gurtin, M.E., Murdoch, A.I.: A continuum theory of elastic material surfaces. Arch. Ration.
Mech. Anal. 57(4), 291–323 (1975)
10. Javili, A., McBride, A., Steinmann, P.: Numerical modelling of thermomechanical solids with
mechanically energetic (generalised) Kapitza interfaces. Comput. Mater. Sci. (2012). doi:10.
1016/j.commatsci.2012.06.006
11. Javili, A., McBride, A., Steinmann, P., Reddy, B.D.: Relationships between the admissible
range of surface material parameters and stability of linearly elastic bodies. Phil. Mag. (2012).
doi:10.1080/14786435.2012.682175
12. Javili, A., Steinmann, P.: A finite element framework for continua with boundary energies. Part
I: the two-dimensional case. Comput. Methods Appl. Mech. Eng. 198, 2198–2208 (2009)
II: the three-dimensional case. Comput. Methods Appl. Mech. Eng. 199, 755–765 (2010)
14. Lebedev, L.P., Vorovich, I.I.: Functional Analysis in Mechanics. Springer, New York (2003)
15. Podstrigach, Y.S., Povstenko, Y.Z.: Introduction to Mechanics of Surface Phenomena in
Deformable Solids (in Russian). Naukova Dumka, Kiev (1985)
16. Rubin, M., Benveniste, Y.: A Cosserat shell model for interphases in elastic media. J. Mech.
Phys. Solids 52(5), 1023–1052 (2004)
17. Schiavone, P., Ru, C.Q.: Solvability of boundary value problems in a theory of plane-strain
elasticity with boundary reinforcement. Int. J. Eng. Sci. 47(11–12), 1331–1338 (2009)
18. Steigmann, D.J., Ogden, R.W.: Elastic surface-substrate interactions. Proc. Royal Soc. Lond.
A 455(1982), 437–474 (1999)
19. Wang, J., Duan, H.L., Huang, Z.P., Karihaloo, B.L.: A scaling law for properties of nano-
structured materials. Proc. Royal Soc. Lond. A 462(2069), 1355–1363 (2006)
20. Wang, J., Huang, Z., Duan, H., Yu, S., Feng, X., Wang, G., Zhang, W., Wang, T.: Surface
stress effect in mechanics of anostructured materials. Acta Mechanica Solida Sinica 24, 52–82
(2011)
On the Influence of Residual Surface
Stresses on the Properties of Structures
at the Nanoscale
Holm Altenbach, Victor A. Eremeyev and Nikita F. Morozov
Abstract We discuss the influence of residual surface stresses on the effective

(apparent) properties of materials at the nanoscale such as the stiffness of rods.
The interest to the investigation of the surface effects is recently grown with respect
to progress in nanotechnologies. The surface and interface effects play an important
role for nanofilms, nanocomposites, nanoporous materials, etc. Here we consider the
Gurtin–Murdoch model of surface elasticity. With the help of the simple problem of
uniaxial tension of a rod with residual surface stresses we analyze the behavior of
the rod under tension and present the effective stiffness.
1 Introduction
Recently, the interest to the model of surface elasticity by Gurtin and Murdoch [6]
grows fast with respect to development of nanotechnologies, see [3, 16]. The model
[6] predicts the size effect observed in the case of nanosized materials [15]. Unlike to
macro- and microsized specimen where the size effect can be explained by various
mechanisms, see the review [2], the size effect in nanomechanics can be related to
surface phenomena only. An elastic body with surface stresses can be considered as a
classical elastic body with glued elastic membrane. The stress resultant tensor acting
H. Altenbach · V. A. Eremeyev (B)

Institut für Mechanik, Fakultät für Maschinenbau, Otto-von-Guericke-Universität
Magdeburg, Universitätsplatz 2, 39106 Magdeburg, Germany
e-mail: holm.altenbach@ovgu.de
V. A. Eremeyev
South Scientific Center of RASci & South Federal University, Rostov on Don, Russia
N. F. Morozov
St. Petersburg State University, Bibliotechnaya sq. 2, 198904 St. Petersburg, Russia
e-mail: morozov@NM1016.spb.edu

in the membrane can be interpreted as surface stresses in the Gurtin–Murdoch model.

The additional surface elasticity influences on the effective properties of materials
[3, 16]. It was shown that within the linear theory of surface elasticity the presence of
surface stresses lead to the stiffening of the material, see, for example, [1, 4, 17, 18].
The residual or initial surface stresses can play an important role with respect to the
material behavior at the nanoscale. In particular, the residual surface stress change
the free vibrations of materials [5, 10, 14]. The finite deformations of elastic solids
within the framework of the model [6] including analysis of residual surface stresses
are considered in [7–9]. Let us note that the using of nonlinear elasticity methods is
necessary for the correct describing of the prestressed state of solids.
The paper is organized as follows. In Sect. 2 we recall the basic equations of
elasticity with surface stresses. In Sect. 3 we discuss the constitutive equations in
more details. Here we introduce the natural configurations for the elastic body and
residual (initial) stresses as a result of mismatch of natural configurations for the
bulk and surface materials, respectively. We formulate the constitutive equations
for surface stresses with non-natural reference configuration. Finally, we illustrate
influence of residual surface stresses on the effective tangent stiffness considering
the uniaxial tension of a rod with surface stresses.
Throughout the paper we use the direct tensor and vector notations as in [11].
2 Boundary-Value Problem for Nonlinear Solids

with Surface Stresses
Following [6] we recall the basic equations of elastic materials taking into account
surface stresses. The deformation of an elastic body is described by the mapping
x = x(X), (1)
where x and X are the position vectors in the actual configuration χ and in the
reference one κ, respectively, see Fig. 1.
The Lagrangian equilibrium equations and the boundary conditions take the
following form:
∇ · P + ρf = 0 , (n · P − ∇s · S)|Ωs = t ,
(2)
u|Ωu = u0 , n · P|Ω f = t .
Here P is the first Piola-Kirchhoff stress tensor, ∇ the Lagrangian three-dimensional

(3D) nabla operator, ∇s the surface (2D) nabla operator, S the surface stress tensor of
the first Piola-Kirchhoff type acting on the surfaces Ωs , u = x − X the displacement
vector, f and t the body force and surface loads vectors, respectively, and ρ the
density. We assume that on the part of the body surface Ωu the displacements are
given, while on Ω f the surface stresses S are absent, see Fig. 1. Equation (2)2 is
Residual Surface Stresses on the Properties of Structures at the Nanoscale 23
n
N
κ χ
ωS
ΩS N Ωf
v
Ωu V
X x
i3
0 i2
i1
Fig. 1 Deformation of a body with surface stresses
the so-called generalized Young-Laplace equation describing the surface tension in

solids. For solution of (2) we have to specify the constitutive equations for the stress
tensors T and S.
3 Constitutive Relations of Surface Elasticity
For the bulk material we use the standard constitutive relations of the nonlinear
elasticity, see [12, 13],
∂W
P= , W = W (F), (3)
∂F
where W is the strain energy density and F = ∇x the deformation gradient. In the
theory of Gurtin and Murdoch [6] the stress tensor S is similar to the membrane stress
resultants tensor and expressed with the use of the surface strain energy density U
∂U
S= , U = U (Fs ), (4)
∂Fs

where Fs = ∇s xΩ is the surface deformation gradient.
s
Hence, to specify a hyperelastic solid with surface stresses one needs two con-
stitutive equations for both the bulk and the surface behavior, that is for W and U .
Formally, in addition to (1) one needs the mapping Ωs → ωs . Since we usually con-
sider joint deformation of the volume V and the surface Ωs in κ to the corresponding
volume v and the surface ωs in χ , it is enough to use the same mapping (1) for the
deformation of the surface and the volume.
In the nonlinear elasticity one usually chooses a natural reference configuration.
This means that W and P vanish without deformation or, in other words, W and P
n
N
κ χ
χ
κ
κ
κs
κ κs
κ
i3
0 i2
i1
κ
Fig. 2 Deformation of body with surface stresses: reference, actual, and two different natural
configurations for surface and bulk material
possess the properties W (I) = 0, P(I) = 0, where I is the 3D unit tensor. In the case
of surface elasticity one has to choose two natural reference configurations for W
and U which do not coincide with each other, in general. This case is schematically
shown in Fig. 2. Here κs◦ and κ ◦ are the natural configurations taken different for the
surface and the bulk material behavior. For example, if κs◦ = κ and κ ◦ = κs◦ there
exist residual (initial) surface energy and surface stresses that is
U (A) = U0 = 0, S(A) = S0 = 0,
where A ≡ I − N ⊗ N is the surface unit tensor. In other words, here we assume

that the reference configuration κ for the bulk material is natural one while for the
attached on Ωs membranes we assume the non-natural reference configuration κ
with natural one κs◦ .
Let Fs◦ and Fs∗ be the surface deformation gradients related to mappings κs◦ → κ
and κs◦ → χ , respectively. Then there is the multiplicative decomposition
Fs∗ = Fs◦ · Fs .
Since Fs∗ corresponds to the mapping from the stress-free configuration to the actual
one it can be used in the constitutive equations for U and S. Keeping the same
notation we re-write the constitutive equation for U as follows
U = U (Fs◦ · Fs ),
where U satisfies the condition U (A) = 0. Tensor Fs◦ can be considered as the
given parametric tensor in the constitutive equations. The surface stress tensor S is
given now by the relation
∂U
S = Fs◦ T · ∗ . (5)
∂Fs
The initial surface energy and surface stresses are given by

∂U
U0 = U (Fs◦ ), S0 = Fs◦ T · .
∂Fs∗ Fs◦
Using the material frame-indifference principle we write the strain energies as a

functions of the right Cauchy–Green strain tensor and its surface analogues
W = W (C), U = U (Fs◦ · Cs · Fs◦ T ), (6)
where C = F · FT and Cs = Fs · FsT . In the case the isotropic material behavior W

and U are expressed via the principal invariants
W = W (I1 , I2 , I3 ), U = U (J1 , J2 ), (7)
where
1 2
I1 = tr C, I2 = tr C − tr C2 , I3 = det C,
2
2
J1 = tr Fs◦ · Cs · Fs◦ T , J2 = tr Fs◦ · Cs · Fs◦ T .
The corresponding surface stress tensor S takes the form

∂U ∂U ◦ T ◦ ∂U ◦ T ◦ ◦T ◦
S=2 · Fs = 2 F · Fs + 2 F · Fs · Cs · Fs · Fs · Fs . (8)
∂Cs ∂ J1 s ∂ J2 s
As an example of the surface strain energy we consider the quadratic function
1 1
U = λs (J1 − 2)2 + μs (J2 − 2J1 + 2), (9)
8 4
where λs and μs are the surface elastic moduli, which are also named the surface
Lamé moduli. For (9) the tensor S has the form

1
S= λs (J1 − 2) − μs Fs◦ T · Fs◦ + μs Fs◦ T · Fs◦ · Cs · Fs◦ T · Fs◦ · Fs . (10)
2
In the case of infinitesimal deformations without initial deformations we have
Cs ≈ A + 2εε , Fs◦ = A,
and Eqs. (9) and (10) reduce to the relations of the linear surface elasticity, see [1],
1
U = λs tr 2ε + μs tr ε 2 , S = λs Atr ε + 2μs ε ,
2
where ε is the linear surface strain tensor.
Let us consider the uniform surface tension Fs◦ = λ◦ A as an example of the initial
surface deformation gradient. Here we have J1 = 2λ2◦ , J2 = 2λ4◦ , and
1
U0 = (λs + μs )(λ2◦ − 1)2 , S0 = (λs + μs )λ2◦ (λ2◦ − 1)A.
2
Further we consider the influence of residual (initial) stresses on effective (apparent)
stiffness of solids.
4 Uniaxial Tension
To illustrate the influence of surface stresses including residual ones let us consider
the uniaxial tension of a circular cylinder made of incompressible material, see Fig. 3.
The mapping (1) is now given by
x1 = λ−1/2 X 1 , x2 = λ−1/2 X 2 , x3 = λX 3 , (11)
where xk , X k , k = 1, 2, 3, are the Cartesian coordinates in the actual and reference

configurations, respectively, and λ is the stretch parameter.
The corresponding deformation gradient F and the right Cauchy–Green strain
tensor C are given by formulas
F = λ−1/2 (i1 ⊗ i1 + i2 ⊗ i2 ) + λi3 ⊗ i3 , C = λ−1 (i1 ⊗ i1 + i2 ⊗ i2 ) + λ2 i3 ⊗ i3 ,
or with base vectors of cylindrical coordinates as follows
F = λ−1/2 (e R ⊗ e R + eΦ ⊗ eΦ ) + λe Z ⊗ e Z ,
C = λ−1 (e R ⊗ e R + eΦ ⊗ eΦ ) + λ2 e Z ⊗ e Z .
The principal invariants of C are
1 2
I1 ≡ tr C = 2λ−1 + λ2 , I2 ≡ tr C − tr C2 = 2λ + λ−2 , I3 ≡ det C = 1.
2
Fig. 3 Uniaxial tension of

a circular rod of radius a F
subjected by the force F
Ωs
i2
Ωf
0
i1
i3
a
F
In a similar way we calculate the surface deformation gradient Fs and the surface
left Cauchy–Green strain tensor Cs
Fs = λ−1/2 eΦ ⊗ eΦ + λe Z ⊗ e Z , Cs = λ−1 eΦ ⊗ eΦ + λ2 e Z ⊗ e Z .
We assume Fs◦ in the form
Fs◦ = λ1 eΦ ⊗ eΦ + λ2 e Z ⊗ e Z ,
where λ1 and λ2 are positive numbers describing initial axisymmetric stretching of

the surface Ωs . Thus Fs◦ · Cs · Fs◦ T is given by
Fs◦ · Cs · Fs◦ T = λ21 λ−1 eΦ ⊗ eΦ + λ22 λ2 e Z ⊗ e Z ,
and its invariants are
J1 = λ21 λ−1 + λ22 λ2 , J2 = λ41 λ−2 + λ42 λ4 .
For the bulk material we use the neo-Hookean model
W = μ(I1 − 3), P = 2μF − pF−T , (12)
where μ is the elastic modulus playing a role of the shear modulus in the case of
infinitesimal deformations, p is the pressure in incompressible materials.
Obviously, in the case of uniaxial tension (2)1 is fulfilled, the boundary condition
(2)2 reduces to
1
PR R + SΦΦ = 0. (13)
a a
Using (12) from (13) we obtain p
1 −1/2
p = 2μλ−1 + λ SΦΦ .
a
This gives us the axial nominal stress
1 −3/2
PZ Z = 2μ(λ − λ−2 ) − λ SΦΦ . (14)
a
From (10) it follows
S = SΦΦ eΦ ⊗ eΦ + S Z Z e Z ⊗ e Z ,
where

1
SΦΦ = λs (J1 − 2) − μs λ21 λ−1/2 + μs λ41 λ−3/2 , (15)
2

1
SZ Z = λs (J1 − 2) − μs λ22 λ + μs λ42 λ3 . (16)
2
Considering the integral equilibrium condition

Fez = ez · P dS + ez · S ds
Ωf ∂Ω f
we obtain the formula

F = πa 2 PZ Z + 2πaS Z Z . (17)
Substituting into (17) Eqs. (14) and (16) we obtain the dependence of the axial force
F on the stretch parameter λ taking into account initial stretching λ1 and λ2
1 −2 1 −3/2
F = 2μ(λ − λ ) + −λ SΦΦ + 2S Z Z . (18)
πa 2 a
Let us first consider the influence of surface stresses without residual ones. In
Fig. 4 we present dependence F on λ.
Here
F
F=
2π μa 2
is the nominal axial stress. The dotted curve corresponds to the function
F 0 (λ) = λ − λ−2 ,
3
F
Fig. 4 F versus λ without residual surface stresses
which describe the uniaxial tension of a rod within the framework of neo-Hookean
model. For simplicity let us assume that λs = μs . We introduce the dimensionless
parameter
μs
α= ,
2μa
so F = F(λ, α) and F(λ, 0) = F 0 . Curves 1, 2, 3 present the dependence F(λ) for

the values α = 0.05; 0.1; 0.2, respectively. It is seen that
F(λ, α1 ) > F(λ, α2 ) for α1 > α2 ≥ 0, λ > 1,
and
F(λ, α1 ) < F(λ, α2 ) for α1 > α2 ≥ 0, λ < 1.
λ 1.2
F
λ 1.1
λ 1
λ 0.9
λ 0.8
λ 1.2
F
λ 1.1
λ 1
λ 0.9
λ 0.8
Fig. 5 F versus λ with uniform residual surface stresses
Calculating the tangent stiffness by the formula
∂F
E(λ, α) =
∂λ
we conclude that
E(λ, α1 ) > E(λ, α2 ) for α1 > α2 ≥ 0, λ = 1.
This means that the rod with surface stresses becomes stiffer then the rod without
the latter.
Let us consider the influence of uniform residual stresses λ1 = λ2 = λ◦ . Here we
have more complicated behavior, see Fig. 5. As in Fig. 4 the dotted curve corresponds
function F 0 (λ), others curves correspond to various values of λ◦ . Here α = 0.1 is
assumed. One can see that initial stretching λ◦ > 1 leads to the increase of F while
initial compression λ◦ < 1 leads to the decrease of F. The detailed analysis near the
λ2 1.2
F
λ2 1.1
λ2 1
λ2 0.9
λ2 0.8
λ
F
Fig. 6 F versus λ with uniaxial residual surface stresses
point λ = 1 shows that the values of λ, when F = 0, are also shifted with respect
to λ◦ .
Finally, we consider the uniaxial initial stretching with λ1 = 1, λ2 = 1. Such state
corresponds to the uniaxial stress tensor S0 = S Z0 Z ez ⊗ ez . Unlike to the uniform
initial stretching this case relates to self-equilibrate initial stresses in the case of a
cylindrical body. This means that S0 and P = 0 satisfy the equilibrium conditions
(2)1,2 . The corresponding dependencies are shown in Fig. 6. Here α = 0.1 again is
assumed.
Both uniform and uniaxial residual (initial) stretching change the effective tangent
stiffness of the rod. In particular, initial compression leads to the decrease of the
tangent stiffness in comparison with the tangent stiffness of the rod with surface
stresses but without residual ones.
5 Conclusion
Here we discussed the influence of residual surface stresses on the effective (apparent)
stiffness of nanodimensional specimen. It was shown that the presence of surface
stresses leads to the increase of stiffness of nanosized specimen in comparison with
bulk material while influence of residual stresses may result in the decrease or the
increase of the effective stiffness of material.
Acknowledgments The second author was supported by the DFG grant No. AL 341/33-1 and by
the RFBR with the grant No. 12-01-00038.
References
with surface stresses. ZAMM 91(9), 699–710 (2011)
2. Bažant, Z.P.: Size effect. Int. J. Solids Struct. 37(1–2), 69–80 (2000)
3. Duan, H.L., Wang, J., Karihaloo, B.L.: Theory of elasticity at the nanoscale. In: Advances in
Applied Mechanics, vol. 42, pp. 1–68. Elsevier, San Diego (2008)
4. Guo, J.G., Zhao, Y.P.: The size-dependent elastic properties of nanofilms with surface effects.
J. Appl. Phys. 98(7), 074306–074311 (2005)
5. Gurtin, M.E., Markenscoff, X., Thurston, R.N.: Effect of surface stress on natural frequency
of thin crystals. Appl. Phys. Lett. 29(9), 529–530 (1976)
Mech. Anal. 57(4), 291–323 (1975)
7. Huang, Z., Sun, L.: Size-dependent effective properties of a heterogeneous material with inter-
face energy effect: from finite deformation theory to infinitesimal strain analysis. Acta Mechan-
ica 190, 151–163 (2007)
8. Huang, Z., Wang, J.: A theory of hyperelasticity of multi-phase media with surface/interface
energy effect. Acta Mechanica 182, 195–210 (2006)
9. Huang, Z., Wang, J.: Micromechanics of nanocomposites with interface energy effect. In:
Handbook on Micromechanics and Nanomechanics, p. 48. Pan Stanford Publishing, Singapore
(2012) (in print)
10. Lagowski, J., Gatos, H.C., Sproles, E.S.: Surface stress and normal mode of vibration of thin
crystals: GaAs. Appl. Phys. Lett. 26(9), 493–495 (1975)
11. Lebedev, L.P., Cloud, M.J., Eremeyev, V.A.: Tensor Analysis with Applications in Mechanics.
World Scientific, New Jersey (2010)
12. Lurie, A.I.: Nonlinear Theory of Elasticity. North-Holland, Amsterdam (1990)
13. Ogden, R.W.: Non-linear Elastic Deformations. Ellis Horwood, Chichester (1984)
14. Wang, G.F., Feng, X.Q.: Effects of surface elasticity and residual surface tension on the natural
frequency of microbeams. Appl. Phys. Lett. 90(23), 231904 (2007)
15. Wang, J., Duan, H.L., Huang, Z.P., Karihaloo, B.L.: A scaling law for properties of nano-
structured materials. Proc. Royal Soc. Lond. A 462(2069), 1355–1363 (2006)
16. Wang, J., Huang, Z., Duan, H., Yu, S., Feng, X., Wang, G., Zhang, W., Wang, T.: Surface stress
effect in mechanics of nanostructured materials. Acta Mechanica Solida Sinica 24, 52–82
(2011)
17. Wang, Z.Q., Zhao, Y.P., Huang, Z.P.: The effects of surface tension on the elastic properties of
nano structures. Int. J. Eng. Sci. 48(2), 140–150 (2010)
18. Zhu, H.X., Wang, J.X., Karihaloo, B.L.: Effects of surface and initial stresses on the bending
stiffness of trilayer plates and nanofilms. J. Mech. Mater. Struct. 4(3), 589–604 (2009)
On the Isotropic Elastic Properties of Graphene
Crystal Lattice
Igor E. Berinskii and Feodor M. Borodich
Abstract Graphene is a monolayer of carbon atoms packed into a two-dimensional

honeycomb lattice. This allotrope can be considered as mother of all graphitic
forms of carbon. The elastic in-plane properties of graphene are studied. Nowadays
graphene often is simulated as a two-dimensional elastic continuum. It is shown
in this work that if this continuum has the same symmetric properties as graphene
crystal, then the continuum is isotropic while the small deformations are considered.
A simple and mathematically rigorous proof of this statement is given. The proof
is based on the orthogonal transformation of the coordinates of the continual stress
and strain tensors and comparison of the elastic tensor components before and after
transformation.
1 Introduction
As it was noted in [7], graphene is a basic building block for graphitic materials of
all other dimensionalities: it can be wrapped up into 0D buckyballs, rolled up into
1D nanotubes or stacked into 3D graphite. Thus, studies of mechanical properties of
graphene are of enormous practical interest.
Mechanical properties of graphite were studied for many years [3, 5]. Although
the term “graphene” as a single carbon layer of the graphitic structure first appeared
in 1987 [14] and became well-established since 1994 [4], the first isolated individual
graphene planes were obtained by using adhesive tape only in 2004 [11]. In fact,
Geim and Novoselov and their coworkers showed that graphene can exist in the free
I. E. Berinskii (B) · F. M. Borodich

School of Engineering, Cardiff University, Cardiff CF24 3AA, UK
e-mail: iberinsk@gmail.com
F. M. Borodich
e-mail: BorodichFM@cardiff.ac.uk

34 I. E. Berinskii and F. M. Borodich
Fig. 1 Graphene crystal

lattice. Circles simulate the
carbon atoms connected by
covalent bonds
state and it is stable. Their further studies of graphene were devoted mainly to its
remarkable electronic properties.
As graphene has the honeycomb structure (see Fig. 1) it was natural to investigate
its mechanical properties using the engineering approaches that were applied to such
structures. Hexagons were used in engineering applications to make light and stiff
constructions such as aircraft panels or skis. An analysis of in-plane properties of
hexagonal structures was presented in [6]. In particular, the effective in-plane elastic
moduli and Poisson ratio were calculated for honeycomb structures loaded in two
orthogonal directions and then some plastic properties of such constructions were
investigated. These structures were also studied in [8] where linear and nonlinear
elastic and plastic properties of honeycombs were considered and compared with
experiments. In [10] previous models were developed and experimentally observed
anisotropy of honeycomb systems was explained using the structural mechanics
approaches.
The experimental results obtained for large scale engineering honeycomb
structures [8] demonstrate the anisotropic behaviour. At the same time, the discrete
models of graphene (see e.g. [9]) demonstrate more symmetrical properties than the
structural models. The contradiction may be explained by the fact that the connec-
tions between joints of the real structural members can not be simulated the same
way as the interatomic bonds in crystal lattice. Hence, it is important to show, that
the 2D continuum model that corresponds to the graphene lattice should be isotropic.
It is known [13] an isotropic elastic material is an elastic material whose symmetry
group does contain the proper orthogonal group for a reference configuration and,
therefore the mechanical response of the material exhibits no preferred direction, the
property that characterizes material isotropy. The symmetry properties of the stiffness
tensor of an elastic continuum are usually studied either by direct consideration of
particular cases of symmetry transformations (see, e.g. paragraph 14 in [16]) or
by using the general group approach [13]. However, the latter approach is rather
sophisticated and it is not a trivial task to apply this approach to particular cases of
symmetry. Of course, one can write the general expression for re-calculating of the
stiffness components in new coordinates (see, e.g. (19) in paragraph 3.5 [12]) and
On the Isotropic Elastic Properties 35
then check if they are the same. However, we believe that it is simpler to follow
the general approach for determination of elastic properties of effective continuum
models for 3D crystals described by Slawinski [15]. Thus, we consider further the
transformation matrices satisfying the symmetry properties of the graphene crystal
lattice and this approach is used to describe elastic properties of effective 2D model
of graphene. Note, that only small deformations, e.g. the linear elastic continuum
is considered in this work. At large deformations a condition of isotropy may be
violated.
2 The Classical Hooke Law After the Orthogonal

Transformation of Strain and Stress Coordinates
If the 2D elasticity of a 0x1 x2 plane is considered then the displacements in orthogonal

(x3 ) direction is u 3 = 0 and the corresponding derivatives ∂u/∂ x3 = 0. Hence, one
has

∂u 1 1 ∂u 1 ∂u 2
ε11 = , ε12 = + ,
∂ x1 2 ∂ x2 ∂ x1

∂u 2 1 ∂u 1 ∂u 3
ε22 = , ε13 = + = 0, (1)
∂ x2 2 ∂ x3 ∂ x1

1 ∂u 2 ∂u 3 ∂u 3
ε23 = + = 0, ε33 = =0
2 ∂ x3 ∂ x2 ∂ x3
The classical Hooke’s law for 2D case may be presented in tensor form as

2
2
σi j = Ci jkl εkl , i, j = 1, 2 (2)
k=1 l=1
One can represent this equation in matrix form using elasticity matrix C. In 3D case
this matrix generally has a dimension 6×6 but in 2D case we can reduce it to 3×3.
Let us construct it taking a symmetry of the elasticity tensor into account. Consider
the pairs of indexes (i, j) and (k, l) such that i ≤ j and k ≤ l for i, j, k, l = 1, 2.
Now we replace each pair by single index such that
(1, 1) → 1, (2, 2) → 2, (1, 2) → 3. (3)
Hence every Ci jkl of the elasticity tensor for i, j, k, l = 1, 2 corresponds to the

element of the elasticity matrix Cnm for n, m = 1, 2, 3. Finally, we can represent
Eq. (2) as ⎛ ⎞ ⎛ ⎞⎛ ⎞
σ11 C11 C12 C13 ε11
⎝ σ22 ⎠ = ⎝ C12 C22 C23 ⎠ ⎝ ε22 ⎠ (4)
σ12 C13 C23 C33 2ε12
Graphene lattice has third order axial symmetry. This means that an in-plane
rotation by the angle Θ = 2π/3 transfers the lattice into itself. This transformation
is determined by transformation matrix
√
cos Θ sin Θ −1/2
√ 3/2
A= = (5)
− sin Θ cos Θ − 3/2 −1/2
One can see that detA = 1 and A−1 = A T , therefore, AA T = E, where E is a unity
matrix.
After rotation the stress matrix σ

σ11 σ12
σ = (6)
σ12 σ22
transforms into σ̃ according to

σ̃ = Aσ A T . (7)
Here and henceforth, we write the components in the new transformed basis with the
tilde sign, e.g. σ̃ .
Readers interested in the underlying aspects of linear algebra might refer to
[1, 2]. The expression (7) can be represented using the components of the stress
vector σ (the single-column matrix of stress tensor) as
σ̃ = Ãσ, σ = [σ11 σ22 σ12 ]T . (8)
To find the transformed elements σ̃i j , one needs to determine the elements of the
rotation matrix Ã. Let us consider the unite elements of the standard basis of the
space of the symmetric 2×2 matrices

10 00 01 00
σ1 = , σ2 = , σ3 = , σ4 = (9)
00 01 00 10
as probe stresses.
It follows from (8) that the transformation applied to σ1 gives

A11 A12 10 A11 A21 A11 A11 A11 A21
σ̃1 = = (10)
A21 A22 00 A12 A22 A21 A11 A21 A21
The elements of the stress basis matrix σ1 could be written using Kroneker’s symbols
as
(σ1 )kl = δk1 δl1 . (11)
Hence, one can represent (10) as


2
2
(σ̃1 )kl = Aki σi j A jl T = Ak1 Al1 (12)
i=1 j=1
To consider the matrix form of stress-strain equations we have to construct

single-column matrix of new stress components using (k, l) = (1, 1), (2, 2), (1, 2)
in following form
σ˜1 = (A11 A11 A21 A21 A11 A21 )T (13)
Let us compare the last equation with (8) taking into account that according to (9)

T
σ1 = 1 0 0 (14)
This means that the expression (13) is the first column of Ã. Similarly we can find
the second column of this matrix using the transformation of σ2 . To find the third
column, we have to consider

01
σ3 + σ4 = (15)
10
that gives us
(σ3 + σ4 )kl = δk1 δl2 + δk2 δl1 (16)
Hence according to (6)

A11 A12 01 A11 A21
σ˜3 + σ˜4 =
A21 A22 10 A12 A22

2 A11 A11 A12 A21 + A11 A22
= (17)
A12 A21 + A11 A22 2 A21 A22
or if we use a single-column matrix representation
σ̃3,4 = (2 A11 A11 2 A21 A22 A12 A21 + A11 A22 )T (18)
that gives us a third column of matrix Ã. Finally, we can combine tree columns to
obtain ⎛ ⎞
A11 A11 A12 A12 2 A11 A11
Ã = ⎝ A21 A21 A22 A22 2 A21 A22 ⎠ (19)
A11 A21 A12 A22 A11 A22 + A12 A21
Now let us find the transformed components of strain-tensor. These components

transform according to
ε̃ = AεA T (20)
where A stands for (4). Strain-tensor may be presented as a square symmetrical

matrix
ε11 ε12
ε= (21)
ε12 ε12
Let us rewrite the strain-tensor components as a single-column matrix to consider the

matrix form of stress-strain equations. From (4) it follows that this matrix contains
factor of 2 for the third component, namely
⎛ ⎞
ε11
ε = ⎝ ε22 ⎠ (22)
2ε12
Hence, the corresponding transformation matrix differs from (19). Let us account
the factors of 2 by
ε̃ = FÃF−1 ε, (23)
where Ã is a matrix (19) and ⎛ ⎞

100
F = ⎝0 1 0⎠ (24)
002
Hence, the transformation matrix for strain-tensor components given by matrix (22)
can be restated as
⎛ ⎞
A11 A11 A12 A12 A11 A12
MA = FÃF−1 = ⎝ A21 A21 A22 A22 A21 A22 ⎠ (25)
2 A11 A21 2 A12 A22 A11 A22 + A21 A12
The last expression gives us an opportunity to represent (23) as
ε̃ = MA ε (26)
In case of A given by (5) the MA has a following view

⎛ √ ⎞
1 3 3
⎜ 4 − ⎟
⎜ 4 √ 4 ⎟
⎜ 3 1 3 ⎟
MA = ⎜ ⎟ (27)
⎜ √4 √
4 4 ⎟
⎝ ⎠
3 3 1
− −
2 2 2
Since now we have obtained σ̃ and ε̃, we can formulate stress-strain equation in
transformed coordinates, namely
σ̃ = C̃ε̃ (28)
3 Isotropy of an Effective Continuum 2D Model of Graphene
Let us use the following definition [15]:

Definition 3.1 the elastic properties of a continuum are invariant under an orthogo-
nal transformation of coordinates if the transformed elasticity matrix is identical to
the original elasticity matrix, i.e. C̃ = C.
Theorem 3.1 Elastic properties of a continuum are invariant under an orthogonal
transformation, given by matrix A, if and only if
C = MA
T
CMA . (29)
The proof of this theorem is given in [15] and for the sake of completeness, it is
repeated here.
First, let us consider the following lemma.
Lemma 3.1 Let Ã be given by matrix (17) and MA be given by matrix (25). It follows
that Ã−1 = MA T .
Proof Let us calculate Ã−1 Since A is an orthogonal matrix, namely, A T = A−1 ,
we can rewrite Eq. (6) as
σ = A T σ̃ A (30)
Hence, in a manner analogous to that used to obtain expression (9) we can rewrite
expression (30) in the desired notation as
σ = A˜T σ̃ , (31)
where A˜T is constructed as matrix (17), but with entries AiTj = A ji of A T used
in place of the entries Ai j of A. Note, that the order of operations matters; namely
A˜T = Ã T . Comparing expression (9) with (31), we see that A˜T = Ã−1 . Hence, we
can write the inverse of matrix Ã explicitly, namely, as
⎛ ⎞
A11 A11 A21 A21 2 A11 A21
Ã−1 = ⎝ A12 A12 A22 A22 2 A12 A22 ⎠ (32)
A11 A12 A21 A22 A11 A22 + A21 A12
Comparing the entries of matrices (32) and (25), we notice that the former one is
equal to the transpose of the latter, as required.
Now let us prove the Theorem 1.
Proof Consider stress-strain equations in transformed coordinate system, namely
(28). Substituting expressions (8) and (26)–(28) we obtain
Ãσ̃ = C̃MA ε̃ (33)
Multiplying both sides by Ã−1 , we get
σ̃ = Ã−1 C̃MA ε̃ (34)
According to Lemma 1 proved above, Ã−1 = MA T . Hence, we can write
σ̃ = MA T C̃MA ε̃ (35)
Examining Eqs. (28) and (35), we conclude that they both hold for any ε, if and only
if
C = MA T C̃MA , (36)
which is relation between C and C̃ under transformation matrix A. In view of defin-

ition of isotropy invariance with respect to A means that
C = MA T CMA (37)
which is expression (29) as required.

For materials having third-order symmetry such as graphene and boron nitride
condition (29) must be satisfied for the matrix A given by (5). Matrix equation
(29) leads us to a system of linear equations
⎧ √ √
⎪
⎪ 15 3 3 9 3 3 3
⎪
⎪ − C11 + C12 + C13 + C22 + C23 + C33 = 0
⎪
⎪ 16 8 √4 16 √ 4 4
⎪
⎪
⎪
⎪ 3 3 3 3 3 3
⎪
⎪ C11 − C12 + C13 + C22 − C23 − C33 = 0
⎪ 16
⎪ 8 4√ 16 4√ 4
⎪
⎪
⎪ 9 C + 3 C − 3 3 C − 15 C − 3 C + 3 C = 0
⎨ 11 12 13 22 23 33
16√ 8 √ 4 16 √ 4 √ 4 √ (38)
⎪
⎪ 3 3 3 3 3 3 3
⎪
⎪ − C11 − C12 − C13 + C22 − C23 − C33 = 0
⎪
⎪ √
16 8√ 2√ 16 √4 4
⎪
⎪
⎪
⎪− 3 3 C + 3 C + 3 C − 3 C + 3 C = 0
⎪
⎪
⎪
⎪ 16
11
8 √
12
16
22
2 √
23
4
33
⎪
⎪
⎪
⎩ 3 C11 − 3 C12 + 3 C13 + 3 C22 − 3 C23 − 3 C33 = 0
16 8 4 16 4 4
Let us note there are only five independent equations to determine six unknown values
because the second and the sixth equations are the same. Therefore, the system can
be solved up to an elastic constant of the material. Let us present result of the solution
of (29) as
C11 = C22 , C12 = C22 − 2C33 , C13 = 0, C23 = 0 (39)
Consequently, taking the symmetry of elasticity tensor into account, we have just
four non-zero coefficients connected by following relations
C1111 = C2222 , C1212 = (C1111 − C1122 )/2 (40)
Hence, only two of elastic modules are independent. This shows that graphene and
other 2D materials with third order symmetry must have the isotropic elastic prop-
erties.
4 Conclusions
The elastic in-plane properties of two-dimensional graphene layer have been studied.
It has been shown that if a graphene is modelled as 2D linear elastic continuum having
the same symmetric properties as a graphene crystal, then the continuum is isotropic.
A simple and mathematically rigorous proof of this statement is given. The results
are of importance mainly in application to the graphene samples in linear tension.
In case of nonlinear deformation the isotropy of continuum will be violated and this
effect is not considered in this work. Another case that is not investigated in this
work is buckling of the structure as a result of compression of the lattice. In this case
some elements will have the out-of-plane deformations while we have considered
just in-plane properties of graphenes. The out-of-plane deformations of honeycombs
have been studied by several authors (see e.g. [17] and references therein), however
these results are out the scope of this paper.
Acknowledgments The authors acknowledge the financial support of MINILUBES (FP7 Marie
Curie ITN network 216011-2) by the European Commission and Russian Foundation for Basic
Research (grant 12-01-00521-a).
References
1. Anton, H.: Elementary Linear Algebra. Wiley, New York (1973)

2. Ayres, F.: Matrices Schaum’s Outlines. McGraw-Hill, New York (1962)
3. Blakslee, O.: Elastic constants of compression-annealed pyrolytic graphite. J. Appl. Phys.
41(8), 3373–3383 (1970)
4. Boehm, H.P., Setton, R., Stumpp, E.: Nomenclature and terminology of graphite intercalation
compounds. Pure Appl. Chem. 66(9), 1893–1901 (1994)
5. Bosak, A., Krisch, M., Mohr, M., Maultzsch, J., Thomsen, C.: Elasticity of single-crystalline
graphite: inelastic x-ray scattering study. Phys. Rev. B 75(15), 153408 (2007)
6. El-Sayed, F.A., Jones, R., Burgess, I.: A theoretical approach to the deformation of honeycomb
based composite materials. Composites 10(4), 209–214 (1979)
7. Geim, A.K., Novoselov, K.S.: The rise of graphene. Nat. Mater. 6(3), 183–191 (2007)
8. Gibson, L.J., Ashby, M.F., Schajer, G.S., Robertson, C.I.: The mechanics of two-ddimensional
cellular materials. In: Proc. R. Soc. Lond. A Math. Phys. Sci. 382(1782), 25–42 (1982)
9. Gillis, P.: Calculating the elastic constants of graphite. Carbon 22(4–5), 387–391 (1984)
10. Masters, I., Evans, K.: Models for the elastic deformation of honeycombs. Compos. Struct.
35(4), 403–422 (1996)
11. Novoselov, K.S., Geim, A.K., Morozov, S.V., Jiang, D., Zhang, Y., Dubonos, S.V., Grigorieva,
I.V., Firsov, A.A.: Electric field effect in atomically thin carbon films. Science 306(5696),
666–669 (2004)
12. Nowacki, W.: Teoria Sprezystosci. Panstwowe Wydawnictwo Naukowe, Warszawa (1970)
13. Ogden, R.: Non-linear Elastic Deformations. Dover Publications, Mineola NY (1984)
14. Mouras, S., Hamm, A., Djurado, D., Cousseins, J.-C.: Synthesis of first stage graphite interca-
lation compounds with fluorides. Revue de Chimie Minerale 24(5), 572–582 (1987)
15. Slawinski, M.: Waves and Rays in Elastic Continua. World Scientific Publishing Co Pte Ltd,
Singapore (2010)
16. Sneddon, I., Berry, D.: Classical Theory of Elasticity. Handbuch der Physik. Springer, Berlin
(1958)
17. Zhang, J., Ashby, M.: The out-of-plane properties of honeycombs. Int. J. Mech. Sci. 34(6),
475–489 (1992)
A Comparison of Atomistic and Surface
Enhanced Continuum Approaches at Finite
Temperature
Denis Davydov, Ali Javili, Paul Steinmann and Andrew McBride
Abstract The surface of a continuum body generally exhibits properties that differ
from those of the bulk. Surface effects can play a significant role for nanomaterials,
in particular, due to their large value of surface-to-volume ratio. The effect of solid
surfaces at the nanoscale is generally investigated using either atomistic or enhanced
continuum models based on surface elasticity theory. Hereby the surface is equipped
with its own constitutive structure. Atomistic simulations provide detailed informa-
tion on the response of the material. Discrete and continuum systems are linked
using averaging procedures which allow continuum quantities such as stress to be
obtained from atomistic calculations. The objective of this contribution is to compare
the numerical approximations of the surface elasticity theory to a molecular dynamics
based atomistic model at finite temperature. The bulk thermo-elastic parameters for
the continuum’s constitutive model are obtained from the atomistic simulation. The
continuum model takes as its basis the fully nonlinear thermo-elasticity theory and is
implemented using the finite element method. A representative numerical simulation
of face-centered cubic copper confirms the ability of a surface enhanced continuum
formulation to reproduce the behaviour exhibited by the atomistic model, but at a far
reduced computational cost.
D. Davydov (B) · A. Javili · P. Steinmann

LTM,University of Erlangen–Nuremberg, Egerlandstr. 5, 91058 Erlangen, Germany
e-mail: denis.davydov@ltm.uni-erlangen.de
A. Javili
e-mail: ali.javili@ltm.uni-erlangen.de
P. Steinmann
e-mail: paul.steinmann@ltm.uni-erlangen.de
A. McBride
CERECAM, University of Cape Town, Rondebosch 7701, South Africa
e-mail: andrew.mcbride@uct.ac.za

44 D. Davydov et al.
1 Introduction
Continuum formulations are widely used for problems at the macro- and meso-scopic
length scales. Progress in nano-technology introduces new challenges of constructing
and modelling structures on the length-scale of angstroms and nano-meters. Such
problems can be modelled using discrete (molecular dynamics, MD) approaches,
which provide a very detailed description of material. Such formulations, however,
are computationally very demanding. Continuum formulations are computationally
far cheaper, but can be applied only if they are enhanced with a surface energy.
Surface effects can be modelled using various continuum approaches. Phenom-
enological models that endow the surface with its own energy date back to the pio-
neering work of Gibbs [6]. More recently, Gurtin and Murdoch [8] described surface
effects using tensorial surface stresses. Daher and Maugin [2] invoked the method
of virtual power to endow the surface with its own thermodynamic constituents. The
continuum approach used here takes as its point of departure the surface elasticity
formulation of Gurtin and Murdoch (see [12] for the extension to thermoelasticity). A
method for the determination of surface elastic properties from atomistic simulations
is developed in [18].
When using a continuum theory the question of its validity inevitably arises. A
comparison with the more detailed discrete models is not straightforward as contin-
uum fields such as the displacement, deformation gradient and stress are not readily
available from MD simulations. In order to bridge the gap and link the two fun-
damentally different approaches, space-probability averaging can be applied. This
framework was originally proposed in [10, 16]. Discrete quantities are essentially
connected with their continuum field counterparts via averaging procedures in the
current (Eulerian) configuration.
As an alternative to Eulerian averaging, Zimmerman et al. [19] proposed to average
in the reference (Lagrangian) configuration. However, their formulation is restricted
to a zero temperature reference configuration, whereby the introduced fields also
lack the probability averaging.
In this contribution we further develop the Lagrangian averaging approach. In
Sect. 2 we formulate a material averaging framework based on statistical mechanics.
From the onset all the continuum fields are introduced as space-probability averaged
atomistic quantities. We obtain an expression for the atomistic reference positions
at non-zero temperature under the condition of zero continuum displacements. Bal-
ance laws of mass and linear momentum are derived. A thermo-elastic formulation
for a continuum enhanced with a surface energy is briefly reviewed in Sect. 3. The
discrete and continuum approaches are compared in Sect. 4 using two numerical
examples. We outline the methodology to obtain thermo-elastic bulk parameters of
a solid from molecular-dynamics simulations. Using these parameters we compare
both averaging (local quantities) and continuum approaches to the results obtained
using conventional expressions for macroscopic stresses in MD (the Virial pressure).
A non-homogeneous example with surface energy—a plate with a hole is studied.
We confirm the validity of the enhanced continuum formulation against the discrete
A Comparison of Atomistic and Surface Enhanced Continuum 45
(MD) model. The discussion and conclusions are presented in Sect. 5. The main nov-
elty of this contribution is a detailed comparison of atomistic simulation at non-zero
temperature to a continuum enhanced formulation with surface energy not only in
the average (integral) sense, but also locally, using the atomistic expressions for the
continuum fields (e.g. displacement, deformation gradient, Cauchy stress). Previ-
ously, that comparison has been done either for homogeneous solutions [19] or by
applying the continuum displacements to a discrete model and comparing the stress
state [1].
Notation and Definitions
Direct notation is adopted throughout. Occasional use is made of index notation,

the summation convention for repeated indices being implied. The N -dimensional
Euclidean space is denoted E N . The scalar product of two vectors a and b is denoted
a · b = [a]i [b]i . The scalar product of two second-order tensors A and B is denoted
A : B = [A]i j [B]i j . The composition of two second-order tensors A and B, denoted
A · B, is a second-order tensor with components [A · B]i j = [A]im [B]m j . The action
of a second-order tensor A on a vector a is given by [A · a]i = [A]i j [a] j . The
tensor product of two vectors a and b is a second-order tensor D = a ⊗ b with
[D]i j = [a]i [b] j . Gradients of a quantity {•} with respect to the material and spatial
configuration are denoted as ∇X {•} and ∇x {•}, respectively. Let I and i denote the
identity tensors in the material and spatial configurations, respectively. Divergence
operators with respect to the material and spatial configurations are denoted as Div
and div, respectively, and are defined by
Div {•} := ∇X {•} : I, div {•} := ∇x {•} : i .
Greek superscripts are used to define the atomistic quantities. Thus, any scalar g or
vector g quantity attributed to particle α is denoted as gα or gα , respectively. Time
is denoted τ . The total time derivative of gα is denoted gα;τ . The L2 norm of a vector
√
g is denoted |g| = g · g.
2 Atomistic to Continuum
In this section we formulate a framework to analyse the results of atomistic simula-

tions. We follow the idea of Lagrangian spatial averaging proposed in [19]. Thereafter
we complete this approach by introducing statistical averaging based on the general
results obtained in [4].
Molecular dynamics simulations deal with systems containing a large number
of particles N . The particles are often considered to have translational degrees of
freedom only. Thus, the state of the system is fully described by the following phase-
space coordinates
z = {x = {xα }, p = {pα }} ∈ E3N × E3N (1)
where xα is the position and pα is the linear momentum of particle α.

The motion of the system follows the second Newton law i.e.
pα;τ = f α (2)
where f α is the total force acting on particle α.

Forces acting on each particle can be decomposed into external (with respect to
the system) f αext and internal f αint forces, i.e.
f α = f αint + f αext . (3)
The probability density function
W = W(z, τ ) (4)
gives the probability of the state z per unit volume in phase space P. The following
normalization holds
W(z, τ ) dz = 1. (5)
P
An expectation value of a quantity f = f(z) is calculated as

f(z) P = f(z)W(z, τ ) dz. (6)
P

We will also denote f(z) P as the probability average of f(z).
2.1 Space-Probability Averaging
In order to link the particle system to the continuum one, we follow the methodology
originally proposed in [10, 16]. However, instead of introducing averaging based
on the current particle position, we use the position of the particles in the reference
configuration Xα , as in [19].
Let us define a (space) ς -distribution of a property g attributed to each particle α
at material position X as
ς ς
g(z, X) = g(z, X) := gα (z)w(Xα − X) (7)
α α
where w(Y) is the space averaging kernel which satisfies the normalization conditions

w(Xα − X) dX = 1 (8)
S
in the vector space of translations S . A discussion on the reference positions for

each particle Xα is defered until the end of this section.
The total property density per unit volume in space

ς ς
g (X, τ ) := g(z, X)W (z, τ ) dz = w(Xα − X) gα (z)W(z, τ ) dz = g(z, X) P
P α P
(9)
is a sum of probabilities to find the property g α (z) attached to the particle α at X; it
can also be considered as a probability average of the ς -distribution of property.
In [4] we have shown that time differentiation and statistical averaging commute:

∂τ f(z) P = f;τ (z) P . (10)
Thus for any continuum field defined by Eq. (9), its material rate equals

g;τ (X, τ ) = gα;τ (z)w(Xα − X) = w(Xα − X) gα;τ (z)W (z, τ ) dz.
P P
α α
(11)
2.2 Balance of Mass and Linear Momentum
Let us define the following continuum fields:

• referential mass density

ρ(X) := Mα w(Xα − X)W(z, τ ) dz = Mα w(Xα − X), (12)
α P α
• displacement

α P Mα [xα − Xα ]w(Xα − X)W(z, τ ) dz
u(X, τ ) := . (13)
ρ(X)
The balance of mass trivially follows from Eqs. (12) and (11) as
ρ;τ (X) = 0. (14)
An obvious requirement for displacement is u(X, 0) = 0, which leads to


Mα xα W (z, 0) dz w(Xα − X) = Mα Xα w(Xα − X). (15)
α P α
Since the equation above should hold for arbitrary averaging kernels and for any
material position X, we can conclude that

α
X = xα W (z, 0) dz. (16)
P
That is, the statistically averaged positions of particles in the reference configuration
has to be taken as the reference positions. Note, that in [19] the authors limit the
choice of the reference configuration to zero temperature (molecular statics). This
precludes the formulation for media for which zero-temperature configurations are
not computationally well defined, e.g. polymers.
Using Eq. (11), the continuum velocity follows as

α P Mα vα w(Xα − X)W(z, τ ) dz
v(X, τ ) := . (17)
ρ(X)
Another kinematic quantity which is readily obtainable is the deformation gradient

F, which follows from Eq. (13) as

α α α
α M [u − u(X)] ⊗ ∇X w(X − X)
P
F=I+ . (18)
ρ(X)
Using the expression for the velocity Eq. (17) and recalling Newton’s law (2), we
obtain the balance of linear momentum as

[ρ(X)v(X, τ )];τ = f αint w(Xα − X)W(z, τ ) dz + b, (19)
α P
where we have denoted the body forces as

b(X, τ ) = f αext w(Xα − X)W(z, τ ) dz. (20)
α P
The first term on the right hand side of the above equation is the divergence of the
Piola stress P, i.e.
αβ
DivP(X, τ ) = f αint w(Xα − X)W(z, τ ) dz = f̄ w(Xα − X), (21)
α P αβ
where

αβ
f̄ = f αβ W(z, τ ) dz = f αβ (22)
P P
is the statistical averaged interaction force between particles.

The obvious theoretical difference between our formulation and the one in [19]
is statistical averaging. However, since
αβ βα
f̄ = −f̄ (23)
all the mathematical developments of [19] are applicable to our formulation. As a

result, one of the possible solutions to Eq. (21) in line with the proposal by Hardy
[9] reads
1 αβ
P(X, τ ) = − f̄ ⊗ Xαβ wαβ (24)
2 α
with
αβ α β
1

w (X, X , X ) = w Xα − X − aXαβ da. (25)
0
wαβ can be pre-calculated once all the atomic

For each material point X, the function
reference positions are available, thereby dramatically reducing the computation
time.
3 Continuum Theory with Surface Energy
The purpose of this section is to summarise certain key concepts in nonlinear con-
tinuum mechanics. Consider a continuum body that takes the material configuration
B0 at time τ = 0 and the spatial configuration Bτ at time τ > 0. The placements x
and X in the spatial and the material configurations, respectively, are related by the
invertible (nonlinear) deformation map x = y (X, τ ). The associated deformation
gradient or rather the invertible linear tangent map between material and spatial line
elements dx and dX, is defined as
F := ∇X y (X, τ ) . (26)
The boundary of the continuum body is described (or rather covered) by a two-
dimensional surface in the three-dimensional embedding Euclidean space defined
by S0 = ∂B0 and Sτ = ∂Bτ . The boundary placements x and
X in the spatial
and the material configurations, respectively, are related by the invertible nonlinear
deformation map x = y(
X, τ ) = y(X ∈ S0 , τ ). The boundary deformation gradient

F, the non-invertible linear surface tangent map, between line elements d X and dx
is defined as

F := ∇Xy(
X, τ ) . (27)
Table 1 Local balance equations in the bulk and on the surface in material configurations
Linear momentum DivP + b = 0 in B0

D iv
P + bp − P · N = 0 on S0
Angular momentum F · P t = P · Ft in B0

F ·Pt = P ·Ft on S0
Energy P : ∇X v − DivQ = ∂τ E in B0

P:∇ Xv−D ivQ+Q
p + Q · N = ∂τ E on S0
Entropy η = ∂τ S + DivH ≥ 0 in B0
η = ∂τ
S+D −H
ivH p + H · N ≥ 0 on S0
Temperature evolution c ∂τ T = −DivQ + Q + T P,T : ∂τ F in B0
= −D
c ∂τ T ivQ+Q+ T P,T : ∂τ
F on S0
Heat flux Q = −J k F−1 · F−t · ∇X T in B0
= − J
Q k
F−1 ·
F−t · ∇
XT on S0
The notation {•}p is employed to denote externally prescribed quantities [11]
Here, ∇ X {•} = ∇X {•} ·

I, where
I := I − N ⊗ N denotes the mixed-variant surface
unit tensor in the material configuration with N being the outward unit normal to S0 .
It is important to note that the boundary is assumed to be permanently attached to
the enclosed continuum, i.e. x = x|∂ Bτ . Furthermore, we assume that the temperature
of the boundary, denoted by T , is related to that of the continuum body, denoted by

T , via the relation T = T |∂ Bτ .
The governing equations in the bulk and on the surface, in the material configura-
tion are summarised in Table 1, see [12, 13] for further details on numerical as well
as theoretical aspects. For the sake of simplicity it is assumed that heat and entropy
sources in the bulk vanish.
The Piola stress and body force in the bulk are denoted as P, b, respectively. P,
bp
p
and Q denote the surface Piola stress, the (externally) prescribed traction and heat
flux on the surface, respectively. The bulk internal energy per unit reference volume
is denoted E and E represents the boundary internal energy per unit reference area.
The boundary entropy flux is denoted H and H p denotes the prescribed entropy flux
to the surface. Moreover, η ≥ 0 and η ≥ 0, assumed to be positive, are the bulk
entropy production and the separate boundary entropy production. The bulk internal
entropy per unit reference volume is denoted by S and S represents the boundary
internal entropy per unit reference area. The coefficient of thermal conductivity in
the bulk and on the boundary associated with the current configuration are denoted
k ≥ 0 and k ≥ 0, respectively. The heat capacities in the bulk and on the surface are
denoted c and c, respectively.
The Piola stress in the bulk is calculated via the relation
1
P = 21 F · [C : [F · Ft − I]] − a [T − T0 ] [1 − ln J ] F−t with J = DetF , (28)
J
in which C denotes the fourth-order constitutive tensor associated with cubic
anisotropic materials and a = 3κζ with ζ being the heat expansion coefficient
and κ the bulk modulus.1 The surface Piola stress, which follows from an isotropic
surface free energy, reads
−t

P= μ
λ ln J− F +μ γ J
F+ F−t with J = D
et
F, (29)
where μ and
γ denotes the surface tension and λ are solid-like (elastic) surface mate-
rial parameters. Without detailed experimental evidence it is assumed that surface
stress is independent of the temperature.
4 Numerical Examples
All the numerical examples are for a face-centered cubic (FCC) crystal of copper.
From the atomistic perspective it is modeled using the EAM potential by Foiles et al.
[5], which has the lattice parameter a = 3.615 Å at zero temperature and the cut-off
radius 4.94 Å. Atomistic simulations are performed using the LAMMPS open-source
software [17]. A time-step of 1 fs was chosen. The quartic averaging kernel was used
for evaluation of continuum quantities from atomistic simulations:
35
w(r ) = (1 − r 2 )2 (30)
8V
with r = |Xα − X|/R and V = 43 π R 3 . The averaging radius R was chosen to be

3.0a. As it is often done based on the ergodicity assumption, a statistical averaging
was replaced by averaging in time.
4.1 Bulk Properties
The cubic anisotropy continuum parameters, obtained from molecular statics sim-
ulations [3], are (in GPa): C11 = 167.26, C12 = 124.15, C44 = 76.44. Periodic
boundary conditions are considered for the representative volume element (RVE) at
both the atomistic and continuum descriptions.
The heat expansion coefficient a is obtained as follows. The RVE is constrained
to remain at the initial volume, while heating to temperature T = 100 K. For such
a case, F = I and the Cauchy and Piola stresses coincide. By evaluation of the
macroscopic Cauchy stress via the Viral expression
1 αβ
− f (τ ) ⊗ xαβ P (31)
V
αβ
1 The procedure in Sect. 4.1 renders a directly.

and using it in the constitutive equation (28), the isotropic heat expansion coefficient
is obtained as a = 0.0066 GPa/K.
The volumetric heat capacity is obtained from the uniaxial tension of the RVE
under adiabatic conditions. The procedure used to heat the RVE and apply the uniaxial
tension in the atomistic representation is as follows:
I Preparation of the initial configuration
(a) heat to a non-zero temperature at constant volume (1 ns);
(b) relax to zero pressure at the prescribed temperature (1 ns). The statistical
averaged volume of the RVE is calculated;
(c) change the RVE volume to the one calculated2 in step (b) and let the system
equilibrate (0.5 ns). The reference atom positions are calculated during this
step;
II Uniaxial loading (1 iteration)
(a) apply a uniaxial deformation rate and deform to the desired state in 0.5 ns.
For a uniaxial tension of 1 % that corresponds to the rate of 2 × 107 1/s;
(b) equilibrate the system (0.5 ns);
(c) calculate the required continuum fields (0.5 ns). The calculation of the sta-
tistical averaged quantities is performed at an interval of 100 steps.
The deformation procedure can be applied iteratively if needed. It was shown pre-
viously for the potential used here [19], that the time averaged quantities converge
within 0.5 ns.
Applying a uniaxial deformation of 4 % to the bulk in 4 iterations, one obtains
the change of stress and temperature with respect to strain. From the continuum
perspective, for the homogeneous case, the heat equation is trivially integrated to
obtain the stationary solution as
T
c ln = −a tr(Δε). (32)
T0
From this, the volumetric heat capacity was calculated as c = 0.00432 GPa/K. The
comparison between the atomistic results and the small and large strain continuum
formulations are presented in Fig. 1.
It is obvious that the material behaves in non-linear elastic way, thus the materi-
ally linear constitutive law here considered, see Eq. (28), agrees well with the MD
simulation only for relatively small strains (about 1 %). The implication of that will
be clear in the next numerical example. From Fig. 1 one can also conclude that the
Cauchy stress, obtained from the Lagrangian space-probability averaging approach
is consistent with the Virial pressure Eq. (31) for the homogeneous deformation of
the crystal.
2 The change of volume to the statistically averaged volume is important to keep the molecular
dynamics system at zero pressures during this step.
7 6
atomistic (local) atomistic (local)
6 continuum (small strain) continuum (small strain)
atomistic (global) 5 atomistic (global)
5 continuum (large strain) continuum (large strain)
4
σ11 [GPa]
σ22 [GPa]
4
3
3
2
2
1 1
0 0
0 0.01 0.02 0.03 0.04 0 0.01 0.02 0.03 0.04
F11 - 1.0 F11 - 1.0
6 101
atomistic (local) atomistic (global)
continuum (small strain) 100 continuum (small strain)
5 atomistic (global)
continuum (large strain) 99
4
σ33 [GPa]
98
T [K]
3 97
96
2
95
1
94
0 93
0 0.01 0.02 0.03 0.04 0 0.01 0.02 0.03 0.04
F11 - 1.0 F11 - 1.0
Fig. 1 A comparison of Cauchy stress and temperature as calculated from the atomistic and con-
tinuum approaches for the uniaxial tension of homogeneous RVE under adiabatic conditions
The Green–Kubo [7, 15] approach is used to calculate the heat conduction coeffi-
cient. The heat conduction coefficient is related to the time integral of the correlation
function as: ∞
V
k= J(0) · J(τ ) dτ (33)
3 kB T 2 0
where heat flux is defined as

1 αβ vα + vβ αβ
J := u α vα + f · x . (34)
α
2 2
αβ
The integral is usually approximated either numerically (e.g. trapezoidal rule), or ana-
lytically by fitting an exponential function. For the studied material, both approaches
give similar values. As a final isotropic value we chose the averaged numerically
integrated value over three dimensions: k = 55.96 W/mK. Note, that the thermal
transport in metals is not due to phonon, but electron transport [14]. Therefore, MD
simulations underestimate the conductivity.
4.2 Plate with a Hole
Having obtained all the bulk properties (both elastic and thermal) from molecular
dynamics, surface effects can be studied. As a numerical example the uniaxial tension
of a plate with a hole is considered. We aim to compare atomistic simulations to
continuum predictions using both integral and local quantities.
From the atomistic simulation perspective, we consider a square plate oriented
along the crystallographic axes. The number of unit cells in the third dimension was
fixed to 3, whereas in the other two dimension it was chosen to be 40. The size effect
was studied in our previous contribution [3]. Here we restrict ourselves to a single
size. The procedure to generate the reference configuration at non-zero temperature,
as well as the application of the uniaxial loading, was explained in the previous
section.
The free surface is introduced as follows: in the initial (zero temperature) config-
uration all atoms which lie within a cylinder of a given diameter (1/5 of the plate
length) are marked for deleting. After obtaining the equilibrium reference config-
uration at non-zero temperature, the atoms are removed from the system. Due to
missing neighbors, surface tension occurs and the system is no longer in local or
global thermodynamical equilibrium. In order to evaluate the continuum quantities
which correspond to the steady state solution, the following procedure is applied.
I→II Relaxation under surface tension
(a) equilibrate the system (0.5 ns)3 ;
(b) calculate the continuum fields from atomistic simulation (0.5 ns).
Starting from the relaxed configuration, the uniaxial loading is applied as described
previously.
The Cauchy stresses and displacements from the atomistic and continuum
approaches are compared in the material (Lagrangian) coordinates. The results below
are evaluated along the x = 0 and y = 0 lines. The stress and displacements profiles
for the relaxed and loaded configuration are presented in Figs. 2 and 3, respectively.
In the continuum model, due to the two-dimensional nature of the problem, the num-
ber of surface material parameters reduce to two, which are obtained by a fitting
procedure of the atomistic results. The procedure to obtain continuum surface mate-
rial parameters is elaborated upon in [3]. In addition to the results obtained from the
molecular dynamics simulations, results from the molecular statics simulations [3]
are included.
Both a quantitatively and qualitatively good agreement between the discrete and
the continuum models is observed. Note, that the continuum solution predicts higher
stresses compared to the ones calculated from the MD simulation. The reason for this
is, most likely, the materially linear constitutive law here used, which corresponds
to the MD simulation only at relatively small strain levels (see Fig. 1).
3 The continuum simulation will verify that this time is sufficient to reach the global thermodynam-
ical equilibrium.
0.3 0.15
continuum 0.1 continuum
0.25 MD MD
σ11 [GPa]
σ11 [GPa]
0.2 MS 0.05 MS
0
0.15
-0.05
0.1 -0.1
0.05 -0.15
0 -0.2
40 60 80 100 120 140 40 60 80 100 120 140
x [A] y [A]
0.2 0.3
0.15 continuum continuum
MD 0.25 MD
σ22 [GPa]
σ22 [GPa]
0.1
MS 0.2 MS
0.05
0 0.15
-0.05 0.1
-0.1
-0.15 0.05
-0.2 0
40 60 80 100 120 140 40 60 80 100 120 140
x [A] y [A]
0 0
-0.02 -0.02
Ux [A]
Uy [A]
-0.04 -0.04
continuum continuum
-0.06 MD -0.06 MD
MS MS
-0.08 -0.08
40 60 80 100 120 140 40 60 80 100 120 140

x [A] y [A]
Fig. 2 Microscopic stresses and displacements due to surface tension
By looking at the evolution of the temperature (Fig. 4) during relaxation under

surface tension, we observe that the steady state is reached in 0.005 ns. Therefore,
we can be confident, that the results measured from the atomistic simulations after
equilibration for 0.5 ns indeed correspond to the steady state solution.
5 Discussion and Conclusions
In this contribution we compared a thermo-elastic continuum mechanics approach

enhanced with surface energy to atomistic (MD) simulations at non-zero temperature.
A Lagrangian statistical averaging formulation was developed and applied to the
numerical examples of a FCC crystal to obtain the continuum fields. The expression
for the atomistic initial configuration at a non-zero temperature was obtained. The
balance laws of mass and linear momentum were derived. The complete procedure
for obtaining the bulk elastic and thermal properties was presented. The procedure
is general and can be applied to other materials, e.g. polymers.
The bulk uniaxial loading under adiabatic conditions showed that the local expres-
sions for the continuum fields (stress, deformation gradient, temperature) give results
consistent with the ones obtained from the global expressions often used in MD sim-
ulations (Virial pressure, temperature). That is the case even for the considerable
2 2.8
2.4
1.6
σ11 [GPa]
σ11 [GPa]
2
1.2 1.6
1.2
0.8
0.8
0.4 0.4
40 60 80 100 120 140 40 60 80 100 120 140

x [A] y [A]
1.6 1.6
σ22 [GPa]
σ22 [GPa]
1.2 1.2
0.8 0.8
0.4
0.4
40 60 80 100 120 140 40 60 80 100 120 140

x [A ] y [A]
1.6 0.25
1.4 MD
0.2 MS
1.2
0.15 continuum (0.50)
Ux [A]
Uy [A]
1 continuum (0.35)
0.8 0.1 continuum (0.15)
0.6 0.05 continuum (0.05)
0.4
0
0.2
0 -0.05
40 60 80 100 120 140 40 60 80 100 120 140
x [A] y [A]
Fig. 3 Microscopic stresses and displacements at uniaxial loading of 1 % (for the continuum
solution, an evolution in time is given)
-0.022 -0.02
t = 0.0000 t = 0.0000
-0.024 t = 0.0005 -0.022 t = 0.0005
t = 0.0010 t = 0.0010
-0.026 t = 0.0015 -0.024 t = 0.0015
t = 0.0020 t = 0.0020
ΔΘ [K]
ΔΘ [K]
-0.028 t = 0.0025 -0.026 t = 0.0025

t = 0.0030 t = 0.0030
-0.03 t = 0.0035 -0.028 t = 0.0035
t = 0.0040 t = 0.0040
-0.032 t = 0.0045 -0.03 t = 0.0045
t = 0.0050 t = 0.0050
-0.034 -0.032
-0.036 -0.034
1 1.5 2 2.5 3 3.5 4 4.5 5 1 1.5 2 2.5 3 3.5 4 4.5 5
x/R [] y/R []
Fig. 4 Evolution of the temperature (continuum solution) in time during relaxation under surface
energy (left) and uniaxial loading (right)
deformation of 4 %. Thus, the validity of the local atomistic expression of the con-
tinuum fields was shown, at least for homogeneous solution.
The investigation of a numerical example with surface energy (plate with a hole)
has confirmed both the applicability of the averaging procedure as well as the validity
of the thermo-elastic formulation of the continuum enhanced with a surface energy.
Local quantities evaluated using both the continuum and discrete approaches are in
excellent agreement.
The extension of this contribution to the case of polymers with nanoparticles is

work in progress.
Acknowledgments The first author is grateful to the German Science Foundation (Deutsche
Forschungs-Gemeinschaft), grant STE 544/46-1, for their financial support. The support of this
work by the ERC Advanced Grant MOCOPOLY is gratefully acknowledged by the second and
third authors.
References
1. Admal, N.C., Tadmor, E.B.: A unified interpretation of stress in molecular systems. J. Elast
100, 63–143 (2010)
2. Daher, N., Maugin, G.A.: The method of virtual power in continuum mechanics application to
media presenting singular surfaces and interfaces. Acta Mech. 60(3–4), 217–240 (1986)
3. Davydov, D., Javili, A., Steinmann, P.: On molecular statics and surface-enhanced continuum
modeling of nano-structures. Comput Mater Sci [accepted]
4. Davydov, D., Steinmann, P.: Reviewing the roots of continuum formulations in molecular
systems. Part I: Particle dynamics, statistical physics, mass and linear momentum balance
equations. Math Mech Solids [accepted]
5. Foiles, S.M., Baskes, M.I., Daw, M.S.: Embedded-atom-method functions for the fcc metals
Cu, Ag, Au, Ni, Pd, Pt, and their alloys. Phys. Rev. B 33(12), 7983–7991 (1986)
6. Gibbs, J.W.: The Scientific Papers of JW Gibbs, vol. 1. Dover Publications, New York (1961)
7. Green, M.S.: Markoff random processes and the statistical mechanics of time-dependent phe-
nomena. II. Irreversible processes in fluids. J. Chem. Phys. 22(3), 398–413 (1954)
Mech. Anal. 57(4), 291–323 (1975)
9. Hardy, R.: Formulas for determining local properties in molecular-dynamics simulations—
shock waves. J. Chem. Phys. 76, 622–628 (1982)
10. Irving, J., Kirkwood, J.G.: The statistical mechanical theory of transport processes. IV. The
equations of hydrodynamics. J. Chem. Phys. 18(6), 817–829 (1950)
11. Javili, A., McBride, A., Mergheim, J., Steinmann, P., Schmidt, U.: Micro-to-macro transitions
for continua with surface structure at the microscale. Submitted
12. Javili, A., Steinmann, P.: On thermomechanical solids with boundary structures. Int. J. Solids
Struct. 47(24), 3245–3253 (2010)
III: The thermomechanical case. Comput. Methods Appl. Mech. Eng. 200(21–22), 1963–1977
(2011)
14. Jones, W., March, N.H.: Theoretical Solid State Physics. Dover Publications, New York (1985)
15. Kubo, R.: Statistical-mechanical theory of irreversible processes. I. General theory and simple
applications to magnetic and conduction problems. J. Phys. Soc. Jpn. 12(6), 570–586 (1957)
16. Noll, W.: Die herleitung der grundgleichungen der thermomechanick der kontinua aus der
statistichen mechanik. J. Ration. Mech. Anal. 4, 627–646 (1955)
17. Plimpton, S.: Fast parallel algorithms for short-range molecular dynamics. J. Comput. Phys.
117(1), 1–19 (1995). http://lammps.sandia.gov
18. Shenoy, V.B.: Atomistic calculations of elastic properties of metallic fcc crystal surfaces. Phys.
Rev. B Condens. Matter Mater. Phys. 71(9), 094104 (2005)
19. Zimmerman, J.A., Jones, R.E., Templeton, J.A.: A material frame approach for evaluating
continuum variables in atomistic simulations. J. Comput. Phys. 229, 2364–2389 (2010)
Surface Mechanics and Full-Field
Measurements: Investigation
of the Electro-Elastic Coupling
Cécile Flammier, Frédéric Kanoufi, Sorin Munteanu, Jean Paul Roger,

Gilles Tessier and Fabien Amiot
Abstract Many proofs of concept studies have established the mechanical sensitivity
of functionalized microcantilevers to a large spectrum of target molecules. However,
moving to real-life applications also requires the monitored mechanical effect to be
highly specific. On the other hand, describing the involved surface effects in the con-
tinuum mechanics framework is still challenging. Several attempts to overcome the
Stoney’s surface stress failure to satisfy field equations tend to show such a descrip-
tion has to be non-local, so that at least one ‘characteristic length’ parameter has to
be used. The consequence is twofold: first, suited modelings have to be developed
to describe the surface effects at the cantilever scale; and second, the involved char-
acteristic length is (thermodynamically) connected to the molecular mechanisms at
the cantilever surface, and may therefore be a key parameter for the target molecules
identification. This requires to experimentally access displacement fields induced
by the molecular interactions under scrutiny. A set-up providing mechanical and
C. Flammier · F. Amiot (B)

FEMTO-ST Institute CNRS-UMR 6174/UFC/ENSMM/UTBM, 24 chemin de l’Épitathe,
Besançon 25000, France
e-mail: cecile.flammier@femto-st.fr
F. Amiot
e-mail: fabien.amiot@femto-st.fr
F. Kanoufi · S. Munteanu
PECSA CNRS-UMR 7195/ESPCI ParisTech, 10 rue Vauquelin, Paris 75005, France
e-mail: frederic.kanoufi@espci.fr
S. Munteanu
e-mail: sorin.munteanu@espci.fr
J. P. Roger · G. Tessier
Langevin Institute CNRS-UMR 7587/ESPCI ParisTecr h, 1 rue Jussieu,
Paris Cedex 0575238, France
e-mail: jean-paul.roger@espci.fr
G. Tessier
e-mail: gilles.tessier@espci.fr

60 C. Flammier et al.
chemical fields along the cantilever is thus implemented focusing on cases where
the cantilever’s surface reacts heterogeneously. The large amount of data obtained
using full-field set-ups is redundant from the mechanical point-of-view, and this
redundancy is used to identify some of the key parameters describing the mechani-
cal surface effects. Results obtained when studying the electro-elastic coupling in a
non-adsorbing case are presented.
1 Introduction
Many proofs of concept studies have established the mechanical sensitivity of func-
tionalized microcantilevers to a large spectrum of target molecules [1], ranging from
alcanethiols [2] and explosives [3] to DNA [4] and proteins [5]. These microcantilever
sensors are therefore considered as a promising tool for diagnostics. The main diffi-
culty to overcome is however to improve the sensor’s specificity in order to make it
usable even though the target molecule is at trace level among highly concentrated
products [6].
From the experimental point of view, most of the above-mentioned studies have
been carried out using the so-called optical lever technique: a laser beam is focused
at the free-end of a cantilever, and the reflected spot position is monitored with a
position sensitive detector (PSD). The chemically induced mechanical effects are
therefore integrated over the cantilever’s length to displace the spot on the PSD.
This scalar information is then translated to a scalar ‘surface stress’ thanks to Stoney
equation [7], but this ‘surface stress’ does not satisfy any field equation and is not
connected to the Cauchy stress in the bulk.
For different reasons, several groups started few years ago to develop set-ups
providing a richer kinematic information. Some of them added a scanning stage
to the optical lever platform [8, 11, 12] when others chose to implement different
interferometric imaging systems [9, 10, 13, 14], thus providing a field information at
the micrometer scale. This contribution therefore proposes to give an insight into the
advances allowed by this redundant information, focusing on the modeling issues one
has to deal with when considering the surface phenomena involved at the micrometer
scale.
The available frameworks for dealing with surface effects are then first reviewed
and an identification method is described. Some results concerning the electro-elastic
coupling are presented.
2 Surface Mechanics
This section is intended to quickly review the main theories available to describe
surface effects.
Surface Mechanics and Full-Field Measurements 61
2.1 Second Gradient of Strain and Surface Tension
In order to describe surface effects within a thermodynamically grounded framework,

Mindlin [15] proposed to describe the thermodynamic state of an isotropic, linear
elastic material point with the usual finite strain tensor ε1 (related to the displacement
u), along with its first gradient ε2 and second gradient ε3 . These are defined by
1
ε1 = ∇u + ∇ut , ε2 = ∇∇u, ε3 = ∇∇∇u (1)
2
The Gibbs free enthalpy G therefore depends on these three parameters

G = G ε1 , ε2 , ε3 (2)
so that it depends on 54 independent variables. The associated stress tensors are

defined by the derivatives of the Gibbs free enthalpy with respect to the corresponding
variables, so that the equilibrium condition reads

div τ 1 − divτ 2 + div divτ 3 + f = 0
with
∂G ∂G ∂G
τ1 = , τ2 = , τ3 = (3)
∂ε1 ∂ε2 ∂ε3
in the volume, in addition to 9 boundary conditions. f denotes the body forces. The
constitutive equation is then obtained by considering all the possible combinations
leading to scalars:
λ 1 1
G = ε ε + μεi1j εi1j
2 ii j j
+ a1 εi2j j εikk
2
+ a2 εiik
2 2
εk j j + a3 εiik
2 2
ε j jk + a4 εi2jk εi2jk + a5 εi2jk εk2ji
+ b1 εii3 j j εkkll
3
+ b2 εi3jkk εi3jll + b3 εii3 jk ε3jkll + b4 εii3 jk εllk
3
j
+ b5 εii3 jk εll3 jk + b6 εi3jkl εi3jkl + b7 εi3jkl ε3jkli
+ c1 εii1 ε3j jkk + c2 εi1j εi3jkk + c3 εi1j εkki
3
j
+ b0 εii3 j j (4)
λ and μ are the Lamé parameters, an , cn and b0 scale as μl 2 and bn as μl 4 , where l

stands for a length, which is shown to scale as the material thickness affected by the
surface [15]. It should be highlighted that the term b0 εii j j is the unique term linear
with respect to one of the observable variables. This describes an equivalent for the
surface tension in liquids. The Gibbs free enthalpy without any external load reduces
to :
b0
G = (divu) (5)
2
This therefore results in a self-equilibrated force without any external load, and
also describes the energy for the creation of a new surface. All the ingredients are
therefore present in this second strain-gradient theory in order to properly describe
surface effects in solids. This description is also firmly thermodynamically grounded,
and involves characteristic lengths (and especially a cohesion modulus b0 ).
Even though closed-form solutions are available for some simple problems [15],
this very rich modeling makes use of 54 independent kinematic variables and 18
constitutive parameters. To the best of our knowledge, implementing an identification
procedure for these constitutive parameters remains a challenge to take up, so that
alternatives and compromises are to be found to describe these surface effects.
2.2 Surface Effects as Special Boundary Conditions
One of these alternatives has been proposed by Gurtin and Murdoch [16, 17]. The
material body is considered to be divided into bulk and surface (see Fig. 1). The
bulk is considered to be isotropic and linear elastic and is described using a classical
framework (solely) based on ε1 . Denoting Θ the surface stress in the membrane, the
equilibrium condition reads
divs Θ = τ 1 n
Denoting Ξ the surface strain defined by
1
Ξ= ∇s u + ∇s ut
2
the constitutive equation reads
Θ = σ I + 2 (μ0 − σ ) Ξ + (λ0 + σ ) (T r Ξ ) I + σ ∇s u,
where λ0 , μ0 are the constitutive parameters of the membrane, and σ is the residual
stress tensor. It should be highlighted that the constitutive parameters of the mem-
brane may be negative. With this modeling, Gurtin and Murdoch propose another
Fig. 1 Surface description for

Gurtin-Murdoch theory
definition for the surface stress, satisfying field equations within a linear elasticity
framework. Again, this modeling implicitly involves internal lengths, even though it
is much simpler than a second strain-gradient theory. However, it has no thermody-
namic grounding, and the surface stiffness identification is questionable [17].
3 Asymptotic Analysis
3.1 Modeling
In order to have at disposal a rather simple, though thermodynamically grounded,

modeling to describe surface effects and to circumvent the difficulties arising with the
previous modelings, one proposes in the following to restrict to a specific geometry
(namely Euler-Bernoulli cantilever beams) and to describe chemical effects at the
surface thanks to a thermodynamically equivalent layer (TE layer), which is ascribed
to deform together with the cantilever [18], see Fig. 2.
The Gibbs free enthalpy is divided into chemical and mechanical parts:
G = Gc + Gm
The TE layer constitutive parameters are thus chosen to (mechanically) mimic the
chemical part of the free enthalpy close by the equilibrium. A Taylor expansion of the
chemical contribution in equilibrium’s vicinity with respect to the surface amount
sets the virtual layer stiffness El and free strain (eigenstrain) εl . Such an expansion
also yields the layer’s stiffness scales as El = K l ξ −1 , where
tl
=ξ
tc
is the ratio of the layer and cantilever thicknesses [18]. Assuming the constitutive
parameters for the cantilever and the TE layer are known, the problem can be solved
for a vanishing TE layer thickness, assuming displacements continuity at the interface
and using an enriched beam kinematic, thanks to an asymptotic analysis [18]. This
provides the displacement field of the cantilever so that the energy conversion (from
Fig. 2 Surface description

for the asymptotic cantilever
theory
chemical to mechanical) can be studied. The conversion efficiency from chemical to

mechanical energy can then be shown to significantly depend on the functionalization
pattern for instance, thereby proving the described modeling is non-local and features
characteristic lengths [18]. It is also quite simple since it only involves two parameters
(K l , εl ) to describe surface effects.
3.2 Identification Procedure
These two parameters can be experimentally approached using a multiple wave-

lengths reflection microscope, for which the collected intensity Im (P, λ) at point P
and for the wavelength λ formally reads
Im (P, λ) = Ia (P, λ) [1 + Rwd (λ, P) + Rwi (P)] (6)
Acquiring images for different wavelengths allows one to simultaneously retrieve

wavelength-dependent Rwd and wavelength-independent Rwi reflectivity change
fields [19]. The former is assumed to qualitatively describe chemical modifications
of the surface under scrutiny, whereas the latter is the consequence of mechanical
effects. A calibration procedure is implemented to quantitatively relate Rwi to the
surface rotation field through a sensitivity field g(x):
dw
= g −1 (x)Rwi (x), (7)
dx
where w(x) denotes the out-of-plane cantilever displacement field and x is the can-
tilever beam axis (the beam corresponds to − 2l ≤ x ≤ 2l ). Assuming the TE layer
eigenstrain is related to the wavelength-dependent reflectivity change field through,
εl = F Rwd , (8)
the in-plane displacement of the lower side of the TE layer may be easily calculated.
The in-plane displacement of the upper side of the cantilever is derived from the
out-of-plane displacement w(x), so that the displacements continuity at the interface
reads, at equilibrium:
x x x η 3
tc d 2w tc Ec d w
− dη = F Rwd (η)dη + 1+ dτ dη
2 − 2l dx2 − 2l 6 Kl l
−2 −2 l dx3
for any x (pixels) along the beam. E c denotes the cantilever’s Young’s modulus.
Assuming the cantilever is at equilibrium, the continuity condition is recast
x
Ec
tc dw dw l
F Rwd (η)dη + 4+
(x) − −
− 2l 6
Kl dx dx 2
2
tc Ec tc l d w l
− 4+ − x+ 2
− =0 (9)
6 Kl 2 2 dx 2
It should be highlighted that the Eq. (9) can written for any x along the beam. Deal-
ing with full-field informations, Eq. (9) is written for any pixel along the cantilever
beam, thus providing an over-determined linear system, whose mainunknowns are
tc Ec
the strain-reflectivity coefficient F and the characteristic length 6 4 + K l . The

other terms Rwd , dw d x are obtained from the measured fields. It should be high-
2
lighted that the curvature at the clamping dd xw2 − 2l is rather difficult to measure
since it requires a fairly rich kinematic description. Instead of solving Eq. (9) in a
least-square sense to get access to the TE layer stiffness parameter K l , it is proposed
to filter Eq. (9) by making use of an orthonormal polynomial basis {Pn (x)}. Pn (x) is
of order n and satisfy
l
2
Pn (x)Pm (x)d x = δmn , (10)
− 2l
where δmn stands for the Kronecker symbol. Let us denote

x
tc Ec dw dw l
J (x) = F Rwd (η)dη + 4+ (x) − − , (11)
− 2l 6 Kl dx dx 2
l
2
a0 = J (x)P0 (x)d x, (12)
− 2l
l
2
a1 = J (x)P1 (x)d x (13)
− 2l
Filtering out the 0 and 1 order terms in Eq. (9) yields
J (x) − a0 P0 (x) − a1 P1 (x) = 0 (14)
for any x along the beam, so that the poorly measured term
vanishes. The left-hand
side of Eq. (14) is linear with respect to F and t6c 4 + EKcl so that gathering all the
equations obtained for all the pixels yields an over-determined linear system featuring
one equation per pixel for only 2 unknowns. As one is looking for a non-trivial solu-
tion, the resulting equations system is solved using a singular value decomposition
to provide an estimate for the ratio
F
r= (15)
tc Ec
6 4+ Kl
The associated singular value should vanish, thus providing a quality estimator for
the solution. It should be highlighted that the situation with an homogeneous Rwd
would make r disappear from the projected equation.
4 Application to the Electro-Elastic Coupling
The above-described procedure is applied to reflectivity fields obtained using gold-

coated silicon-nitride cantilevers (80 × 15 × 0.75 µm) placed in a 3-electrodes
arrangement with a KCl (5 × 10−3 M) electrolyte. The gold layer is used as the
working electrode. Every cantilever is electrically addressed independently, and its
potential is swept in the double-layer regime (0.04 V ≤ V ≤ 0.4 V vs. Ag/AgCl) at
4 mV/s after an electrochemical cleaning. 160 images are recorded for each wave-
length during three potential cycles, and the obtained reflectivity change fields are
presented on Fig. 3. The cantilever base is at the top of the figure, and the can-
tilever’s tip is at the bottom. The wavelength-dependant Rwd is averaged across the
cantilever’s width and is displayed on Fig. 3a. The Rwd field, which is assumed to
translate the local chemical activity, is rather heterogeneous, thus highlighting the
need for full-field measurements. The Rwi field, which is assumed to represent the
mechanical surface rotation, is plotted after width-averaging on Fig. 3b and denotes
an upward (gold-side) bending of the cantilever when the potential is increased. A
slight strain accumulation can be noticed during the 3 potential cycles.
The sensitivity field is obtained from a calibration procedure to translate the Rwi
field into a rotation field dw
d x . The Rwd and rotation fields are then used to identify
the r ratio, using the identification procedure detailed in Sect. 3.2 for each time step,
independently. The identified r ratio is displayed on Fig. 4. This ratio is found to
be rather independent on the time step and establishes to r −7 × 10−6 µm−1 .
The larger deviations from this value occur for the first steps, where a bubble travels
(a) (b)
Fig. 3 Measured reflectivity changes measured when sweeping the potential: (a) wavelength-
dependant (red) reflectivity change, (b) wavelength-independent reflectivity change
Fig. 4 Identified ratio r along the experiment
nearby the cantilever’s base, as seen in Fig. 3, and at the time steps the sweeping
direction changes, leading to sharp changes in the cantilever’s electrical state so that
the equilibrium condition can no longer be ensured. This tends to indicate that a
single reaction occurs at the cantilever’s surface, and that the chosen modeling is
adequate to describe the involved effects.
5 Conclusion
Considering the fact full-field set-ups are now available to examinate chemo-
mechanical coupling at the micrometer scale, a modeling as well as an identifi-
cation procedure have been proposed to describe chemo-mechanical couplings. It is
demonstrated that parameters (thermodynamically) describing the reaction occuring
at a micro-cantilever’s surface may be retrieved using fields obtained from a multiple-
wavelength reflection microscope. This first attempt also clearly shows that the tra-
ditional cantilever geometry with an homogeneous reaction is not the most adequate
to yield reliable parameters, so that further developments require new geometries to
be devised.
Acknowledgments This work was funded through the NRA grants μ-Ecoliers ANR-08-JCJC-
0088 (F.K., S.M.) and CheMeCo ANR-11-JS09-019-01 (F.A.).
References
1. Lavrik, N.V., Sepaniak, M.J., Datskos, P.G.: Cantilever transducers as a platform for chemical
and biological sensors. Rev. Sci. Instrum. 75(7), 2229–2253 (2004)
2. Berger, R., Delamarche, E., Lang, H.P., Gerber, C., Gimzewski, J.K., Meyer, E., et al.: Surface
stress in the self-assembly of alkanethiols on gold probed by a force microscopy technique.
Appl. Phys. A. 66, S55–S59 (1998)
3. Pinnaduwage, L.A., Boiadjiev, V., Hawk, J.E., Thundat, T.: Sensitive detection of plastic
explosives with self-assembled monolayer-coated microcantilevers. Appl. Phys. Lett. 83(7),
1471–1473 (2003)
4. Fritz, J., Baller, M.K., Lang, H.P., Rothuizen, H., Vettiger, P., Meyer, E., et al.: Translating
biomolecular recognition into nanomechanics. Science 288, 316–318 (2000)
5. Arntz, Y., Seelig, J.D., Lang, H.P., Zhang, J., Hunziker, P., Ramseyer, J.P., et al.: Label-free
protein assay based on a nanomechanical cantilever array. Nanotechnology 14, 86–90 (2003)
6. Arlett, J.L., Myers, E.B., Roukes, M.L.: Comparative advantages of mechanical biosensors.
Nat. Nanotechnol. (2011). doi:10.1038/nnano.2011.44
7. Stoney, G.: The tension of metallic films deposited by electrolysis. Proc. R. Soc. Lond. Ser. A
82, 172 (1909)
8. Mertens, J., Álavarez, M., Tamayo, J.: Real-time profile of microcantilevers for sensing appli-
cations. Appl. Phys. Lett. 87, 234102 (2005)
9. Helm, M., Servant, J.J., Saurenbach, F., Berger, R.: Read-out of micromechanical cantilever
sensors by phase shifting interferometry. Appl. Phys. Lett. 87, 064101 (2005)
10. Wehrmeister, J., Fuss, A., Saurenbach, F., Berger, R., Helm, M.: Readout of micromechanical
cantilever sensor arrays by Fabry-Perot interferometry. Rev. Sci. Instrum. 78, 104105 (2007)
11. Jeon, S., Jung, N., Thundat, T.: Nanomechanics of self-assembled monolayer on microcan-
tilever sensors measured by a multiple-point deflection technique. Sens. Actuators B 122,
365–368 (2007)
12. Watari, M., Lang, H.-P., Sousa, M., Hegner, M., Gerber, C., Horton, M.A., McKendry, R.A:
Investigating the molecular mechanisms of in-plane mechanochemistry on cantilever arrays.
J. Am. Chem. Soc. 129, 601–609 (2007)
13. Amiot, F., Roger, J.P., Boccara, A.C.: Towards massive parallel reading of sensitive mechani-
cal microsensors. SPIE Proceedings: Advanced Biomedical and Clinical Diagnostic Systems,
vol. 4958, pp. 183–188. SanJose CA, USA (2003)
14. Amiot, F., Roger, J.P.: Nomarski imaging interferometry to measure the displacement field of
micro-electro-mechanical systems. Appl. Opt. 45(30), 7800–7810 (2006)
15. Mindlin, R.D.: Second gradient theory of strain and surface tension in linear elasticity. Int. J.
Solids Struct. 1, 417–438 (1965)
Mech. Anal. 57, 291–323 (1975)
17. Gurtin, M.E., Murdoch, A.I.: Surface stress in solids. Int. J. Solids Struct. 14, 431–440 (1978)
18. Amiot, F.: A model for chemically-induced mechanical loading on MEMS. J. Mech. Mater.
Struct. 2(9), 1787–1803 (2007)
19. Garraud, N., Fedala, Y., Kanoufi, F., Tessier, G., Roger, J.P., Amiot, F.: Multiple wavelengths
reflectance microscopy to study the multi-physical behavior of MEMS. Opt. Lett. 36(4),
594–596 (2011)
Effect of a Type of Loading on Stresses
at a Planar Boundary of a Nanomaterial
Mikhail A. Grekov and Yulia I. Vikulina
Abstract A two-dimensional model of an elastic body at nanoscale is considered as

a half-plane under the action of a periodic load at the boundary. An additional surface
stress, and constitutive equations of the Gurtin–Murdoch surface linear elasticity are
assumed. Using Goursat–Kolosov complex potentials and Muskhelisvili technique,
the solution of the boundary value problem in the case of an arbitrary load is reduced
to a hypersingular integral equation in an unknown surface stress. For the case of a
periodic load, the solution of this equation is found in the form of Fourier series. The
influence of the surface stress on the stresses at the boundary of the half-plane under
the tangential and normal periodic loading is analyzed. In particular, it is found out
the size effect which becomes apparent in the dependence of the stresses on a length
of the load period of the order 10 nm. Moreover, the tangential stresses appear under
the action of the normal loads.
1 Introduction
The near-surface effects which are intrinsic to nanomaterials can cause an essential
difference of physical properties of these nanomaterials from the same properties of
macroscale bodies. Thus, physical properties of a nanometer specimen depend on its
size (size effect). For example, Young’s modulus of a cylindrical specimen increases
significantly, when the cylinder diameter becomes very small [10].
As a rule, an ideal effect of a surface stress on the elastic body is not taken
into account at the macroscale because it is insignificant in comparison with the
M. A. Grekov (B) · Y. I. Vikulina

Faculty of Applied Mathematics and Control Processes, Saint-Petersburg State University,
Universitetski pr. 35, St. Petersburg 198504, Russia
e-mail: magrekov@mail.ru
Y. I. Vikulina
e-mail: julia.vikulina@gmail.com

70 M. A. Grekov and Y. I. Vikulina
effect of external forces. The Gurtin–Murdoch theory of surface elasticity which has
obtained rapid development in recent years is extensively used to consider the surface
properties of nanoobjects [1, 7].
In this paper the Gurtin-Murdoch model is applied to a semi-infinite linear elastic
body with plane surface under plane strain conditions. The action of external forces at
the boundary and surface stresses is assumed. Based on Goursat–Kolosove complex
potentials and the Muskhelisvili approach, the boundary value problem is reduced
to a hypersingular integral equation. The solution of this equation in the case of a
periodic loading is used to analyze an influence of the surface stress on the stress
state of the boundary in relation to a change of the period and type of loading.
2 Basic Equations
Consider the elastic half-space its surface has elastic properties differing from those
in the bulk material. We assume that the media is in conditions of plane strain under
the action of an external surface load and the additional surface stress. Thus, we
come to the formulation of the boundary value problem for the half-plane Ω = {z :
Im z < 0, Re z ∈ R1 }, z = x1 + i x3 ,with the rectilinear boundary Γ .
In general case, the boundary condition is described by generalized Young–
Laplace law [7]
n · Σ − ∇s · τ = p, (1)
where n is the unit vector normal to the boundary surface, Σ is the tensor of volume
stress, τ is the tensor of surface stress, p is the vector of an external surface load. Equa-
tion (1) means that the action of the surface stress is replaced by the corresponding
load ts (z) = ∇s τ defined by the surface gradient operator ∇s = ∇ − n∂/∂n, where
∇ is the Hamilton operator [2]

τ11 τ22 e1 ∂ ∂ ∂h 1 ∂h 2
ts (z) = − + n+ (h 2 τ11 )+ (h 1 τ21 )+ τ12 − τ22
R1 R2 h 1 h 2 ∂α1 ∂α2 ∂α2 ∂α1

e2 ∂h 1 ∂ ∂h 2 ∂
+ − τ11 + (h 2 τ12 ) + τ21 + (h 1 τ22 ) (2)
h1h2 ∂α2 ∂α1 ∂α1 ∂α2
Here e1 , e2 are the basis vectors of a curvilinear coordinate system α1 and α2 ;

h 1 , h 2 are the corresponding metric factors, R1 , R2 are the principal radii of the
coordinate lines curvature, τi j (i, j = 1, 2) are the components of the surface stress
tensor.
Let α1 = x1 , α2 = x2 in the plane surface x3 = 0. For the plane strain, we have
h 1 = h 2 = 1, 1/R1 = 1/R2 = 0. As τ12 = 0, τ22 = τ22 (x1 ), then according to
Eq. (2), the boundary condition (1) in complex variables takes the form
σ33 (z) − iσ13 (z) = −i p(x1 ) − its (x1 ), z ∈ Γ, (3)

Effect of a Type of Loading on Stresses at a Planar Boundary of a Nanomaterial 71
where p(x1 ) = p1 (x1 ) + i p3 (x1 ); p1 , p3 are the projections of the load vector p on
the Cartesian coordinate axes x1 , x3 ; ts = ∂τ11 /∂ x1 . Note that Eq. (3) can be directly
derived considering an equilibrium of an element of a boundary surface under applied
forces [5].
Generally, we assume that p(x) is the periodic function with the period a
x1+a/2
p(x1 ) = p(x1 + a), p(t)dt = P, P = P1 + i P3 , (4)

x1 −a/2
and satisfies the Hölder’s condition on whole Γ . The following conditions are
realized at infinity
lim (σ33 (z) − iσ13 (z)) = −i P/a, lim σ11 (z) = σ1 , lim ω(z) = ω∞ ,
x3 →−∞ x3 →−∞ x3 →−∞
(5)
where ω is the rotation angle of material particles; σi j are the stress components in
the x1 , x3 coordinate system.
The constitutive equations of linear elasticity for the surface [2, 7] and the bulk
material in the case of the plane strain are reduced to the following
τ11 = γ0 + (λs + 2μs − γ0 )ε11

s
, τ22 = γ0 + (λs + γ0 )ε11
s
, (6)
σ11 = (λ + 2μ)ε11 + λε33 , σ33 = (λ + 2μ)ε33 + λε11 ,

(7)
σ31 = 2με31 , σ22 = λ(ε33 + ε11 ),
where γ0 is the residual surface stress in an unstrained state; λs , μs are the moduli
of surface elasticity similar to the Lamé constants λ, μ for 3D elasticity; εi j are the
components of the strain in the bulk material; ε11 s is the component of the surface
strain.
3 Construction of Integral Equation
Proceeding from the volume Ω to the boundary Γ , we assume that the continuity
condition of displacements is satisfied
lim u j (z) = u sj (x1 ),

z→x1
where u sj is the displacement of points at the boundary Γ along the x j -axis ( j = 1, 3).
This equality yields the continuity condition for the strain ε11
lim ε11 = ε11

s
. (8)
z→x1
Relations (6) and (8) lead to the equation for finding the surface stress τ11
τ11 (x1 ) = γ0 + (λs + 2μs − γ0 )ε11 (x1 ), (9)
Thus, the problem is reduced by defining a stress-strain state of the half-plane with
the rectilinear boundary Γ on which the surface stress τ11 is acting. Equation (9)
connects the unknown surface stress τ11 with the strain ε11 arising under external
loading and conditions at infinity (5).
The expression for the strain ε11 can be found by solving the boundary problem
(3), (5). The stress vector σn = σnn + iσnt on the area with the normal n and the
displacement vector u = u 1 + iu 2 of the point z are connected with the complex
Goursat–Kolosov potentials by following formulas [4]

σn (z) = Φ(z) + Φ(z) − Φ (z) + Φ(z) − (z − z) Φ (z) e−2iα , (10)
du
− 2μ = −κΦ(z) + Φ(z) − Φ (z) + Φ(z) − (z − z) Φ (z) e−2iα , (11)
dz
where α is the angle between the area and the axis x1 ; κ = 3 − 4ν; ν is the Poisson’s
ratio. The derivative in Eq. (11) is taken along the area, i.e. in the direction of the
vector t which is perpendicular to basis vector n so that n and t define the right-hand
coordinate system.
The values of the function Φ at infinity follow from Eqs. (5), (10) and (11)
∞ + σ∞
σ11 2iμ ∞ σ1 P3
lim Φ(z) = 33
+ ω = + ,
x3 →−∞ 4 κ+1 4 4a
(12)
∞ ∞ σ1 P3 iP
lim Φ(z) = lim Φ(z) − (σ33 − iσ13 )= + + .
x3 →−∞ x3 →−∞ 4 4a a
Let z → x1 ∈ Γ and α = 0 in Eq. (10). Then subjecting to the boundary

conditions (3), we obtain Riemann–Hilbert’s jump problem
Φ + (x1 ) − Φ − (x1 ) = its (x1 ) + i p(x1 ), (13)
where Φ ± (x1 ) = lim Φ(z). According to [9], the solution of the problem (13)
Im z→±0
can be written as
∞ ∞
1 its (t) 1 i p(t)
Φ(z) − c± = dt + dt = Iτ (z) + I p (z). (14)
2πi z−t 2πi z−t
−∞ −∞
Here c± = lim Φ(z) ∓ i P/(2a), as lim Iτ (z) = 0 and lim I p (z) =

x3 →±∞ x3 →±∞ x3 →±∞
±i P/(2a).
Assuming α = π/2 in Eq. (10) and then α = 0 in Eq. (11), we obtain
σ11 (x1 ) + iσ13 (x1 ) = Φ − (x1 ) + 2Φ − (x1 ) + Φ + (x1 ), (15)
du
− 2μ = −κΦ − (x1 ) − Φ + (x1 ). (16)
d x1
Substituting Eq. (16) into (9) yields

τ11 = γ0 + MRe κΦ − + Φ + , (17)
λs + 2μs − γ0
where M = .
2μ
Using the Sokhotski–Plemelj formulas, one can show that (17) is the integro -
differential singular equation in surface stress τ11 . After differentiating Eq. (17) and
using Eq. (14), we get the equation of the unknown function ts

ts (x1 ) − MRe κ Iτ− (x1 ) + Iτ+ (x1 ) = MRe κ I p− (x1 ) + I p+ (x1 ) . (18)
In view of Eq. (14) and formulas Sokhotski-Plemelj, Eq. (18) yields
+∞ +∞
M(κ + 1) ts (t) M(κ − 1) M(κ + 1) p1 (t)
ts (x1 )− dt = p 3 (x1 )− dt.
2π (t − x1 ) 2 2 2π (t − x1 )2
−∞ −∞
(19)
This hypersingular integral equation is obtained without using periodicity conditions
of the function p(x1 ) and, therefore, it is valid for an arbitrary loading. In case of non-
periodic loadings, the function p should vanish at infinity and satisfy to conditions
of integral existence in the right hand side of Eq. (19) in sense of Hadamard’s finite
part [8].
It should be noticed that the homogeneous equation corresponding to Eq. (19) has
only zero solution. Otherwise, under the absence of external loadings, there would
be a surface stress τ11 differing from a constant that for an infinite plane surface
is unreal. Therefore, if a derivative of function ts satisfies to Hölder’s condition,
then Eq. (19) always has the unique solution for any continuous right hand side of
Eq. (19) [8].
4 Solution of Equation (19) in the Case of Periodic Loading
Find the solution of the integral Eq. (19) in the case of the action of a self-balanced
periodic loading at the boundary Γ , i.e. assume that P = 0 in Eq. (4). Consider a
special case when tangential load p1 is described by an odd function, and normal
load p3 – by an even one. Then function p can be expanded into the following Fourier
series
∞ ∞

p(x) = p1 (x) + i p3 (x) = Ck sin bk x + i Dk cos bk x, (20)
k=1 k=1
where
a/2 a/2
2 2
Ck = p1 (x) sin bk xd x, Dk = p3 (x) cos bk xd x, bk = 2π k/a.
a a
−a/2 −a/2
Hereinafter, we will denote x ≡ x1 instead of x1 .

We also calculate the function ts in the form of Fourier series
∞

ts (x) = Ak sin bk x + Bk cos bk x. (21)
k=1
It is possible to find unknown factors Ak , Bk from Eq. (19) by substituting the expres-
sions (20) and (21) into it and computing corresponding integrals. However, there
exists a more convenient method to find these factors. Using Eqs. (14), (20) and (21),
we derive the following expression for complex potential Φ
∞
∞
σ11 1 (Ck − Dk + Ak + i Bk )eibk z , Im z > 0,
Φ(z) = + (22)
4 2 (Ck + Dk + Ak − i Bk )e−ibk z , Im z < 0.
k=1
After substituting Eq. (22) into Eq. (18) and using definitions (14), we equate the
coefficients at the same harmonics that yields
π k (Ck (κ + 1) + Dk (κ − 1))
Ak = − , Bk = 0, k = 1, 2, . . . (23)
a/M + π k(κ + 1)
Thus, we have obtained analytical expressions for all coefficients in the Fourier series
(21) of function ts . In other words, we have got the exact solution of the integral
Eq. (19) in the form of Fourier series.
Integrating Eq. (21), we derive the expression for the surface stress
∞
Ak
τ11 (x) = − cos bk x + τ0 . (24)
bk
k=1
The constant τ0 can be found from Eq. (17). For this purpose, substitute Eqs. (22)
and (24) into Eq. (17). Then, assuming Ck = Dk = 0 (k = 1, 2, . . .) in the derived
equation, that corresponds to a free boundary of the half-plane, we obtain
M(1 + κ)
τ0 = γ0 + σ1 . (25)
4
The quantity τ0 is the surface stress corresponding to a homogeneous stress-strain
state of bulk material with a plane boundary. As one can see from Eq. (25), τ0 = γ0
for an unloaded body. So if σ1 = 0, and the boundary is free from the external
loading, there are no deformations in the bulk material and in the surface as well.
In the general case, surface stress τ0 depends on both γ0 and σ1 .
Substituting Eq. (22) into Eq. (15), we obtain expressions for longitudinal σ11 and
tangential σ13 stresses at Γ
∞

σ11 (x)|x3 =0 = 2Ck + Dk + 2 Ak cos bk x + σ1 ,
k=1 (26)
∞

σ13 (x)|x3 =0 = − Ck + Ak sin bk x.
k=1
5 Example
Let the external loads at the boundary of the half-plane be defined by one of the
following functions
p1 (x) = q1 Im f / max |Im f |, p3 (x) = q3 Re f / max |Re f |, (27)
where f (x) = − sh−2 (y + iπ x/a) ; q1 , q3 are the dimension factors equal to max-
imum of absolute values of corresponding loads; the parameter y defines a form of
corresponding curves. The plots of functions (27) for y = 0.5 are represented in
Fig. 1 by continuous lines. As it follows from Eq. (27), p1 (x) → q1 sin (2π x/a),
p3 (x) → −q3 cos (2π x/a), when y → ∞.
Approximating each of functions (27) by a piece of the corresponding Fourier
series (20) with prescribed accuracy ε, one can obtain the numerical solution of the
problem with the same accuracy. As a criterion of accuracy, we accept an integral
criterion. According to it, a ratio of the difference between the integral of function
p j and of its approximation over an interval of positive changing to the first one does
not exceed a given value ε.
The results of calculations show that the number of members of the corresponding
series required for an achievement of the prescribed accuracy depends on the value
of parameter y. This number increases if y decreases. In particular, when y = 0.5,
five members of the series are enough to approximate the function p1 (x1 ) with the
accuracy ε = 0.01.
For comparison, we also consider the action of tangential and normal loads defined
by the first member of Fourier series (20) and represent them by dashed lines in Fig. 1.
They also can be defined by the corresponding functions (27) if y → ∞
Fig. 1 Two types of tangential p1 and normal p3 external loads for y = 0.5 (continuous lines) and
y = ∞ (dotted) in (27)
p1 (x) = q1 sin b1 x, p3 (x) = −q3 cos b1 x. (28)
Using formulas (20), (23) and (26), we calculated stresses at the boundary for var-
ious values of geometrical parameters a and y with and without surface stress τ11 .
Loads are defined by functions (27) when y = 0.5 and functions (28). Besides, for
simplification of an analysis, it is assumed that σ1 = 0. The Poisson’s ratio ν = 0.3.
The plots of dependencies of the maximum absolute values of longitudinal and
tangential stresses on the period of loadings a are displayed in Fig. 2. Continuous
lines correspond to loadings (27), dotted - to (28). These dependencies illustrate the
so-called size effect noticed at the nanoscale in many works (see, for example, [1, 3,
6, 11]) if the surface stress is taken into account.
From Fig. 2 one can conclude that the most significant influence of the period
of loading a on stresses is in the limits of changing a approximately from 10 M to
300 M. For aluminum M = 0.113 nm [2], and this period is in the interval from 1 to
34 nm. If a > 1000 M, the size effect almost disappears and the stress-strain state of
a body does not depend on the surface stress.
Note that the size effect becomes more apparent for the tangential loads (curves 1)
than for the normal (curves 2). Besides, the maximum values of stress σ11 are less for
sinusoidal loadings (28) than for loadings (27). For the maximum values of stress σ13 ,
the behavior is opposite. The plots of stresses at the boundary of the half-plane are
Fig. 2 The maximum of the absolute value of the longitudinal stress s11 = max |σ11 | and tangential
stress s13 = max σ13 versus the period a for tangential loadings (curves 1) and normal loadings
(curves 2)
Fig. 3 The distribution of longitudinal σ11 and tangential σ13 stresses on the half-plane boundary
in the range of one period for tangential loadings and a = 10 M
Fig. 4 The distribution of longitudinal σ11 and tangential σ13 stresses on the half-plane boundary
in the range of one period for normal loadings and a = 10 M
represented in Figs. 3 and 4 by continuous and dashed lines calculated with and with-
out the surface stress respectively. Curves 1 corresponds to the action of loads (28),
curves 2—loads (27) if y = 0.5. The plots are constructed for a = 10 M. For
such value of a, the difference between the solution with the surface stress and the
traditional solution is clearly expressed.
It is apparent from the dependencies given in Figs. 3 and 4 that the existence of a
surface stress reduces the concentration of longitudinal and tangential stresses. As a
result, changing of these stresses becomes more smooth than it predicts the solution
in traditional statement. It is remarkable that, owing to the surface stress, normal
loadings cause tangential stresses at the boundary (Fig. 4). This fact follows directly
from boundary conditions (3) which mean that if the surface stress is not constant,
then the tangential stress will always be at the half-plane boundary independently of
the type of loading.
It is interesting to estimate an influence of the surface stress on the relative changes
of extremal values of stresses at the boundary for loadings defined by function (27)
with y = 0.5 and function (28). This influence can be seen from Table 1 in which
these extremal values corresponding to graphs in Figs. 3 and 4 are given.
Table 1 Extreme values of stresses at the boundary versus type of loading for a = 10, M = 1.13 nm
Type of loading Tangent loading p1 /q1 Normal loading p3 /q3
Type of function Func. Eq. (27) Func. Eq. (28) Func. Eq. (27) Func. Eq. (28)
max σ11 τ11 =0 2.855 2 0.064 1
τ11 = 0 1.120 1.064 0,047 0.733
min σ11 τ11 =0 −0.668 −2 −0.272 −1
τ11 = 0 −0.394 −1.064 −0.177 −0.733
max σ13 τ11 =0 0.882 1 0 0
τ11 = 0 0.401 0.532 0.034 0.134
6 Conclusion
The stress state of an elastic half-plane under the action of periodic surface forces at
the nanoscale is investigated. Based on the Gurtin–Murdoch model of surface elas-
ticity, complex potentials of Goursat–Kolosov, the Muskhelishvili representations
and boundary properties of analytical functions, the solution of the boundary value
problem in the general case of arbitrary loading at the boundary is reduced to the
hypersingular integral equation. The exact solution of this equation in the form of
Fourier series is obtained in the case of periodic loading. The numerical results for
aluminum show that the highest influence of the surface stress on the stress state of
the boundary takes place when the loading period does not exceed approximately
40 nm. As follows from the solution (23), this influence depends on elastic constants
of the surface and bulk material, by which the parameter M is expressed.
It is important to note one feature of the considered boundary problem. The
surface stress arises in a planar surface as a reaction on the changing the external
load along the surface. So far, the existence of the surface stress has been considered
in a curved boundary surface that is free from external loading (see, for example,
[2, 3, 5–7, 11]). In this sense, there is a steady opinion that the surface stress appears
only on a curvilinear surface, and the size effect is expressed in a dependence of
physical properties and stress-strain state of a body on the change of the surface
curvature.
In this work, the surface of the body has a zero curvature and a geometrical linear
dimension to which the size effect is related is the period of the load. Another feature
of the solution obtained is that the surface stress appears from the action of normal
loads and this leads to arising tangential stresses. As in the case with size effect,
these stresses become negligible when the period a is 100 nm. or more.
References
1. Altenbach, H., Eremeev, V.A., Morozov, N.F.: On equations of the linear theory of shells with
surface stresses taken into account. Mech. Solids 45, 331–342 (2010)
42, 1–68 (2009)
3. Goldstein, R.V., Gorodtsov, V.A., Ustinov, K.V.: Effect of residual stress and surface elasticity
on deformation of nanometer spherical inclusions in an elastic matrix. Phys. Mesomech. 13,
318–328 (2010)
4. Grekov, M.A.: A singular plane problem in the theory of elasticity. St. Petersburg University,
St. Petersburg (2001) (in Russ.)
5. Grekov, M.A., Kostyrko, C.A.: Instability of a flat surface of a film coating due to surface
diffusion. Vestnik St. Petersburg University, Ser. 10(1):46–54 (2007) (in Russ.)
6. Grekov, M.A., Morozov, N.F.: Surface effects and problems of nanomechanics. J. Ningbo Univ.
25, 60–63 (2012)
7. Gurtin, M.E., Murdoch, A.I.: A continuum theory of elastic material surfaces. Arch. Rational
Mech. Anal. 57, 146–147 (1975)
8. Linkov, A.M.: Boundary Integral Equations in Elastisity Theory. Kluwer, Dordrecht (2002)
9. Muskhelishvili, N.I.: Some Basic Problems of the Mathematical Theory of Elasticity. Noord-
hoff, Groningen (1963)
10. Podstrigach, Y.S., Povstenko, Y.Z.: Introduction to Mechanics of Surface Phenomena in
Deformable Solids. Naukova Dumka, Kiev (1985) (in Russ.)
11. Tian, L., Rajapakse, R.K.N.D.: Analytical solution for size-dependent elastic field of a
nanoscale circular inhomogeneity. Trans. ASME J. Appl. Mech. 74, 568–574 (2007)
Surface Stress in an Elastic Plane
with a Nearly Circular Hole
Mikhail A. Grekov and Anna A. Yazovskaya
Abstract A boundary value problem on a nanometer hole in an elastic plane under

arbitrary remote loading is solved. It is assumed that complementary surface stress is
acting at the boundary of the hole. The outer surface of the hole is supposed to be con-
formally mapped on the outer surface of the circle by means of a power function. The
Gurtin–Murdoch surface elasticity model is applied to take into account the surface
stress effect. Based on the Goursat–Kolosov complex potentials and Muskhelishvili’s
technique, the solution of the problem is reduced to a singular integro-differential
equation in an unknown surface stress. For a nearly circular hole, the boundary
perturbation method is used that leads to successive solutions of hypersingular inte-
gral equations. Numerical results based on the first-order approximate solution are
specified for an elliptical nearly circular hole.
1 Introduction
In traditional continuum mechanics, the effect of surface energy is ignored as it is

small compared to the bulk energy. At the same time, the surface effects become
significant for nanoscale materials and structures due to the high surface/volume
ratio [6, 19, 22]. In particular, the surface stresses are directly related to the
size effect, that means the material properties of a specimen depend on its size
[4–6, 19]. Besides the results of theoretical analysis and numerical calculations, the
size effect was observed in a number of experimental measurements of nanowires
and nanotubes [22].
M. A. Grekov (B)
Saint-Petersburg State University, Universitetski pr. 35 ,198504 St.-Petersburg, Russia
e-mail: magrekov@mail.ru
A. A. Yazovskaya
Saint-Petersburg State University, Universitetski pr. 28 ,198504 St.-Petersburg, Russia
e-mail: yazovskaya_aa@mail.ru

82 M. A. Grekov and A. A. Yazovskaya
Solid surface stresses are known like the prestress in a prestressed membrane
that is perfectly fitted to the bounding surface of a bulk material. The inclusion
of such a surface stress in an otherwise traction-free surface of the bulk material
leads to additional loads applied to this surface [7]. The general form of a boundary
condition with surface stress, namely, the generalized Young–Laplace equation, was
derived in [11, 17]. Surface effects have been considered in numerous theoretical
investigations based on the generalized Young-Laplace equation and the theory of
surface elasticity which had been developed in [11, 15]. In particular, the theory of
elasticity with surface stresses is applied to the modifications of the two-dimensional
theories of nanosized plates and shells in [1–3].
Various application of the Gurtin–Murdoch model [11, 15] in nanomechanics is
presented in the literature which is reviewed by Wang et al. in [22]. Using this model,
Tian and Rajapakse [19, 20] gave the solutions of two-dimensional size-dependent
elastic fields of a matrix with a nanoscale circular inhomogeneity [19] and elliptical
hole [20] under arbitrary remote loading. These solutions are fundamental in nanome-
chanics like the similar solutions in traditional solid mechanics. At the same time,
the shape of real holes and inhomogeneities differs from the circular or elliptical one.
Moreover, they can have relief surfaces and a circular/elliptical shape of them is an
example of an approximation. In this context, the aim of the present paper is to study
the surface stress effects for a nearly circular nanosized hole in an elastic plane under
remote loading. The outside of the hole can conformally be mapped on the outside
of the circle by means of a power function. Based on Goursat-Kolosov’s complex
potentials, in this paper Muskhelishvili’s approach [16] is extended to a nanoscale
hole for which the surface stresses give rise to nonclassical boundary conditions, i.e.
a generalized Young–Laplace equation. Application of the Gurtin-Murdoch surface
elasticity model and the mapping function leads to the singular integro differential
equation in an unknown surface stress. The solution of this equation by means of the
boundary perturbation method [8] for the case of a nearly circular hole leads to suc-
cessive solutions of hypersingular integral equations. The derivation of the solution
and numerical results for a special case of an elliptical hole are presented in the next
sections.
2 General Formulation of the Problem
Consider an infinite linear elastic body containing a single cylindrical hole of the
nanoscale, the surface of which has elastic properties differing from the same prop-
erties of the volume. This surface is represented as a thin film inseparably con-
nected with the bulk material. Plane strain conditions are assumed and the body
is subjected to remote loading and extra surface stresses at the boundary surface.
So, it leads to the 2D boundary value problem for the elastic plane of the com-
plex variable z = x1 + i x2 (i is the imaginary unit) with a single nanosized hole
(Fig. 1).
Surface Stress in an Elastic Plane with a Nearly Circular Hole 83
Fig. 1 Conformal mapping from z-plane with the pentagonal hole to ζ -plane with the circular hole
of the unit radius
It is supposed that the outside of the hole can be conformally transformed into the
exterior region of a unit circle (|ζ | > 1) by the mapping function υ:

N
z = υ(ζ ) = a1 ζ + a−k ζ −k , |ζ | ≥ 1, N ≥ 1. (1)
k=1
When N = 1 and a1 = (a + b)/2, a−1 = (a − b)/2, the hole has an elliptical shape
with axes 2a and 2b (a > b). In Fig. 1 the values N = 4, a1 = a0 , a−1 = a−2 =
a−3 = 0, a−4 = 0.1a0 are taken for Eq. (1).
It is important to note that ζ = ρeiθ in the polar coordinates ρ, θ at the plane
of the complex variable ζ . On the other hand, ρ = const and θ = const are two
sets of orthogonal coordinate curves at the plane of the complex variable z. So, the
boundary of the hole defined by the equality ρ = 1 is one of these coordinate curves.
In the general case of a three-dimensional body, the boundary conditions are
described by the Young–Laplace generalized law [4, 17]
n · S − ∇s · T = p, (2)
where n is the vector of the unit normal to the boundary surface, S is the volume stress
tensor, T is the surface stress tensor, and p is the vector of a given load. Equation (2)
means that the presence of surface stresses in a boundary surface of a solid yields
an action of the corresponding traction ts = ∇s · T defined by the nabla operator on
the surface ∇s which relates with the 3D gradient operator ∇ (3D nabla operator)
as ∇s = ∇ − n ∂/∂n. Here n is the coordinate along the normal to the surface. For
a curved surface with two orthogonal unit base vectors e1 and e2 , vector ts can be
expressed as follows [4]

τ 11 τ 22
ts = − + n
R1 R2

e1 ∂ ∂ ∂h 1 ∂h 2
+ (h 2 τ 11 ) + (h 1 τ 21 ) + τ 12 − τ 22
h 1 h 2 ∂α1 ∂α2 ∂α2 ∂α1

e2 ∂h 1 ∂ ∂h 2 ∂
+ − τ 11 + (h 2 τ 12 ) + τ 21 + (h 1 τ 22 ) . (3)
h1h2 ∂α2 ∂α1 ∂α1 ∂α2
Here, α1 and α2 are two parameters defining the surface that α1 = const and α2 =
const give two sets of mutually orthogonal curves on the surface, and h 1 and h 2
are the corresponding metric coefficients, R1 and R2 are the radii of the principal
curvatures. Quantities τ 11 , τ 22 , τ 12 are the components of the surface tensor T.
We introduce local Cartesian coordinates n, t in an arbitrary point z ∈ outside
of the hole. If z is a point of the hole boundary
= {z : α2 = const} free from
an external load, and the axis t is tangential to
, then using Eq. (3) and taking into
account the plane strain conditions, one can express the boundary conditions (2) in
the following complex variable form:
σtts 1 ∂σtts
σnn + iσnt = −i ≡ q(z), z ∈
, (4)
R1 h 1 ∂α1
where, σnn , σnt are the components of the volume stress tensor T, σtts = τ11 , and
if x1 = x1 (α1 ), x2 = x2 (α1 ) is the parametric form of representation of the hole
boundary
in Cartesian coordinates x1 , x2 , then
x1 x2 − x2 x1

h 21 (α1 ) = x12 + x22 , R1−1 (α1 ) = . (5)
h 31
Note that Eq. (4) can be directly derived considering an equilibrium of an element of
a boundary surface under applied forces [7].
The stresses σ jk ( j, k = 1, 2) in coordinates x1 , x2 and the rotation ω are specified
at infinity
lim σ jk (z) = s jk , lim ω(z) = 0 (6)
|z|→∞ |z|→∞
The linear elastic constitutive equations for the surface [4, 11] and bulk material in
the case of plane strain are reduced to the following
σtts = γs + (2μs + λs − γs )εtts , σ33

s
= γs + (λs + γs )εtts (7)
and
σtt = (2μ + λ)εtt + λεnn , σnn = (2μ + λ)εnn + λεtt ,

σnt = 2μεnt , σ33 = λ(εtt + εnn ). (8)
In Eqs. (7) and (8), εnn , εtt , εnt denote strains in the bulk material, λs and μs are the
moduli of the surface elasticity, similar to the Lamé constants λ and μ of the bulk
material, γs is the residual surface stress under unstrained conditions.
The non-separability conditions which state that the displacements of the film
coincide with the displacements of the bulk material at the boundary lead to the
equation for the hoop strains
εtts = εtt , z ∈
. (9)
3 Direct Solution of the Problem
3.1 Complex Potentials
Following Muskhelisvili’s technique [16], one can obtain the joint expression for
stresses and displacements
dz
G(z) = ηΦ(z) + Φ(z) + zΦ (z) + Ψ (z) , (10)
dz
where
σn (z), η = 1,
G(z) = du
−2μ , η = −κ,
dz
and Φ and Ψ are two holomorphic functions outside of the hole, σn = σnn + iσnt ,
u = u 1 + iu 2 , u 1 and u 2 are the displacement components in coordinates x1 , x2 ,
κ = (λ + 3μ)/(λ + μ).
In Eq. (10) and below, a quantity with the bar denotes complex conjugation and
the prime denotes the derivative with respect to the argument. The increment dz is
taken in the direction of the axis t. Thus in Eq. (10), dz = |dz|eiα , dz = dz, and α
is the angle between axes x1 and t.
Taking into account the mapping function υ, Eq. (10) is transformed to the fol-
lowing

υ (ζ ) dζ υ(ζ )
G(ζ ) = ηΦ(ζ ) + Φ(ζ ) + Φ (ζ ) + Ψ (ζ ) , |ζ | > 1. (11)
υ (ζ ) dζ υ (ζ )
Here, it is taken Φ(ζ ) = Φ(υ(ζ )), Φ(ζ ) = Φ(υ(ζ )), G(ζ ) = G(υ(ζ )), so the
functions Φ(ζ ) and Ψ (ζ ) are holomorphic in the region |ζ | > 1.
We introduce the new function Φ(ζ ) in the region |ζ | < 1 by means of the equality

υ (ζ )Φ(ζ ) = − υ (ζ )Φ ζ −1 + ζ −2 υ(ζ )Φ ζ −1 − ζ −2 υ ζ −1 Ψ ζ −1
(12)
By replacing ζ with ζ −1 in the equality (12) and after the complex conjugation, we
get

υ (ζ )Ψ (ζ ) = ζ −2 υ ζ −1 Φ(ζ ) + Φ ζ −1 − υ ζ −1 Φ (ζ ), |ζ | > 1.
(13)
Substituting Eq. (13) into (11), we obtain the basic expression for the stresses and
displacements
dζ

υ (ζ )G(ζ ) = υ (ζ ) ηΦ(ζ ) + Φ(ζ ) + ζ −2 υ ζ −1 Φ(ζ ) + Φ ζ −1
dζ

+ υ(ζ ) − υ ζ −1
Φ (ζ ) , |ζ | > 1. (14)
Let ζ → ζ0 = eiθ , dζ = |dζ |ei(θ+π/2) and η = 1 in Eq. (14). Then in view of the
boundary conditions (4), we come to the Riemann–Hilbert boundary problem with
regard to the function Υ (ζ ) = υ (ζ )Φ(ζ )
[Υ (ζ0 )]+ − [Υ (ζ0 )]− = −υ (ζ0 )q(ζ0 ), |ζ0 | = 1, (15)
where Υ ± (ζ0 ) = lim Υ (ζ ), q(ζ0 ) ≡ q(υ(ζ0 )).

|ζ |→1∓0
Analysis shows that for the mapping function (1), the right hand side of Eq. (12)
and consequently function Υ has a single pole of N + 1 order in the point ζ = 0.
So, the solution of the problem (15) can be written as
υ (ζ )Φ(ζ ) = −I (ζ ) + R(ζ ), |ζ |
= 1, (16)
where

N +1
1 υ (η)q(η)
I (ζ ) = dη, R(ζ ) = ck ζ −k . (17)
2πi η−ζ
|η|=1 k=0
Note that according to [16], the complex potentials Φ and Ψ have the following
representations at infinity
F 1 κF 1
Φ(z) = D1 − +o(z −1 ), Ψ (z) = D2 + +o(z −1 ), (18)
2π(κ + 1) z 2π(κ + 1) z
where F is the stress resultant at the boundary of the hole, lim z o(z −1 ) = 0 and
z→∞
4D1 = s11 + s22 , 2D2 = s22 − s11 + 2is12 .
In view of Eqs. (18), the form of the mapping function υ and F = 0, we obtain
from Eqs. (13)–(16) that
c0 = a1 D1 , c−1 = 0, c−2 = a1 D2 + a−1 D1 . (19)
The remaining coefficients of the function R(ζ ) can be found from the analyticity
condition of the function υ (ζ )Ψ (ζ ) at infinity.
3.2 Integral Equation
The integral in Eq. (17) contains the function q defined by Eq. (4) in terms of the
unknown surface stress σθθ s . To find this stress, we use Eq. (9) and constitutive rela-
tions (7) and (8) that yields
Ms
σθθ
s
= γs + (λ + 2μ)σθθ − λσρρ , z ∈
, (20)
4μ(λ + μ)
where Ms = 2μs + λs − γs .
Stresses σθθ and σρρ at the boundary of the hole
(i.e. at the circle |ζ | = 1) can
be obtained from Eq. (14) by taking η = 1, ζ → ζ0 = eiθ , and dζ = |dζ |eiθ and
dζ = |dζ |ei(θ+π/2) , respectively:
σθθ − iσρθ = Φ − (ζ0 ) + 2Φ − (ζ0 ) + Φ + (ζ0 ),

(21)
σρρ + iσρθ = Φ − (ζ0 ) − Φ + (ζ0 ).
Substituting σθθ and σρρ from (21) into (20) yields

σθθ
s
= γs + M Re κΦ − (ζ0 ) + Φ + (ζ0 ) , |ζ0 | = 1, (22)
where M = Ms /(2μ).
We will show that Eq. (22) is an integral equation. For simplicity, denote τ = σθθ
s .

As ∂τ/∂θ = iη∂τ/∂η = iητ (η) for η = e , then q(η) = τ (η)/R1 + ητ (η)/ h 1
iθ
and the Sokhotski–Plemelj formulas (see, for example, [16]) for Cauchy type integral
I (ζ ) in Eq. (17) take the form

± υ (ζ0 )q(ζ0 ) υ (η)q(η)
I (ζ0 ) = ± + dη. |ζ0 | = 1. (23)
2 η − ζ0
|η|=1
In view of Eq. (16), we introduce (23) into Eq. (22). As a result, we obtain the
following singular integro-differential equation in the unknown surface stress τ
⎧ ⎫
⎪
⎨ M(κ + 1) ⎪
⎬
−1 υ (η)q(η)
2 − M(κ − 1)R1 τ (ζ0 ) + Re
dη
⎪
⎩ πiυ (ζ0 ) η − ζ0 ⎪
⎭
|η|=1
R(ζ0 )
= 2γs + 2M(κ + 1)Re , |ζ0 | = 1. (24)
υ (ζ0 )

To derive Eq. (24) we used the evident equality Re ηh −1 1 (η)τ (η) = 0. Note that
the integral in Eqs. (23) and (24) is understood in the sense of the Cauchy principle
value [16].
There are some ways to solve Eq. (24). One of them is a direct solution by means
of one of the computing methods. But if a shape of the hole is close to the circular
one, we can apply the perturbation technique to the solution of this equation.
4 Solution of the Problem by the Boundary

Perturbation Method
4.1 General Approach
The boundary perturbation technique is widely used by a number of scientists for

the solution of different problems in continuum mechanics. For example, the recent
works in which we applied this technique are the papers [7, 8, 21].
Let ε be a maximal relative deviation of a hole boundary from the circle of the
radius a0 , and ε 1. In this case, we can represent the function τ as
∞
τm (ζ ) m
τ (ζ ) = ε . (25)
m!
m=0
Since functions υ, R1−1 and h −1 depend on the small parameter ε as well, the
right and left hand sides of Eq. (24) can also be expressed as power series in ε. After
equating the sum of polynomial coefficients of each power εm (m = 0, 1, . . . ) to
zero, Eq. (24) is reduced to the following sequence of hypersingular integral equations

M(κ + 1) η + ζ02 η−1 τm (η)
[2a0 − M(κ − 1)] τm (ζ0 ) + dη = f m (ζ0 ),
2πi (η − ζ0 )2
|η|=1
(26)
were |ζ0 | = 1, m = 0, 1, . . . ; f m (m > 0) is the function depending on all previous

solutions τ0 , τ1 , . . . , τm−1 , and
D1 D2
f 0 = 2a0 γs + Ma0 (κ + 1)Re R0 (ζ0 ), R0 (ζ0 ) = + 2. (27)
4 2ζ0
The integral in Eq. (26) is understood in the sense of a finite-part integral firstly
introduced by Hadamard in 1923 [12].
For m = 0, Eqs. (26) and (27) correspond to the problem of the circular hole
with the surface stress. In this case, it is evident that the integral Eq. (26) can not
have nontrivial solutions if f 0 = 0, i.e. if γs = 0 and without remote loading. So
according to [13], the inhomogeneous Eq. (26) has a unique solution.
The solution of Eq. (26) in the zero-order approximation (m = 0) derived in [10]
is defined by the formula
τ0 = d0 + d2 ζ02 + d2 ζ0−2 , (28)
where
a0
d0 = γs + a0 H1 D1 , d2 = a0 H2 D2 ,
a0 + M
(29)
M(1 + κ) M(1 + κ)
H1 = , H2 = .
a0 + M 2a0 + M(3 + κ)
4.2 First-Order Approximation for an Elliptical Hole
Consider an elliptical hole with the semi-axes
a = a0 (1 + ε), b = a0 (1 + kε), ε 1, |k| < 1, (30)
and find the first-order perturbation solution of the initial problem. First of all, sub-
stitute expressions x1 = a cos θ and x2 = b sin θ , defining the boundary of the hole,
into Eqs. (5). It yields

1 2 a 2 − b2
2 ab
h 21 = a + b2 − ζ0 + ζ0−2 , R1−1 = . (31)
2 2 h 31
Instead of searching for the function f 1 and solving Eq. (26) with this function,
insert following expressions of the first-order approximation
R1−1 = a0−1 + ε f (ζ0 ), h −1 −1

1 = a0 + εg(ζ0 ),
k+ k−
2 k+ k−
2
f (ζ0 ) = − + ζ0 + 3ζ0−2 , g(ζ0 ) = − + ζ0 + ζ0−2 ,
a0 2a0 a0 2a0
1 + k 1 − k b − a
k+ = , k− =
0
, k= ,
2 2 a − a0
Φ(ζ ) = Φ0 (ζ ) + εΦ1 (ζ ).
into Eqs. (16) and (17). As a result, the exact Eq. (16) is transformed to the following
approximate one

1 + k + ε − k − εζ −2 [Φ0 (ζ ) + εΦ1 (ζ )]

= −I0 (ζ ) − ε I1 (ζ ) + (1 + k + ε)D1 + (1 + k + ε)D2 + k − ε D1 ζ −2 . (32)
Here

t0 (η) t1 (η) + k + − k − η−2 t0 (η)
I0 (ζ ) = dη, I1 (ζ ) = dη, (33)
η−ζ η−ζ
|η|=1 |η|=1
and
1
t0 (η) = τ0 (η) + ητ0 (η) ,
a0
1
t1 (η) = f (η)τ0 (η) + ηg(η)τ0 (η) + τ1 (η) + ητ1 (η) . (34)
a0
Collecting members with ε0 and ε1 in Eq. (32), we obtain equations for complex
potentials Φ0 and Φ1 of the zero-order and first-order approximations respectively
Φ0 (ζ ) = −I0 (ζ ) + D1 + D2 ζ −2 , (35)

Φ1 (ζ ) = k − ζ −2 − k + Φ0 (ζ ) − I1 (ζ ) + k + D1 + k + D2 + k − D1 ζ −2 . (36)
Function Φ0 in (35) corresponds to the solution of the problem on a circular hole

[9, 10] and can be written by taking into account the expression for τ0 in Eq. (28)
and computing the integral I0 in (33).
To find τ1 and therefore function Φ1 , represent τ1 in the form of the power series
with the unknown coefficients
+∞

τ1 (ζ0 ) = dn1 ζ0n , |ζ0 | = 1. (37)
n=−∞
Substituting (36) into the equation


τ1 (ζ0 ) = M Re κΦ1− (ζ0 ) + Φ1+ (ζ0 ) , (38)
which arises from Eq. (22), and taking into account Eqs. (33)–(35) and (37), one can
get

2
τ1 (ζ0 ) = ζ0 ,
1 2n
d2n (39)
n=−2
where
M
d01 = k + d0 , d−2
1
= d21 , d−4
1
= d41 ,
a0 + M
2a0 + (3 + κ)M 1 1
d2 = (3 + κ)k + d2 + k − κd0 − 5d0 + 2(1 + κ)a0 k − D1 ,
M 2
2a0 + (5 + κ)M 1 1
d4 = (κ − 5)k d2 + (1 + κ)a0 k − D2 .
−
M 2
5 Numerical Results for an Elliptical Hole and Discussions
In this section, selected numerical results for the hoop stress σtt (σθθ in the mapped
plane) at the boundary of an elliptical hole with the ratio of the longer axis to the
shorter one a/b = 1.5 (Fig. 2) under plane strain are presented.
The values of the surface elastic parameters γs and Ms taken from theoretical cal-
culations for cubic metals [14, 18] are: γs = 1 N/m, Ms = 5.19 N/m. Besides, there
are used the bulk elastic constants for aluminum λ = 58.17 GPa, μ = 26.13 GPa.
In order to calculate σtt , the general Eq. (21) and the equations of the first-order
approximation (28), (29), (33)–(36), and (39) are applied. The effect of surface stress
is studied if γs = 0 and γs = 1 N/m. Comparison of these two cases allows to reveal
the influence of the residual surface stress γs and the surface stress Ms εtt arising due
to the action of the remote uniaxial tensions s22 and s11 (normal to the longer and
shorter axis of the ellipse respectively) on the hoop stress σtt . In the case γs = 0, the
Fig. 2 An example of the

boundary of an elliptical hole
with ratio a/b = 1.5 and the
reference circle of radius a0
in dimensionless coordinates.
Maximum deviation of the
ellipse from the circle is 0.3a0
Fig. 3 Stress σtt at θ = 0 vs. radius of reference circle a0 under uniaxial remote loading s22
(· · · ε = 0.2, − − ε = 0.3, − · − ε = 0.5, — exact solution under σtts = 0)
Fig. 4 Stress σtt at θ = π/2 vs. radius of reference circle a0 under uniaxial remote loading s11
(· · · ε = 0.2, − − ε = 0.3, − · − ε = 0.5, — exact solution under σtts = 0)
ratios σtt /s11 and σtt /s22 do not depend on the magnitude of the applied loading but
if γs = 1 N/m, they depend. An analysis shows that for a value of tension less than
1 GPa, the influence of γs on the stress distribution at the boundary and especially
on the stress concentration is comparable with the influence of the remote loads. To
highlight the effect of the residual surface stress γs , s22 = 0.1 GPa and s11 = 0.1 GPa
in the case γs = 1 N/m are taken (see Figs. 3, 4, 5).
Figures 3, 4, 5 show the dependence of σtt on the parameter ε but the discrepancy
is in the limits of the first-order approximation and not higher than 20–30 %. It is
important to note that deviations of the approximate solution for a0 > 100 nm from
the exact classical one does not exceeds 5 % (Fig. 3) and 14 % (Fig. 5).
Fig. 5 Variation of stress σtt along boundaries of the circular hole of radius a0 (— σtts = 0, — —
σtts
= 0) and the elliptical hole (· · · ε = 0.2, − − ε = 0.3, − · − ε = 0.5) under uniaxial loading
s22 = 0.1 GPa or s11 = 0.1 GPa (a0 = 5 nm, γs = 1 N/m)
The main effect induced by the surface stress is the size-dependence of stress σtt
especially if a0 < 30 nm. This phenomenon being readily apparent from Figs. 2 and
3 has been discovered in many studies (see review [22]). The curve designated by
γs = 0 and ε = 0.5 in Fig. 2 coincides practically with the same curve in [20].
The influence of residual surface stress γs on the stresses and strains in nanosized
materials recently ascertained in [6] is less examined. It can be seen from Fig. 3 and
Fig. 4 that the surface stress not only decreases the maximum values of σtt but, for
a very small hole (a0 < 3 nm), causes σtt to be negative all over at the boundary
in spite of tension at infinity, if γs is taken into account. The reason of this effect is
that the residual surface stress produces negative hoop stresses on the surface of an
unloaded nanomaterial. It follows for example from the exact solution of the problem
on a nanosized circular hole in an infinite plane [10]. As it is clear from this solution,
the hoop stress is negative everywhere at the boundary if remote loading does not
exceed some particular value.
Distribution of the stress σtt at the boundary of the circular hole of radius
a0 = 5 nm in classical solution and in view of surface stress σtts , and at the boundary
of the elliptical hole is represented in Fig. 5. As it is shown in Fig. 5, the presence of
the surface stress takes a third of the stress concentration and triples maximal com-
pressive stress in the case of the circular hole and chosen values of applied loading
and γs . The similar effect can be seen in Fig. 5 with regard to the elliptical hole.
Acknowledgments The work was supported by the Russian Foundation for Basic Research (grant
11-01-00230) and St.-Petersburg State University (project 9.37.129.2011).
References
1. Altenbach, H., Eremeyev, V.A., Morozov, N.F.: Linear theory of shells taking into account
surface stresses. Doklady Phys. 54, 531–535 (2009)
3. Altenbach, H., Eremeyev, V.A.: On the shell theory on the nanoscale with surface stresses. Int.
J. Eng. Sci. 49, 1294–1301 (2011)
42, 1–68 (2009)
5. Eremeyev, V.A., Morozov, N.F.: The effective stiffness of a nanoporous rod. Doklady Phys.
55, 279–282 (2010)
6. Goldstein, R.V., Gorodtsov, V.A., Ustinov, K.V.: Effect of residual stress and surface elasticity
on deformation of nanometer spherical inclusions in an elastic matrix. Phys. Mesomech. 13,
318–328 (2010)
7. Grekov, M.A., Kostyrko, C.A.: Instability of a flat surface of a film coating due to surface
diffusion. Vestnik St. Petersburg University, Ser. 10(1), 46–54 (2007) (Russian).
8. Grekov, M.A., Makarov, S.N.: Stress concentration near a slightly curved part of an elastic
body surface. Mech. Solids 39(6), 40–46 (2004)
9. Grekov, M.A., Morozov, N.F.: Solution of the Kirsch problem in view of surface stresses. Mem.
Differ. Equ. Math. Phys. 52, 123–129 (2011)
10. Grekov, M.A., Morozov, N.F.: Surface effects and problems of nanomechanics. J. Ningbo Univ.
25(1), 60–63 (2012)
11. Gurtin, M.E., Murdoch, A.I.: A continuum theory of elastic material surfaces. Arch. Rational
Mech. Anal. 57, 146–147 (1975)
12. Kaya, A.C., Erdogan, F.: On the solution of integral equations with strongly singular kernals.
Quart. Appl. Math. 45(1), 105–122 (1987)
13. Linkov, A.M.: Boundary Integral Equations in Elastisity Theory. Kluwer Academic Publishers,
Dordrecht (2002)
14. Miller, R.E., Shenoy, V.B.: Size-dependent elastic properties of nanosized structural elements.
Nanotechnology 11, 139–147 (2000)
15. Murdoch, A.I.: A thermodynamic theory of elastic material interface. Q. J. Mech. Appl. Math.
29, 245–275 (1976)
16. Muskhelishvili, N.I.: Some Basic Problems of the Mathematical Theory of Elasticity. Noord-
hoff, Groningen (1963)
17. Povstenko, Y. Z.: Theoretical investigation of phenomena caused by heterogeneous surface
tension in solids. J. Mech. Phys. Solids 41, 1499–1514 (1993)
18. Shenoy, V.B.: Atomic calculations of elastic properties of metalic fcc crystal surfaces. Phys.
Rev. B 71, 94–104 (2005)
19. Tian, L., Rajapakse, R.K.N.D.: Analytical solution for size-dependent elastic field of a
nanoscale circular inhomogeneity. Trans. ASME J. Appl. Mech. 74, 568–574 (2007)
20. Tian, L., Rajapakse, R.K.N.D.: Elastic field of an isotropic matrix with nanoscale elliptical
inhomogeneity. Int. J. Solids Struct. 44, 7988–8005 (2007)
21. Vikulina, Y., Grekov, M.A., Kostyrko, S.A.: Model of film coating with weakly curved surface.
Mech. Solids. 45(6), 778–788 (2010)
22. Wang, J., Huang, Z., Duan, H., et al.: Surface stress effect in mechanics of nanostructured
materials. Acta Mechanica Solida Sinica 24, 52–82 (2011)
Glass Spheres: Functionalization, Surface
Modification and Mechanical Properties
Zinaida Kutelova, Hendrik Mainka, Katja Mader, Werner Hintz

and Jürgen Tomas
Abstract Hydrophobic micro-glass particles were obtained by chemical

modification with organosilanes. Particles were treated by peroxymonosulfuric acid
to obtain a hydrophilic surface, which was the first step of the modification process.
Different silanes were used, each of them with a different functional group, effecting
variable degrees of hydrophobicity. The successful chemical modification process
was established using Fourier transform infrared spectroscopy (FTIR) and water drop
interaction with the modified particles surfaces. The morphology of the modified
particles was studied using scanning electron microscope (SEM). The degree of the
hydrophobicity was established with static contact angle measurements. The micro-
scopic adhesion and particle contact properties of the comparatively stiff (amorphous)
micro-glass beads and the macroscopic powder flow behavior were investigated with
ring shear tests and evaluated by constitutive models on physical basis.
Z. Kutelova (B) · H. Mainka · K. Mader · W. Hintz · J. Tomas

Mechanical Process Engineering, Department of Process Engineering,
Otto-von-Guericke-University of Magdeburg, Magdeburg, Germany
e-mail: zinaida.kutelova@ovgu.de
H. Mainka
e-mail: hendrik.mainka@ovgu.de
K. Mader
e-mail: katja.mader@ovgu.de
W. Hintz
e-mail: werner.hintz@ovgu.de
J. Tomas
e-mail: juergen.tomas@ovgu.de

96 Z. Kutelova et al.
1 Introduction
A hydrophobic surface is advantageous in several applications, such as car wind-

screens and self-cleaning glass surfaces [1]. In aqueous medium in dependence on
the pH value the unmodified glass particles have polar functional hydroxyl-groups
Si-OH or charged Si-O− or Si-OH+ 2 groups on their surfaces. These groups can
interact with water molecules and thus the surface is wettable. The organo silanes
are substances, commonly used for modification of glass or silica surfaces, providing
them with various properties [2]. Their typical applications are as: coupling agents,
adhesion promoters, silicate stabilizers, crosslinking agents, hydrophobing and dis-
persing agents. Similar to the process of Hegde, Hirashima and Rao [3] for silica
particles, it is possible to obtain hydrophobic glass particles using different silanes.
The silanization process has been used for decades for the surface hydrophobization
of silicate and other materials [4], and thus, could also be used for nano and micro
glass particles. There is a wide range of commercially available silanes for silaniza-
tion [5]. Silanes are silicon chemicals that possess a hydrolytically active center that
can react with inorganic substrates such as glass and form stable covalent bonds.
They also possess an organic substitution that modifies the physical interactions of
treated substrates [5]. In recent years fluorinated silanes are of great interest in the
surface modification processes. The fluoroalkyl modified glass surfaces become not
only hydrophobic but lipophobic as well [6]. The hydroxyl groups, presented on the
particle surfaces, are the sites at which silanes are adsorbed and eventually react [7].
Thus it is important for the particles to be subjected to reasonable pretreatment
process in which the glass particles surface has to be hydroxylated [8].
2 Experimental Setup
For fresh glass particles in the modification process a commercially available prod-
uct Spheriglass 5000 CP00 purchased from Potters Europe GmbH was used with
the following particle size distribution: d10 % = 3.8 µm, d50 % = 10.9 µm, d90 % =
21.7 µm. The chemical composition of the glass-particles is as follows: 72.5 % sil-
icon dioxide, 13.7 % sodium oxide, 9.8 % calcium oxide, 3.3 % magnesium oxide,
0.4 % aluminum oxide, 0.2 % iron oxide and 0.1 % potassium oxide. The raw mate-
rial consist of spherical, smooth, non-porous glass particles with a specific surface
area of 0.3 m2 /g. To ensure that the surface is free of contaminations, the parti-
cles were cleaned and activated by 24-h treatment with peroxymonosulfuric acid
(Fig. 1), which is a mixture of concentrated (95–97 %) sulfuric acid (H2 SO4 ) and
concentrated (50 %) hydrogen peroxide (H2 O2 ) in ratio 3:1. It is a strong oxidiz-
ing agent that removes most of the organic matter from the particles surface and
hydroxylates (adds OH-Groups) making them extremely hydrophilic [9]. After that
the particles were washed with deionized water and dried in an oven. After this
treatment the particles are ready for the surface modification process. The follow-
ing silanes were used for the modification: Chlorodimethylphenylsilane, purchased
Glass Spheres: Functionalization, Surface Modification and Mechanical Properties 97
Fig. 1 Preparation of the particles for the modification process
Fig. 2 Modification of the glass particle surfaces by example of Chlorodimethylphenylsilane
from Alfa Aesar GmbH, 3,3,3-Trifluoropropyltrimethoxysilane and 1H,1H,2H,2H-

Perfluorooctyltriethoxysilane both purchased from ABCR GmbH. The amount of the
modifying agent was chosen to be sufficient for building of a silane-monolayer on
the particles surface and for the surface to be effectively shielded. The reaction pro-
ceeds as shown in Fig. 2. Methanol was used as a solvent. The suspension was stirred
with a magnetic stirrer for one hour and afterwards heated slowly up to 65 ◦ C. After
evaporating the most part of the solvent, the residue was dried overnight in drying
oven at temperature of 70 ◦ C. By the silanization process agglomerates are formed.
In comparison to the unmodified beads, which were non adhesive, the hydrophobic
beads were formed as clumps.
3 Results and Discussion
The successful particles modification was established at first with an empirical test
of the hydrophobicity and then confirmed with Fourier transform infrared (FTIR)
spectroscopic studies. The degree of hydrophobicity was determined by static con-
tact angle measurements. Ring shear tests (Schulze Ring shear apparatus RST-XS)
have been made to investigate how the particles surface modification affects their
macroscopic powder consolidation behavior and their flowability. The results from
the ring shear tests were used to calculate their micromechanical contact and adhesion
properties as well as macroscopic powder flow behavior [18].
3.1 Wettability of the Modified Particles
Right after the silanization process the particles were subjected to an empirical
hydrophobicity test to establish the modification success. The wettability of the
Fig. 3 Interaction between a water droplet and particles modified with: a unmodified parti-
cles, b Chlorodimethylphenylsilane, c 3,3,3-Trifluoropropyltrimethoxysilane and d 1H,1H,2H,2H-
Perfluorooctyltriethoxysilane
particles was tested with water droplet, colored with methylene blue (Fig. 3). The
same test was conducted with the non-modified particles. The results show a suc-
cessful surface modification.
3.2 Fourier Transform Infrared (FTIR) Studies
Figure 4 shows the FTIR spectra of the silanized glass particles as a function of
wave number. The peaks observable at around 2950 and 1400 cm−1 are due to the
C-H bonds [10, 11] and the band at around 850 cm−1 occurs due to Si-C bond [10].
The peaks visible at 3400 and 1600 cm−1 are caused by OH groups. The strong
absorption band at 1100 cm−1 is caused by siloxane bonds Si-O-Si [11, 12]. The
FTIR spectra clearly depict the replacement of surface H from the Si-OH groups by
Fig. 4 Fourier transform infrared absorption spectra of the modified particles

Fig. 5 Static contact angle measurements of microscope glass slides modified with:
a Chlorodimethylphenylsilane, b 3,3,3-Trifluoropropyltrimethoxysilane, c 1H,1H,2H,2H-
the non-hydrolysable silane groups and therefore confirm the surface modification.
The spectra also reveal that not all of the hydrogen atoms have been replaced by the
silane groups.
3.3 Contact Angle and Layer Thickness
The degree of hydrophobicity was determined by measuring the contact angles.

Planar slides with the same composition as the glass particles were modified with
the studied silanes and measurement of the static contact angles was conducted. The
results showed that the surface with the best hydrophobicity, achieved at contact angle
of 114◦ (Fig. 5a), was modified with 1H,1H,2H,2H-Perfluorooctyltriethoxysilane,
followed by 3,3,3-Trifluoropropyltrimethoxysilane with 71◦ (Fig. 5b). The smallest
contact angle of 53◦ was measured after the modification with Chlorodimethylphe-
nylsilane (Fig. 5c). The contact angle of unmodified slides, measured as reference,
was 32◦ and slides, cleaned with peroxymonosulfuric acid, showed a contact angle of
nearly 15◦ . The modified glass slides were also subjected to surface layer thickness
measurement. The results showed that theoretically the layer thickness of the par-
ticles, modified with Chlorodimethylphenylsilane, 3,3,3-Trifluoropropyltrimetho-
xysilane and 1H,1H,2H,2H-Perfluorooctyltriethoxysilane, would be respectively 10,
13 and 20 nm, if an ideal surface monolayer is achieved.
3.4 SEM-Investigation
To characterize the samples scanning electron microscopy was performed (Fig. 6).
The aim was to detect optically any change in the surface structure after the mod-
ification process. Since theoretically the layer thickness is in the range of a few
nanometers, the scanning electron microscope would not be able to reveal its struc-
ture due to the low maximum resolution. But nevertheless on the particles, modified
with Chlorodimethylphenylsilane, was observed a much rougher surface. On the
Fig. 6 SEM-pictures of a Unmodified particles and particles modified with: b Chlo-

rodimethylphenylsilane, c 3,3,3-Trifluoropropyltrimethoxysilane and d 1H,1H,2H,2H-
other hand the particles modified with 1H,1H,2H,2H-Perfluorooctyltriethoxysilane

had some irregularities on their surfaces, which are clearly the places where the long
chain modifying agents were attached to the surface. The SEM-pictures also showed
a changed particles dispersity caused by the modification process. In case of some
surface modifications is caused an agglomerate formation due to the increased adhe-
sion forces. A solid bridge formation at the particles contact points is observed by
the particles modified with 1H,1H,2H,2H-Perfluorooctyltriethoxysilane.
3.5 Macro and Micro Mechanical Properties
Shear experiments are widely performed to determine the various macro-mechanical

properties of cohesive powders [13]. In order to assess the flow behavior of the
modified micro-glass particles a Schulze ring shear tester [14] was used. The stresses,
which lead to flow or irreversible plastic deformation or breakage, are measured [15].
This stress limit is known as yield locus and it can not be crossed. The shear test
procedure implies tree steps: pre-consolidation, preshear and shear [16]. At the first
step the specimen is stressed applying a pre-consolidation normal stress σ , afterwards
it is sheared until reaching a steady state flow (flow under a constant shear stress
and constant bulk density at the applied normal force). Further the measurement
proceeds under reduced shear stress (the last measurement step) until incipient flow,
characterized with failure of the bulk solid or powder, is attained. If one connects the
shear and normal stress points that will cause incipient failure of the bulk powder,
pre-consolidated to a major principal stress, with the stress point at which steady state
flow is attained, the result would be the yield locus [17]. Beside the flow function f f c ,
which numerically describes the flow behavior, other important characteristics, that
could be derived from the shear tests are: the internal friction between the particles
at incipient flow (the slope of the yield locus), stationary angle of internal friction,
which characterizes the cohesive steady-state flow ϕst , the effective angle of internal
friction ϕe associated with cohesionless steady-state flow, the isostatic tensile strength
of the unconsolidated particle without any contact deformation σ0 , which equals a
cohesion force between particles in an unconsolidated powder [18]. By measuring
Table 1 Classification of the

Flow function ffc Description
flowability of bulk solids
according to Jenike[19] and 10 < ff c Free flowing
Tomas[13] 4 < ff c < 10 Easy flowing
2 < ff c < 4 Cohesive
1 < ff c < 2 Very cohesive
ff c < 1 Hardened
the specimen weight and its volume at steady state flow (maximum consolidation) the
bulk density ρb of the bulk powder sample could also be calculated. The classification
of the powder flow behaviors according to Jenike’s flow function ff c = σ1 /σc [19]
is shown in Table 1.
The consolidation function correlates characteristically the uniaxial compressive
strength σc and the major principal stress at stationary flow, σ1 [13]. According to this
function the free flowing powders have a flow function larger than 10 and the non-
flowing less than 1 [20]. The macroscopic consolidation function of the modified and
unmodified particles is shown on Fig. 7. It is obvious that the modification has caused
a decrease of the flowability of the modified particles. Only the flow behavior of
the particles modified with 3,3,3-Trifluoropropyltrimethoxysilane resembles at some
point that of the unmodified particles. The application of the external consolidation
Fig. 7 Macroscopic consolidation function of treated and untreated glass particles measured at
ambient conditions of temperature and humidity
stress σ1 (normal force in the contact) however may result as mutual penetration
and interlocking of the surface modification agent chains of the respective contact
partners and consequently cause a remarkable change in powder flow behavior. On
the other hand the decrease of the flowability of the cleaned (hydrophilic) particles
is caused by the increased tendency for water absorption resulting from the hydroxyl
groups on the particle surfaces. A reverse micro-macro transition is made using the
shear test results to calculate the micromechanical contact and adhesion properties
of the glass particles shown in [18]. The adhesion force is calculated as a linearized
function of contact normal force F N , the adhesion force without any contact flattering
F H 0 and the dimensionless elastic-plastic contact consolidation coefficient κ [20].
This coefficient results from the slope of the adhesion force F H and characterizes
the level of contact flattering and resulting increase of adhesion. Thus a low slope
means low adhesion level, resulting from a stiff particle contact. But a large slope
means comparatively large contact flattering e.g. the contact is soft. The soft contact
is distinctive for a cohesive powder flow behavior [21]. The particles modified with
Chlorodimethylphenylsilane show the largest adhesion force F H increasement with
increasing normal force F N (Fig. 8) and therefore the softest particle contact (Fig. 9).
The surface modification with 3,3,3-Trifluoropropyltrimethoxysilane caused a
decrease of the adhesion force F H 0 of the unconsolidated particles in comparison
to the unmodified glass particles. Particles modified with the other silanes show
the opposite tendency, which is probably due to the characteristics of the sur-
face chains. By reducing the adhesion force due to surface treatment with 3,3,3-
Trifluoropropyltrimethoxysilane, the rearrangement of micro-glass particles into a
denser packing is obviously encouraged and thus an increased bulk density was
established. The particles treated with Chlorodimethylphenylsilane, on the other
hand, have the highest adhesion force and respectively the lowest bulk density.
Fig. 8 Microscopic adhesion

function of the studied micro-
glass particles
Fig. 9 Attractive adhesion force (F H 0 ) and elastic-plastic consolidation coefficient (κ) calculated
for the unmodified and the modified particles
4 Conclusion
Hydrophobic surfaces on the micro-glass particles were successfully generate using

a wet chemical modification e.g. the silanization process. Three different silane cou-
pling agents were used, each of them provided the particles with different hydropho-
bicity degree. By using this modification technique the particles were not uniformly
modified, as can be seen on the SEM-pictures. An increased tendency for agglom-
eration of the modified particles was observed. The investigation of the powder
behavior of the coated glass particles showed a decreased flowability in comparison
to the unmodified particles and increased adhesion forces. FTIR spectroscopy was
used to prove the success of the modification process.
References
1. Rajala, M.: Hydrophobic glass surface. US 2009/0095021 A1, 16 April 2009

2. Owen, M.J., Williams, D.E.: Surface modification by fluoroalkyl-functional silanes: a review.
J. Adhesion Sci. Technol. 5, 307–320 (1991)
3. Hegde, N.D., Hirashima, H., Rao, A.V.: Two step sol-gel processing of TEOS based hydropho-
bic silica aerogels using trimethylethoxysilane as a co-precursor. J. Porous Mater. 14, 165–171
(2007). doi:10.1007/s10934-006-9021-2
4. Plueddemann, E.P.: Silane Coupling Agents. Plenum Press, New York (1991)
5. Arkles, B.: Hydrophobicity and Hydrophilicity and Silane Surface Modification. Gelest Inc,
Morrisville (2011)
6. Yoshino, N., Sato, T., Miyao, K., et al.: Synthesis of novel highly heat-resistant fluorinated
silane coupling agents. J. Fluor. Chem. 127, 1058–1065 (2006). doi:10.1016/j.jfluchem.2006.
04.017
7. Liu, X.M., Thomason, J.L., Jones, F.R.: The concentration of hydroxyl groups on glass surfaces
and their effect on the structure of silane deposits. In: Silanes and Other Coupling Agents.
Koninklijke Brill NV, Leiden (2009)
8. Cras, J.J., Rowe-Taitt, C.A., Nivens, D.A., et al.: Comparison of chemical cleaning methods of
glass in preparation for silanization. Biosens. Bioelectron. 14, 683–688 (1999). doi:10.1016/
S0956-5663(99)00043-3
9. Jradi, K., Daneault, C., Chabot, B.: Chemical surface modification of glass beads for the
treatment of paper machine process waters. Thin solid films 519, 4239–4245 (2011). doi:10.
1016/j.tsf.2011.02.080
10. Hering, N., Schreiber, K., Riedel, R., et al.: Synthesis of polymeric precursors for the formation
of nanocrystalline Ti-C-N/amorphous Si-C-N composites. Appl. Organomet. Chem. 15, 879–
886 (2001). doi:10.1002/aoc.241
11. Hedge, N.D., Rao, A.V.: Organic modification of TEOS based silica aerogels using hexade-
cyltrimethoxysilane as a hydrophobic reagent. Appl. Surf. Sci. 253, 1556–1572 (2006)
12. Jeong, A.Y., Goo, S.M., Kim, D.P.: Characterization of hydrophobic SiO2 powders prepared
by surface modification on wet gel. J. Sol-Gel Sci. Technol. 19, 483–487 (2000)
13. Tomas, J., Kleinschmidt, S.: Improvement of flowability of fine cohesive powders by flow
additives. Chem. Eng. Technol. 81, 1470–1483 (2009). doi:10.1002/ceat.200900173
14. Schulze, D.: Pulver und Schüttgüter, Fließeigenschaften und Handhabung. Springer, Berlin
(2006)
15. Hintz, W., Antonyuk, S., Schubert, W., et al.: Determination of physical properties of fine
particles, nanoparticles and particle beds. In: Modern Drying Technology: Experimental Tech-
niques, vol. 2. (2009). doi:10.1002/9783527631643.ch6
16. Haack, A., Tomas, J.: Untersuchungen zum Dämpfungsverhalten hochdisperser kohäsiver Pul-
ver. Chem. Ing. Technik 75, 1646–1649 (2003)
17. Rabinovich, Y., Esayanur, M., Johanson, K., et al.: The flow behavior of the liquid/powder
mixture, theory and experiment. I. The effect of the capillary force (bridging rupture). Powder
Technol. 204, 173–179 (2010)
18. Tomas, J.: Product design of cohesive powders mechanical properties, compression and flow
behavior. Chem. Eng. Technol. 27, 605–618 (2004). doi:10.1002/ceat.200406134
19. Jenike, A.W.: Storage and flow of solids. Eng Exp Station, University of Utah Bull No.123
(1964)
20. Tomas, J.: Fundamentals of cohesive powder consolidation and flow. Granul. Matter 6, 75–86
(2004). doi:10.1007/s10035-004-0167-9
21. Tomas, J.: Adhesion of ultrafine particles-a micromechanical approach. Chem. Eng. Sci. 62,
1997–2010 (2007)
Spectral Properties of Piezoelectric Bodies
with Surface Effects
Andrey V. Nasedkin and Victor A. Eremeyev
Abstract We consider the problems of natural oscillations of nanosize piezoelectric

bodies taking into account surface stresses and surface electric charges. The spec-
tral properties of the boundary-value problems are determined by the combination
of approaches developed earlier for piezoelectric bodies and for elastic bodies with
surface stresses. We formulate theorems on the changes of the natural frequencies
under the changes of boundary conditions and material characteristics. We also dis-
cuss finite element approaches for determination of the natural frequencies, the res-
onance and antiresonance frequencies of nanosize piezoelectric bodies. The paper
provides the results of finite element computations of the model problems that illus-
trate some of the observed trends for the frequency changes.
1 Introduction
As it is known, nanomaterials has abnormal mechanical properties which differ con-

siderably from conventional macromaterials. One of the factors that are responsible
of the behavior of nanomaterials can be surface effects. As recent investigations
(see, for example, [1–3]) show, for the bodies of submicro- and nanosize the surface
stresses play an important role and influence the deformation of the bodies in gen-
eral. Similar to the elastic bodies, when analyzing the piezoelectric nanosize media
one can introduce surface stresses and distributed electric charges into the model by
A. V. Nasedkin (B)
Southern Federal University, Miltchakova str., 8a, Rostov on Don 344090, Russia
e-mail: nasedkin@math.sfedu.ru
V. A. Eremeyev
Institut für Mechanik, Fakultät für Maschinenbau, Otto-von-Guericke-Universität Magdeburg,
Universitätsplatz 2, Magdeburg 39106, Germany
V. A. Eremeyev
South Scientific Center of RASci and South Federal University, Rostov on Don, Russia

106 A. V. Nasedkin and V. A. Eremeyev
adding to the surface the corresponding elastic membranes and dielectric films. This
approach is used in the present work for investigation of the natural oscillations of
piezoelectric nanosize bodies.
2 Problem Setting for Natural Oscillation of Piezoelectric

Bodies with Surface Effects
Let V be a bounded in IR3 region, occupied by the piezoelectric body; S = ∂ V is

the boundary of the region, n is the vector of the external unit normal to S. Let us
consider that the region V and its boundary S satisfy the following conditions: V is
the sum of the finite number of subsets that are star-like relatively to arbitrary balls
contained in them, S is the Lipschitz boundary of C1 class. In more details these
conditions (V, S) can be found in [4].
Restricting ourselves to stationary oscillations regimes with the circular frequency
ω, we will use only amplitude values of all physico-mechanical variable hereinafter
without special reference. Let us consider the vector of mechanical displacements
u = u(x) and the electric potential ϕ = ϕ(x) as the main variables for the piezo-
electric medium. Using these functions, we can define the second-order strain tensor
ε = ε (u) and the electric field vector E = E(ϕ)
ε = (∇u + (∇u)T )/2, E = −∇ϕ, (1)
where by (. . .)T we define the transpose operation.

In the linear approximation we use the following standard constitutive relations
for the piezoelectric medium
σ = c · ·εε − eT · E, (2)
D = e · ·εε + d · E, (3)
where σ is the second-order stress tensor; D is the electric displacement vector;

c = c E is the forth-order tensor of elastic moduli, measured at constant electric
field; e is the third-order tensor of piezomoduli; d = d S is the second-order tensor of
dielectric permittivities, measured at constant strain. Here the conventional for the
piezoelectricity theory superscripts at c E and d S are omitted for the compactness of
further notations.
For the homogeneous problem of natural oscillation of a piezoelectric body we
have the following field equations in quasielectrostatic approximations
−∇ · σ = ρ ω2 u, (4)
∇ · D = 0, (5)
Spectral Properties of Piezoelectric Bodies with Surface Effects 107
where ρ is the material density. The density ρ(x) is assumed to be piecewise-

continuous and ∃ρ0 > 0 : ρ(x) ≥ ρ0 . The material moduli of the medium in
(2), (3) are piecewise-continuous together with their first derivatives by x with usual
symmetry conditions (ci jkl = c jikl = ckli j , eikl = eilk , εkl = εlk ), and for the
positive definite volume density of internal energy W (εε , E) the following inequality
should satisfy ∀εε = ε T , E
1 T
∃ cW > 0 : W (εε , E) = (εε · ·c · ·εε + ET · d · E) ≥ cW (εε T · ·εε + ET · E). (6)
2
Equations (1)–(5) give the coupled system of equations for a piezoelectric body
with respect to the components of the vector of mechanical displacements u and the
function of the electric potential ϕ. The complete statement for the piezoelectricity
(electroelasticity) problem of natural oscillations should also include the correspond-
ing boundary conditions. These boundary conditions can be divided in two types,
mechanical and electric.
To formulate the mechanical boundary conditions we assume that the boundary
S is divided in two subsets Sσ and Su (S = Sσ ∪ Su ). We will assume that at the part
of the boundary Sσ there are only surface stresses τ s
n · σ = ∇s · τ s , x ∈ Sσ , (7)
where ∇s is the surface gradient operator, associated with surface nabla-operator by

the formula ∇s = ∇ − n(∂/∂r ), where r is the coordinate, measured by the normal
to Sσ ; τ s is the second-order tensor of surface stresses.
As for purely elastic body, when taking into account the surface stresses, we adopt
that the surface stresses τ s are related to the surface strains ε s by the formula
τ s = cs · ·εε s , ε s = (∇s u · A + A · (∇s u)T )/2, (8)
where cs is the forth-order tensor of surface elastic moduli; A = I − n ⊗ n, I is

the unit tensor in IR3 . The properties of the tensor of surface elastic moduli cs are
analogous to the corresponding properties of the tensor c, i.e. csi jkl = cs jikl = cskli j ,
1
∃cU > 0, ∀εε s = ε sT : U (εε s ) = ε sT · ·cs · ·εε s ≥ cU ε sT · ·εε s , (9)
2
that follow from the condition of the positive definiteness of the surface energy
density U (εε s ).
The remaining part of the boundary Su is assumed to be rigidly fixed
u = 0, x ∈ Su , Su = ∅. (10)
To set the electric boundary conditions we assume that the surface S is also divided
in two subsets: S D and Sϕ (S = S D ∪ Sϕ ). The regions S D does not contain electrodes
and hold the following conditions
n · D = ∇s · Ds , x ∈ SD . (11)
where Ds = A · ds · A · Es , Es = −∇s ϕ, ds is the second-order dielectric permittivity

tensor that is symmetrical positive definite relatively to the vectors Es . As it can
be seen, here analogously to the boundary condition (7) at the free of electrodes
boundaries S D we take into account the influence of the surface films with the vectors
of surface electric displacement Ds and with the vectors of surface electric field Es .
The subset Sϕ is the union of M + 1 regions Sϕ j ( j ∈ J Q ∪ JV ),
J Q = {1, 2, ..., m}, JV = {0, m, m + 1, ..., M}), that does not border on each other
and are covered with infinitely thin electrodes. At these regions we set the following
boundary conditions
ϕ = Φ j , x ∈ Sϕ j , j ∈ J Q , (12)

Sϕ j n · D d S − Cϕ j nc · Ds dC = 0, Cϕ j = ∂ Sϕ j , j ∈ JQ , (13)
ϕ = 0, x ∈ Sϕ j , j ∈ JV , S j0 = ∅, (14)
where the variables Φ j do not depend on x, and nc is the vector of the unit external
normal to the contour Cϕ j . By (12)–(14) we have m open free electrodes from
Sϕ j , j ∈ J Q , the potentials Φ j at which are not initially known, but the total charge
is equal to zero and (M + 1 − m) short-circuit grounded electrodes from Sϕ j , j ∈ JV
with zero electric potentials. The limit cases of m = 0 and m = M can be considered
in the framework of unified approach (12)–(14). In the case of m = 0 all the electrodes
are short-circuit. With m = M the electric potential appears to be defined up to an
arbitrary constant, and without loss of generality for the uniqueness of the solution
we can adopt the condition (14) for one of the electrodes, for example, for S j0 .
We note that all of the electrodes are equipotential surfaces, i.e. electric potentials
on these electrodes are constant and do not depend on x. The integral conditions
(13) are analogous to the contact boundary conditions for the rigid body. However,
a distinctive property of the piezoelectric devices is the necessity of the boundary
conditions (12)–(14) and their analogues for heterogeneous problems, as they defined
the dependence between electric potentials and electric charges when piezoelectric
bodies are included in external electric circuits. Here we assume that all the areas
Sσ , Su , S D and Sϕ j have Lipschitz boundaries of the class C1 [4], and electric surface
effects are taken into account by introducing the members with the vector of surface
electric displacement Ds into relations (11), (13).
Problem (1)–(14) is the eigenvalue problem of natural oscillations of a piezoelec-
tric body with surface effects and consists in the determination of natural frequen-
cies ω and corresponding eigenfunctions u, ϕ, that deliver nontrivial solution of the
homogeneous problem. The spectral properties of this problem will be set using the
approaches applied in [5–7].
3 Generalized Problem Settings
We transfer from the classical settings (1)–(14) of the eigenvalue problems for piezo-
electric bodies with surface effects to their generalized or weak settings. Previously
we introduce the space of functions ϕ and the vector functions u, defined on V . We
denote by Hρ0 the space of vector functions u ∈ L2 with the scalar product

(v, u)Hρ0 = ρ v T · u d V. (15)
V
On the set of vector functions u ∈ C1 which satisfy homogeneous boundary condition

(10) on Su we introduce the scalar product

(v, u)Hu1 = (∇v) · ·∇u d V +
T
(A · ∇s v)T · ·∇s u · A d S. (16)
V Sσ
The closure of this set of vector functions u in the norm generated by the scalar
product (16) will be denoted by Hu1 .
On the set of functions ϕ ∈ C1 which satisfy the homogeneous boundary condition
(14) on Sϕ j , j ∈ JV , and (12) for arbitrary Φ j on Sϕ j , j ∈ J Q we introduce the scalar
product
(χ , ϕ)Hϕ1 = (∇χ )T · ∇ϕ d V + (∇s χ )T · ∇s ϕ d S. (17)
V SD
The closure of this set of functions ϕ in the norm generated by the scalar product
(17) will be denoted by Hϕ1 .
In order to formulate the generalized or weak solution of the eigenvalue problem
we apply the scalar multiplication to Eq. (4) by arbitrary vector function v ∈ Hu1 ,
and we multiply Eq. (5) by some function χ ∈ Hϕ1 . By integrating the obtained
equations on V , and by using the standard technique of the integration by parts with
Eqs. (1)–(3), (7), (8), (10)–(14), we obtain the following integral relations
c(v, u) + e(ϕ, v) = ω2 ρ(v, u), (18)

−e(χ , u) + ε(χ , ϕ) = 0, (19)
where
ρ(v, u) = (v, u)Hρ0 , c(v, u) = cv (v, u) + cs (v, u), (20)

cv (v, u) = V ε (v) · ·c · ·εε (u) d V, cs (v, u) = Sσ ε s (v) · ·cs · ·εε s (u) d S, (21)

e(ϕ, v) = − V E(ϕ) · e · ·εε (v) d V, ε(χ , ϕ) = εv (χ , ϕ) + εs (χ , ϕ), (22)

εv (χ , ϕ) = V E(χ ) · d · E(ϕ) d V ,

εs (χ , ϕ) = S D Es (χ ) · ds · Es (ϕ) d S. (23)
As it can be easily noted, the account of surface effects for piezoelectric bodies in
relations (18)–(23) is reduced to adding the forms cs (v, u) and εs (χ , ϕ). Therefore
we can use the approaches suggested in [7] for conventional eigenvalue problems of
piezoelectric bodies.
In particular the forms ρ(v, u), c(v, u), ε(χ , ϕ) are symmetrical, bilinear and
positive definite in L2 , Hu1 , Hϕ1 , respectively, while e(ϕ, v) is a bilinear form. For
fixed u ∈ Hu1 , ϕ ∈ Hϕ1 the forms e(ϕ, v) and ε(χ , ϕ) are linear-bounded functionals
in Hϕ1 . Then by the Riesz’ theorem [8] the elements eu, εϕ exist and are unique for
all χ ∈ Hϕ1
e(χ , u) = (χ , eu)Hϕ1 , ε(χ , ϕ) = (χ , εϕ)Hϕ1 . (24)
For variable u ∈ Hu1 , ϕ ∈ Hϕ1 it is obvious that eu and εϕ are linear operators acting
from Hu1 into Hϕ1 and from Hϕ1 into Hϕ1 , respectively, and an inverse exists for the
operator εϕ. Then, from (19), (24) we obtain that
εϕ = eu, ϕ = Au, A = ε−1 e, (25)
where the operator A act from Hu1 into Hϕ1 , and is linear and bounded.
Using (24), (25) we can represent the system (18), (19) in the reduced form
c̃(v, u) = ω2 ρ(v, u), (26)
where
c̃(v, u) = c(v, u) + ε(Av, Au). (27)
Note that the form c(v, u) + ε(Av, Au) is positive definite in Hu1 , that is provided by
conditions (6), (9).
Definition 9.1 We will call the triple of quantities (ω2 , u ∈ Hu1 , ϕ ∈ Hϕ1 ), which
satisfy (26) for arbitrary vector function v ∈ Hu1 or, which is equivalent (18), (19) for
arbitrary v ∈ Hu1 , χ ∈ Hϕ1 a generalized or weak solution of the eigenvalue problem
(1)–(14) for a piezoelectric body with surface effects.
By repeating arguments presented in [4] for a piezoelectric body with m = 0 and
without surface effects and in [6] for elastic body with surface stresses, we can show
that the space Hc1 , which is the closure of the set of vector function u ∈ C1 , satisfying
(10) in the norm generated by the scalar product (27), is the equivalent to Hρ0 , and
the next theorems follow from the complete continuity of the operator of embedding
from Hu1 into Hρ0 , as in the general situation [9].
Theorem 9.1 The operator equation (26) of eigenvalue problem for piezoelectric
body with account for surface effects has a discrete spectrum 0 < ω12 ≤ ω22 ≤ · · · ≤
ωk2 ≤ . . . ; ωk2 → ∞ as k → ∞, and the corresponding eigenfunctions u(k) form a
system that is orthogonal and complete in the spaces Hρ0 and Hc1 .
Theorem 9.2 The Courant–Fisher minimax principle

ωk2 = max min R(v) ,
w1 ,w2 ,...,wk−1 ∈Hu1 v =0, v∈Hu1
ρ(v,w j )=0; j=1,2,...,k−1
where R(v) is the Rayleigh quotient
c̃(v, v)
R(v) = . (28)
ρ(v, v)
We observe that the orthogonality conditions in Theorem 1 can be presented in the

forms
(u(i) , u( j) )Hρ0 = 0 , (u(i) , u( j) )Hc1 = 0 , i = j,
and also in the extended representation
c(u(i) , u( j) ) + e(ϕ ( j) , u(i) ) = 0,

−e(ϕ (i) , u( j) ) + ε(ϕ (i) , ϕ ( j) ) = 0,
where i = j, ϕ (i) = A u(i) .
4 Theorems of Natural Frequency Changes
We establish some consequences from variational descriptions of natural frequencies

of piezoelectric bodies with surface effects. We will formulate these consequences in
the form of the theorems about the change of natural frequencies under the changes
of boundary conditions and material parameters of the medium. At the beginning we
investigate the influence of the surface effects. Along with the problem formulated
we will consider the corresponding problems without surface stresses and without
surface electric charges.
So that in the problem without surface stresses instead of the boundary condition
(7) we use conventional natural condition
n · σ = 0, x ∈ Sσ ,
and for the solution u and the projection functions v we introduce the functional
space H1f u with the norm

(v, u)H1 = (∇v)T · ·∇u d V.
fu
V
By analogy, in the problem without surface electric charges (without surface dielec-
tric films) we change the boundary conditions (11) and (13), setting Ds = 0, and
also introduce the functional space H1f ϕ with the norm

(χ , ϕ)H1 = (∇χ )T · ∇ϕ d V.
fϕ
V
By comparison of the space Hu1 , Hϕ1 and H1f u , H1f ϕ , it can be concluded that
Hu1 ⊂ H1f u , Hϕ1 ⊂ H1f ϕ , (29)
as for the functions from Hu1 and Hϕ1 additional smoothness at the boundaries Sσ and
S D is required.
We note different variables for problems without surface stresses by subscripts
“ f u”, and the variables for problems without surface dielectric films by subscripts
“ f ϕ”. Obviously, the following inequalities take place
c f u (v, v) = cv (v, v) ≤ c(v, v), ∀ v ∈ Hu1 , (30)

ε f ϕ (χ , χ ) = εv (χ , χ ) ≤ ε(χ , χ ), ∀ χ ∈ Hϕ1 . (31)
The relations (29)–(31) allow to formulate the following theorems.

Theorem 9.3 The natural frequencies ωk for the problem of oscillations with surface
stresses are not less than the corresponding natural frequencies ω f uk for the problem
without surface stresses, i.e. ω2f uk ≤ ωk2 for all k.
Theorem 9.4 The natural frequencies ωk for the problem of oscillations with surface
electric charges are not greater than the corresponding natural frequencies ω f ϕk for
the problem without surface electric charges, i.e. ωk2 ≤ ω2f ϕk for all k.
To prove Theorem 3 we write the chain of inequalities that follow from (28)–(30)
ω2f u1 = inf R f u (v) ≤ inf R f u (v) ≤ inf R(v) = ω12 .

v∈H1f u v∈Hu1 v∈Hu1
The generalization of the inequality ω2f ϕ1 ≤ ω12 to the higher frequencies now follows
from the known arguments [6, 9], that use Theorem 2.
The proof of Theorem 4 is based on the inequality that follow from (29), (31)
ε(Av, Av) ≤ ε f ϕ (A f ϕ v, A f ϕ v), ∀ v ∈ Hu1 ,
with generalization of the obtained inequality ω12 ≤ ω2f ϕ1 to the higher frequencies
from the arguments [6, 7, 9], that use Theorem 2. The proofs of further theorems
will also be omitted as they can be carried out analogously to the corresponding
theorems for piezoelectric bodies described in the framework of the classical models
of electroelasticity [7] and for electric bodies with surface stresses [6].
Theorem 9.5 The substitute of short-circuited electrodes by free electrodes leads

to the increase of natural frequencies for both problems with and without surface
effects.
We note that Theorem 5 also states that the electric antiresonance frequencies are
not less than the electric resonance frequencies with the same ordinal numbers [7].
Let us also investigate the natural frequencies under the changes of certain para-
meters of the problem. We will explicitly point these changes in the formulations of
the theorems, and all the variables related to the modified problems will be marked
with a star. As above, for the initial and modified problems the parameters that are
not specified in the theorem formulations are assumed to be identical.
Theorem 9.6 If the regions with the conditions of rigid fixing are identical for two
problems, so that Su ⊃ S∗u , then ωk2 ≥ ω∗k
2 for all k.
Theorem 9.7 If the elastic modules, piezomoduli and densities for two problems
are such that c̃(v, v) ≥ c̃∗ (v, v), ρ(v, v) ≤ ρ̃∗ (v, v) for ∀ v ∈ Hu1 , then ωk2 ≥ ω∗k
2
for all k.
Theorem 9.8 If the regions with boundary conditions of electric type are identical
for two problems, so that Sϕ ⊃ S∗ϕ , Sϕ j ⊃ S∗ϕ j ; j = 0, 1, ..., M, then ωk2 ≤ ω∗k
2 for
all k.
Theorem 9.9 If the dielectric permittivities for two problems are such that ε(χ , χ ) ≥
ε∗ (χ , χ ) for ∀ χ ∈ Hϕ1 , then ωk2 ≤ ω∗k
2 for all k.
We note that the results of Theorems 5–9 are valid for both problems with and without
surface effects. From comparison of the theorems provided it can be concluded that
the same type of changes in the mechanical and electrical boundary conditions or
stiffness properties and dielectric permittivities lead to the opposite changes in the
natural frequencies of piezoelectric bodies.
5 Finite Element Approximations
For numerical solution of the eigenvalue problems (18)–(23) we can use classical
technique of finite element modal analysis [10, 11]. Let Vh be the region occupied
by the corresponding finite element mesh Vh ⊆ V, Vh = ∪k V ek , where V ek is a
separate finite element with number k. At the boundary Sh = ∂ Vh we select the
regions Shσ , Shu , Sh D , Shϕ j that approximate the boundaries Sσ , Su , S D , Sϕ j , j =
1, 2, ..., M. Then at Sh with the corresponding boundaries finite functional spaces
1 , H1 can be introduced analogously to spaces H1 , H1 .
Hhu hϕ u ϕ
We will search the approximate solution {uh ≈ u, ϕh ≈ ϕ} of the problem (18),
(19) at the finite element mesh Vh = ∪k V ek in the form
uh (x) = NuT (x) · U, ϕh (x) = NϕT (x) · Φ , (32)

where NuT is the matrix of the form functions (basis functions) for the displacements,
NϕT is the row vector of the form functions for electric potential, U, Φ are the global
vectors of nodal values for displacements and electric potential, respectively. The
projection functions v ∈ Hhu1 , χ ∈ H1 can be presented in the form
hϕ
Φ.
v(x) = NuT (x) · δU, χ (x) = NϕT (x) · δΦ (33)
According to the conventional finite element technique we will write the weak
1 and H1 for the region V =
setting of the problem (18), (19) in finite spaces Hhu hϕ h
∪k V ek with corresponding boundaries. Substituting (32), (33) in problem (18), (19)
for Vh , we will have
−ω2 Muu · U + Kuu · U + Kuϕ · Φ = 0, (34)

−Kuϕ ∗ ·U+K
ϕϕ · Φ = 0, (35)
a ek ek em
where Muu = a Muu ek , K
uu = Kuu , Kuϕ = a Kuϕ , Kϕϕ = a Kϕϕ are
the global finite element matrices obtained from the corresponding element matrices
as the result of the assembly procedure ( a ). According to (20)–(23) the element
matrices

Muuek
= ρNue · NueT d V, Kuu
ek
= Kvuu
ek
+ Ksuu
ek
, (36)
V ek

ek
Kvuu = u · c · Bu d V, Ksuu =
BeT e ek
su · cs · Bsu d S ,
BeT e
(37)
V ek Sσek

ek
Kuϕ = u · e · Bϕ d V, Kϕϕ = Kvϕϕ + Ksϕϕ ,
BeT T e ek ek ek
(38)
V ek

ek
Kvϕϕ = Bϕ · d · Bϕ d V, Ksϕϕ =
eT e ek
sϕ · ds · Bsϕ d S ,
BeT e
(39)
V ek S ek
D
Be(s)u = L(∇(s) ) · NueT , Be(s)ϕ = ∇(s) NϕeT , (40)

⎡ ⎤
∂(s)1 0 0 0 ∂(s)3 ∂(s)2
∂
LT (∇(s) ) = ⎣ 0 ∂(s)2 0 ∂(s)3 0 ∂(s)1 ⎦ , ∂s j = ∂ j − n j , (41)
0 0 ∂(s)3 ∂(s)2 ∂(s)1 0 ∂r
where Sσek , S D
ek are the edges of finite elements facing the regions S , S
hσ h D with
eT eT
given surface effects, Nu , Nφ are the matrices and row vectors of approximating
basis functions, respectively, that are defined at separate finite elements; j = 1, 2, 3.
In (36)–(41) we used matrix-vector notations: c, cs are 6 × 6 matrices of elastic bulk
and surface modules, c(s)αβ = c(s)i jkl ; α, β = 1, ..., 6; i, j, k, l = 1, 2, 3 with the
correspondence law α ↔ (i j), β ↔ (kl), 1 ↔ (11), 2 ↔ (22), 3 ↔ (33), 4 ↔
(23) = (32), 5 ↔ (13) = (31), 6 ↔ (12) = (21); e is 3 × 6 matrix of piezomoduli
(eiβ = eikl ).
We note that in (34)–(41) the global and element matrices of mass and stiffness
Muu , Muuek , K
vuu , Kvuu are formed in the same way as for purely elastic body, and the
ek
matrices Kuϕ , Kuϕ , Kvϕϕ , Kvϕϕ

ek ek are identical to the corresponding matrices of piezo-
electric bodies. The matrices Ksuu and Ksϕϕ are defined by the surface stresses and
surface electric charges, respectively. These matrices are analogous to the stiffness
matrices for surface elastic membranes and the matrices of dielectric permittivities
for surface dielectric films. Hence, for implementing the finite element piezoelectric
modal analysis for the bodies with surface effects it is necessary to have surface
structural membrane elements and surface finite elements of dielectric films along
with ordinary solid piezoelectric finite elements. The rest of the modal analysis tech-
nique for the piezoelectric bodies with surface effects can repeat similar technique
for piezoelectric bodies without surface effects. In particular, the same solvers for
eigenvalue problems can be used to determine practically important frequencies of
electric resonances and antiresonances (see for example [13, 15]).
6 Numerical Examples
Let us consider the problem of natural oscillations of a longitudinally polarized

piezoelectric rod with a circular cross-section made of zinc oxide (Z n O). We adopt
that the rod has the length l = 1 × 10−6 (m) and the radius R = 0.05 × 10−6 (m).
For zinc oxide (which is 6 mm-class material) we set the following bulk moduli
[12]: ρ = 5.676 × 103 (kg/m3 ), c11 E = 2.097 × 1011 , c E = 1.211 × 1011 , c E =
12 13
1.051 × 1011 , c33
E = 2.109 × 1011 , c E = 0.425 × 1011 (N/m 2 ), e = −0.61, e =
44 31 33
1.14, e15 = −0.59 (Cl/m2 ), ε11 S = 7.38 ∗ ε , ε S = 7.83 ∗ ε , ε = 8.85 ×
0 33 0 0
10−12 (F/m). We will assume that the ends of the rod are covered with electrodes,
for the upper electrode ϕ = 0, and for the lower electrode ϕ = 0 in the problem of
determination of the electric resonance frequencies fr k . Conditions (12), (13) on the
lower electrode are satisfied for the problem of determination of the electric antires-
onance frequencies f ak , where f = ω/(2π ). The lower end of the rod is considered
to be rigidly fixed.
We refer the rod to the cylindric coordinate system 0r θ z, directing z-axis along
the symmetry axis of the rod and choosing the coordinate system origin to lay in the
plane of the lower end of the rod. At the lateral surface of the rod we select an area
0 ≤ z ≤ l1 , and we will vary the boundary conditions at the lateral surface r = R
when 0 ≤ z ≤ l1 , l1 ≤ z ≤ l and at the upper end z = l.
The part of the boundary z = l (the upper end) and r = R, l1 ≤ z ≤ l (the
upper part of the lateral surface) is considered to be free of mechanical stresses or as
boundary with surface stresses. In order to illustrate Theorems 3 and 7 let us compare
the first two frequencies of electric resonance and antiresonance at the absence and
the presence of the surface stresses and under the increase of the stiffness moduli
of the surface membrane that define the surface stresses. For both problems we set
l1 = 0, i.e. we refer to all lateral surface and top end z = l as Sσ . Moreover, in both
problems we do not take into account the effect of the dielectric films.
These and further problems on the natural frequencies will be solved as axisym-
metric problems using ANSYS finite element software. We divide the meridian
section of the rod into quadrilateral eight-node finite elements PLANE13 with the
options for axisymmetric piezoelectric analysis. Let us choose the number of finite
elements along the radius to be equal to 30, and along the rod length to be equal to 600.
We note that, as the computations have shown, such sufficiently fine mesh provided
enough accuracy of the computations under different variations of the input parame-
ters in all the examples considered. To model surface stresses on the lateral surface
Sσ we place axisymmetric elastic shell finite elements SHELL208 with options of
only membrane stresses. Such membrane elements will approximate the boundary
conditions (7) with the constitutive Eq. (8) with appropriate choice of elastic moduli
of membrane and its thickness. If we adopt that the surface elastic moduli cs are
the elastic moduli of isotropic body, than it is enough to set only the surface elas-
tic Young’s modulus E s and surface Poisson’s ratio νs . Then for equivalent elastic
membrane and the corresponding membrane finite elements in ANSYS finite ele-
ment package it is necessary to set the Young’s modulus of the membrane E m , the
Poisson’s ratio of the membrane νm and the thickness of the membrane h m so that
the equalities E s = h m E m , νs = νm take place. Therefore, for the equivalent mem-
brane the values E m and h m are not significant separately but in their multiplication
E s = h m E m . Formally putting h m = lm , we will set the surface Young’s modulus
E s in the form E s = h m E m , E m = ks E 0 , varying the proportionality ratio ks . For
the computations we set that E 0 = 2 × 1011 (N/m2 ), νs = 0.3.
We note that to insure the accuracy of the finite element computations in ANSYS
due to the smallness of the geometric sizes of the rod here it is convenient to transfer
to dimensionless coordinates and parameters that can be introduce as following:
ũ = u/l, x̃ = x/l, ϕ̃ = ϕ/(E d l), c̃ E = c E /c33 S S 11 σ =
E , ẽ = eE /c E , d̃ = d /ε S , σ̃
d 33
E , Ẽ = E/E , D̃ = DE /c E , ω̃ = ωT , T = l/v E , v E =
σ /c33 d d 33 d d 3 3
E /ρ, E =
c33 d

E /ε S . Then the problem can be solved in dimensionless form for the variables
c33 11
marked with (. ˜. .), and after solving this problem we can return to dimensional
quantities.
Figure 1 (left) illustrates the graphs of the dependencies of the first two electric
resonance frequencies fr 1 and fr 2 on the coefficient ks , plotted on the x-axis in
a logarithmic scale, i.e. at ks = 0 = 10−∞ , 10−6 , 10−4 , 10−2 etc. As it follows
from Theorems 3, 7 under the increase of elastic stiffness of the surface member the
natural frequencies also increase, and they increase considerable at ks ≥ 10−2 .
If along with the electric resonance frequencies we find the electric antireso-
nance frequencies f a1 and f a2 , then we can find dynamic frequency coefficients of
electromechanical coupling by formulas kd j = 1 − ( fr j / f a j )2 , j = 1, 2. These
coefficients are responsible for electric activation of the corresponding oscillation
modes and for effectiveness of the mechanical and electric energy transformation.
The corresponding graphs of the dependence of the electromechanical coupling coef-
Fig. 1 Resonance frequencies fr j (left) and coupling coefficients kd j (right) versus stiffness
coefficient ks
ficients kd j on ks are shown in Fig. 1 (right). As it can be seen, for the example
considered the electromechanical coupling coefficients decrease with the increase of
the value of surface stiffness, also more considerably at ks ≥ 10−2 . This property
is quite expected for the example considered, but in general it does not follow from
the established theorems and for other problems and oscillation modes can not take
place.
Account of surface charges and dielectric films is illustrated by Fig. 2 for
the problems without surface stresses. Here for computations the simulation of
the dielectric film is implemented by adding to the edge r = R, 0 ≤ z ≤ l the
dielectric layer consisting of finite elements PLANE13 with options of axisym-
metric piezoelectric analysis at negligible small elastic stiffness and piezomoduli.
Basic dielectric permittivities of three-dimensional dielectric layer are set as follows:
ε0v11 = 0, ε0v22 = 7.38 ∗ ε0 , ε0v33 = 7.83 ∗ ε0 (F/m). Therefore, basic dielectric
permittivities along the radial axis (axis 1) are set to be zero and the other values
coincided with the dielectric permittivities of the material Z n O. The thickness h f
Fig. 2 Resonance frequencies fr j (left) and coupling coefficients kd j (right) versus permittivity
coefficient k f
Fig. 3 Resonance frequencies fr j (left) and coupling coefficients kd j (right) versus boundary sizes
Su
of the dielectric layer is assumed to be equal to 1 × 10−8 (m). Such layer simulates
the dielectric film and boundary conditions (11) with surface dielectric permittivities
εs j j = h f εv j j , j = 1, 2, 3 when εv j j = k f ε0v j j , k f = 1. Further in numerical com-
putations the multiplier k f was changed from 0 to 104 , and accordingly the dielectric
permittivities εv j j were changed. The results of the computations are shown in Fig. 2,
where Fig. 2 (left) illustrated the graphs of the dependencies of the first two electric
resonance frequencies fr 1 and fr 2 on coefficient k f , plotted along the horizontal axis
in logarithmic scale, and Fig. 2 (right) illustrated the graphs of the dependency of the
electromechanical coupling coefficients kd j on k f . As it follows from Theorems 4,
9 with the increase of dielectric permittivities of surface film the natural frequencies
decrease. A small decrease of the resonance frequencies can be explained by small
dielectric permittivity coefficients and small piezomoduli for piezoelectric material
zinc oxide Z n O. As it is seen from Fig. 2 (right), for the example considered the
electromechanical coupling coefficients decrease with the increase of the variable
k f faster than the resonance frequencies.
Figure 3 illustrate the dependencies of the resonance frequencies on mechanical
boundary conditions and Theorem 6. Here we consider the problem for the rod
with additionally varied boundary of rigid fixing at the lateral border r = R at 0 ≤
z ≤ l1 . Figure 3 (left) shows the graphs of the dependencies of the electric resonance
frequencies fr 1 and fr 2 on the value of the ratio l1 /l, and Fig. 3 (right) illustrates the
graphs of the dependencies of the electromechanical coupling coefficients kd1 and
kd2 on the same ratio l1 /l for the rod without taking into account the surface effects.
As it follows from Theorem 6, the resonance frequencies increase with the increase
of the size of the rigid fixing. As it is seen from Fig. 3 (right), the electromechanical
coupling coefficient decrease with the increase of the rigid boundary, as the oscillation
modes with big boundaries of rigid fixing should more likely to be less electrically
active.
Finally, the dependencies of the resonance frequencies on the electric boundary
conditions and the statement of Theorem 8 are demonstrated in Fig. 4. Here in the
problem for the rod we vary the boundary of the lower electrode under the increase
Fig. 4 Resonance frequencies fr j (left) and coupling coefficients kd j (right) versus boundary
sizes Sϕ
of this boundary along the lateral surface r = R, when 0 ≤ z ≤ l1 . Figure 3 (right)

shows the graphs of the dependencies of the electric resonance frequencies fr 1 and
fr 2 on the value of the ratio l1 /l, and Fig. 4 (right) shows the graphs of the depen-
dencies of the electromechanical coupling coefficients kd1 and kd2 on the same ratio
l1 /l for the rod without surface effects. As it follows from Theorem 8, the reso-
nance frequencies decrease with increase of the size of the electrode surface. As for
Fig. 2 (left), small decrease of the resonance frequencies can be explained by small
coefficients of dielectric permittivities and piezomoduli for piezoelectric material
zinc oxide Z n O. Moreover, as it is seen from Fig. 4 (right), the electromechani-
cal coupling coefficient kd1 decreases with the increase of the electrode surface,
and kd2 increases at the beginning and decreases only at l1 /l ≥ 0.4. Therefore,
in the example considered different oscillation modes show different activity under
the change of the sizes of electrode surfaces. Figure 4 (right) reflects the fact that
the dependencies of the electric resonance frequencies fr k and the electric antires-
onance frequencies
f ak established by Theorems 3–9 not necessarily apply to the
values kdk = 1 − ( fr k / f ak )2 , i.e. the electromechanical couplings of the oscilla-
tion modes in general case do not have to satisfy the inequalities of Theorems 3–9
for the resonance frequencies fr k or f ak .
7 Concluding Remarks
This paper has considered the problems of natural oscillations of piezoelectric bod-
ies of nanosizes in the framework of the piezoelasticity theory with account for
surface effects induced by surface stresses and surface dielectric films. Classical
and generalized settings of the spectral problems were formulated in expanded and
reduced forms. For generalized settings the corresponding functional spaces were
introduced. It was proved that the spectrum was discrete and real and the eigenvectors
were orthogonal.
The theorems that establish the dependencies of natural frequencies of piezoelec-

tric nanosize bodies were formulated with account for surface stresses and surface
dielectric films, and the change of the rigidly fixed boundary, boundaries with elec-
trodes and material parameters of piezoelectric nanosize bodies. The differences in
the problems of determining the frequencies of electric resonance and antiresonance
were noted.
It was noted that the same changes of mechanical and electric boundary conditions
of stiffness characteristics and dielectric permittivities lead to the opposite changes
in the natural frequencies. All dependencies were established for the piezoelectric
bodies without surface effects, as well as for the bodies with account for surface
stresses and surface dielectric films.
Finite element approximations and the corresponding generalized matrix prob-
lems were suggested for numerical solution of the spectral problems for piezoelectric
bodies with surface effects. The results were illustrated with a numerical example for
obtaining the natural frequencies of nanosize piezoelectric rod made of zinc oxide
under different varied parameters of the problem. It was shown that here standard
finite element software could be used with additional introduction of surface mem-
brane elements and surface dielectric films in the computation models.
The results obtained generalize the known results for purely piezoelectric bodies
and piezoelectric and elastic bodies with voids without surface effects [7, 13–15] and
for the elastic bodies with surface stresses [5, 6] to the problems for piezoelectric
nanosize bodies with surface effects.
References
1. Duan, H.L., Wang, J., Huang, Z.P., Karihaloo, B.L.: Size-dependent effective elastic constants
of solids containing nano-inhomogeneities with interface stress. J. Mech. Phys. Solids 53,
1574–1596 (2005)
2. Duan, H.L., Wang, J., Karihaloo, B.L., Huang, Z.P.: Nanoporous materials can be made stiffer
than non-porous counterparts by surface modifcation. Acta mater. 54, 2983–2990 (2006)
3. Jing, G.Y., Duan, H.L., Sun, X.M., et al.: Surface effects on elastic properties of silver
nanowires: contact atomic-force microscopy, Phys. Rev. B 73, 235409-1–235409-6 (2006)
4. Belokon, A.V., Vorovich, I.I.: Some mathematical problems of the theory of electroelastic
solids, In: Current problems in the mechanics of deformable media. Izv. Dnepropetr. Gos.
Univ., Dnepropetrovsk (1979)
5. Altenbach, H., Eremeyev, V.A., Lebedev, L.P.: On the existence of solution in the linear elas-
ticity with surface stresses, Z. Angew. Math. Mech. 90(3), 231–240 (2010)
with surface stresses, Z. Angew. Math. Mech. 91(9), 699–710 (2011)
7. Belokon, A.V., Nasedkin, A.V.: Some properties of the natural frequencies of electroelastic
bodies of bounded dimentions. J. Appl. Math. Mech. (PMM) 60, 145–152 (1996)
8. Riesz, F., Szokefalvi-Nagy, B.: Functional Analysis. Dover, New York (1990)
9. Mikhlin, S.G.: Variational Methods in Mathematical Physics. Pergamon Press, Oxford (1964)
10. Bathe, K.J.: Finite Element Procedure. Prentice-Hall, Englewood Cliffs, NJ (1996)
11. Zienkewicz, O.C., Morgan, K.: Finite Elements and Approximation. Wiley, NY (1983)
12. Dieulesaint, E., Royer, D.: Ondes elastiques dans les solides. Application au Traitement du
Signal. Masson, Paris (1974)
13. Iovane, G., Nasedkin, A.V.: Some theorems about spectrum and finite element approach for
eigenvalue problems for elastic bodies with voids. Comput. Math. Appl. 53, 789–802 (2007)
14. Iovane, G., Nasedkin, A.V.: Modal analysis of piezoelectric bodies with voids. I. Mathematical
approaches, Appl. Math. Model. 34(1), 60–71 (2010)
15. G. Iovane, A.V. Nasedkin, Modal analysis of piezoelectric bodies with voids. II. Finite element
simulation, Appl. Math. Model. 34(1), 47–59 (2010)
Stability and Structural Transitions in Crystal
Lattices
Ekaterina Podolskaya, Artem Panchenko and Anton Krivtsov
Abstract The advance in nanotechnology has lead to necessity to determine strength

properties of crystal structures. Stability of a structure under finite deformations is
closely connected with its strength. In this work stability of plane triangular (single
atomic layer of FCC and HCP) and FCC lattices under finite strain is investigated.
Analysis and modeling based on discrete atomistic methods is proposed. The medium
is represented by a set of particles which interact by a pair force central potential,
e.g. Lennard-Jones and Morse. Direct tensor calculus is used. Dynamic stability
criterion is established: frequency of elastic waves is required to be real for any
real wave vector. The considered approach allows to describe structural transitions
in solids on the base of stability investigation of pre-strained crystal lattices. The
results of direct MD simulation do not contradict the results of the calculations.
1 Introduction
Recent advance in nanotechnology has lead to the necessity of determining mechan-

ical properties of the minute objects. Due to being small in size such objects are often
without defects, thus their strength, for instance, is close to ideal. According to [1],
ideal strength is the maximum applied stress that an object can endure. Under this
E. Podolskaya (B) · A. Krivstov

Institute for Problems in Mechanical Engineering, Russian Academy of Sciences,
61, Bolshoy pr., V.O., St. Petersburg 199178, Russia
e-mail: katepodolskaya@gmail.com
A. Krivstov
e-mail: akrivtsov@bk.ru
A. Panchenko
e-mail: artemqt@yandex.ru
E. Podolskaya · A. Krivstov · A. Panchenko
St. Petersburg State Polytechnical University, Politekhnicheskaya str. 29,
St. Petersburg 195251, Russia

124 E. Podolskaya et al.
definition it is assumed that the object remains stable under minor strain or stress
deviations along the loading path. On the other hand, it is crucial to make sure that the
object does not loose stability in terms of arbitrary minor perturbations at each strain
or stress increment. If the object is described within continuum mechanics approach,
analysis of ellipticity of equilibrium equations is to be carried out in order to find first
failure strains [2]. However, continuum analysis is not always valid for nanoscale
objects [3], because at this level influence of internal structure cannot be neglected.
An ideal crystal lattice is one of the simplest models to consider within atomistic
approach. The theory has been developed since works of Born [4], where a criterion
for infinitesimal uniform deformations is established. However, it was shown in [5]
that this criterion does not give adequate results if finite deformation is imposed.
Moreover, further problems appear if the deformation field is inhomogeneous. For
this case in continuum mechanics certain apparatus is developed, e.g. in [6]. As for
atomistic approach, there are at least three ways to find the solution: homogenization
(long-wave approximation etc.) and application of continuum methods, direct inves-
tigation of, e.g. corresponding spring system, and computer simulation. There has
been a number of works, e.g. [7] for FCC (face-centered cubic) lattice under triax-
ial compression, which showed structural transition to BCC (base-centered cubic).
Another series of works [8, 9] is devoted to both 2D (square lattice) and 3D (cubic
lattice) structures, for which macro-(continuum) and microscopic criteria are used
to obtain failure surfaces, both in case of homogeneous and inhomogeneous initial
deformation. Recently, there have appeared independent investigations of graphene
stability [10, 11]; its lattice should be described with more sophisticated interaction
forces.
Tensor Notation
Let us introduce the following notation concerning direct tensor calculus [2] used in
this work. Vectors are denoted by lower-case letters in boldface, e.g. a, tensors are
denoted by upper-case letters in boldface with a digit specifying the rank (if the rank is
not equal to two), e.g.4 A, and for scalars italics is used, e.g. A. No special sign denotes
tensor, or dyadic, product, i.e. ab is a dyad, abc is a third-rank tensor etc. For scalar
product symbol · is used, and abc · · · def = (c · d) (b · e) (a · f). The notation for the
∂ax ∂a y ∂az
divergence of vector a is ∇ ·a = + + , where ∇ is Del operator. Gradient
∂x ∂y ∂z
∂a ∂a ∂a ∂a ∂ax ∂a y ∂az
of vector a is ∇a = ix + iy + iz , where = ix + iy + iz ,
∂x ∂y ∂z ∂x ∂x ∂x ∂x
x, y, z are Cartesian coordinates and ix , i y , iz form the corresponding basis of unit
vectors. Transposed gradient of vector a is denoted by a∇.
2 Statement of the Problem
In this work mixed approach is proposed, which includes homogenization and is

similar to [10], but regards simpler objects in order to diminish computational diffi-
culty and obtain as much as possible analytically. Firstly, only simple lattices are
Stability and Structural Transitions in Crystal Lattices 125
considered not to be distracted by sublattices-induced instabilities. Secondly, as

shown in [12], central pair force interaction is applicable for stability analysis of
close-packed lattices. Following the works mentioned in the previous section, we do
one more simplification, which is consideration of periodic, i.e. infinite, structures,
thus no surface effects will be observed. We use Lennard-Jones and Morse potentials
(1), because they depend only on interatomic distance, they have only 2 and 3 para-
meters respectively and also they provide repulsion upon compression and attraction
upon stretching. For 2D case triangular lattice is regarded, which is an atomic layer
of FCC and HCP (hexagonal close-packed) lattices. For 3D case FCC is considered,
as BCC is non-close-packed, HCP is complex and others are not so widespread as
these three.
a 6
−2θ ( ar −1) −θ ( ar −1) a 12
M (r ) = D e − 2e , L J (r ) = D −2 .
r r
(1)
Parameters D and θ are responsible for the potential well depth and width. Near
the equilibrium position if θ = 6 Morse potential is equivalent to Lennard-Jones
potential with the same values of the potential well depth and equilibrium distance
a [13]. An important distinction of Morse potential from Lennard-Jones potential
is that during the compression of the material towards r = 0 the interaction force
remains finite, e.g. if θ = 6, the repulsion force has the order of 106 D/a, which
is preferable for computer simulations under strong compression. In addition, rapid
attenuation of exponents in Morse potential allows us to take into account the smaller
number of coordination spheres.
The procedure of stability criterion derivation and explicit results for 2D case
can be found in [14, 15]. The main idea is as follows. Let us consider a lattice
which is infinite and without defects, not to account for boundary conditions and
inhomogeneities. Using long-wave approximation [4]
◦
ak ek ≈ ak0 ek0 · ∇ r, (2)
we can write equilibrium equations in Piola form [13]
◦ 1
ρ0 ü =∇ ·P, P=− Fk ak0 ek0 ek , (3)
2V0
k
where u is displacement vector, P is Piola stress tensor, ak0 and ek0 are the reference
bond lengths and directions respectively, Fk , ak and ek are the current forces, the
◦
current bond lengths and directions, ρ0 is density and ∇ is Del operator, both in the
◦
reference configuration, and ∇ r is transposed deformation gradient. Then, let us find
the first variation of (3) which takes the form of the following wave equation (4) for
arbitrary homogeneous deformation field
v̈ = 4 Q · · · ∇∇v, (4)
where v = δu is the first variation of the displacement vector, ∇ is Del operator

in current configuration. 4 Q is a fourth-rank tensor that depends on the first and
second derivatives of the interaction potential (current forces in bonds Fk and bonds’
stiffness Ck ) as well as on the geometry of particle surroundings (5)
1
4
Q= E + 4 B ,
ρ0 V0

1 1 2 Fk
=− Fk ak ek ek , 4 B = ak C k + ek ek ek ek . (5)
2 2 ak
k k
Here V0 is the unit cell volume in reference configuration, E is second rank unit
tensor.
The solution of (4) in the wave form is
v = v0 eiωt eik·r , (6)
where k is wave vector and ω is frequency. Thus, for any real wave vector frequency
has to be real, i.e. ω2 > 0, so that additional minor solution v does not contain expo-
nential growth. This demand leads to positive definiteness of tensor D = 4 Q · ·kk
which looks similar to acoustic tensor that is to be positive definite to provide ellip-
◦
ticity [2], but it is not, because 4 Q is not in fact equal to ∂P/∂ ∇ r. If they were
◦
equal, wave equation (5) would contain ∇, not ∇. Hence, instability is associated
with exponential growth of the solution for perturbed state.
3 Triangular Lattice
For biaxial strain along the axes, shown in the bottom of Fig. 1, it is possible to obtain
analytical solution [14, 15] in terms of components of 4 Q
√
Q 11 > 0, Q 21 > 0, Q 12 > 0, Q 12 > 0, B > − AC,
A = Q 11 Q 21 , C = Q 12 Q 22 , 2B = Q 11 Q 22 + Q 12 Q 21 − 4Q 244 , (7)
where two indices instead of four are used due to the symmetry. All vectors and
tensors introduced in the previous section are two-dimensional. In the Fig. 1 the
stability regions of the 2D triangular lattice are plotted gray, ε1 and ε2 are the linear
parts of the Cauchy-Green deformation tensor, the interaction is described by Morse
potential (1) with θ = 6.
To check the adequacy of these results for all configurations elastic modulae are
calculated using the formula for Cauchy stress tensor [13]
1
σ =− ak Fk ek ek , (8)
2V
k
√
where V = 3a 2 /2 (1 + ε1 ) (1 + ε2 ) is current volume of the unit cell. It turns
out that the boundaries of stability regions correspond to the loss of positivity of
Young modulae and shear modulae (plural is due to anisotropy). Note that during the
analysis it is crucial to take at least two coordinational spheres into account, despite
the cliche that if you deal with a close-packed lattice only first sphere is sufficient.
As shown in the bottom of the Fig. 1, structural transition from vertical to horizontal
orientation of the lattice is described within stability analysis. Consideration of larger
amount of atoms does not lead to major alterations.
Analysis similar to macroscopic [8] was carried out which showed that real ellip-
ticity condition is necessary but not sufficient (at least less sufficient, than this) for
2D case. Nearly the same results were achieved with Lennard-Jones potential. The
main difference is that this interaction provides stability during compression right
up to deformations arbitrarily close to point ε1 = ε2 = −1. This effect contradicts
the results for FCC lattice, achieved in [7].
In addition, an MD (molecular dynamical) simulation is carried out. The sim-
ulation technique is described in [13]. For a series of deformed configurations
we perform the following computational experiment. As the initial condition, we
Fig. 1 Top stability regions of the triangular lattice in deformation space ε1 , ε2 , Morse potential
√
with θ = 6. On the boundaries positivity is lost (7) by: 1 Q 11 , 2 Q 22 , 3 Q 21 , 4 Q 12 , 5 B + AC,
also 1, 2, 5 by Young modulae, 3, 4 by shear modulae. Bottom transition from vertical to horizontal
orientation of the triangular lattice. Digits denote the coordinate axes. The unit cell is gray, the
reference atom is marked by a circle, the atoms of the first coordination sphere—by circles of a
smaller radius, the atoms of the second coordination sphere—by empty circles
Fig. 2 Unstable configuration

after 105 (left) and 3 × 105
(right) integration steps. Black
ovals mark “crack” initiation
zones
construct a triangular lattice in the deformed state with periodic boundary condi-
tions, that account for infinite lattice. The interaction is described by means of the
same Morse potential. The initial kinetic energy of the particles does not exceed
0.0002D. The system evolution is described by the solution of the Cauchy problem
for the set of ordinary differential equations

N
rk − rn
m r̈k = F (|rk − rn |) , (9)
|rk − rn |
n=1
where N is the number of particles, m is the particle mass, and rk is the radius-vector
of the kth particle. If further we observe oscillations of the kinetic energy around a
certain value not exceeding 0.0002D, we conclude that this configuration is stable.
If we observe a sudden growth of the kinetic energy, the deformed configuration
is considered unstable. A very good agreement with analytical results is observed.
However, in MD one can only distinguish between 100 % unstable cases and cases,
when instability has not been reached. In addition, the more accurate regions’ bor-
ders are needed, the longer lasts the calculation. Stable regions endured 3 × 105
integration steps, whereas others—not more than 105 steps, excluding border zones.
MD experiment shows, what exactly happens after stability is lost: either the material
may become liquid, or a crack may appear (see Fig. 2).
Similar results were achieved for deformation including shear [15], described by
deformation gradient with the following affine transformation

◦ 1 + ε1 tgϕ21
r ∇∼ . (10)
0 1 + ε2
There are only three elements in the tensor (10) in order to exclude solid-body
rotations from consideration.
4 FCC Lattice
In 3D case more or less analytical results can be obtained only for diagonal affine
transformation, whose eigenvectors coincide with axes of cubic symmetry, and are
partially presented in [16]
⎛ ⎞
◦
1 + ε1 0 0
r ∇∼ ⎝ 0 1 + ε2 0 ⎠ . (11)
0 0 1 + ε3
We use Morse potential, because Lennard-Jones allows infinite compression

which contradicts [7]. Three coordinational spheres are considered, because the dis-
tance between the reference atom and the atoms of the third sphere is the same as the
distance between the reference atom and the atoms of the second sphere in triangular
lattice. Positive definiteness of tensor D, i.e. stability, is ensured if
D11 > 0, D11 D22 − D12 D21 > 0, det D > 0. (12)
Left parts of (12) are homogeneous functions of wave vector components of degree
two, four and six respectively, and contain only even degrees. Inequalities (12) should
hold for any real wave vector. In this case we cannot fully exclude wave vector com-
ponents from consideration and obtain stability criterion only in terms of components
of 4 Q. However, first of all, we have a necessary condition of D11 positive definite-
ness. Moreover, we can write a series of sufficient conditions by extracting quadratic
forms from left parts of inequalities (12). Then, for those cases, when only necessary
condition shows stability, Monte-Carlo method is used.
Proposition 1 Suppose a homogeneous polynomial P(x, y, z) is positive for x > 0,
y > 0, z > 0. Then substitution z = 1 − x − y leads to positivity of P̃(x, y) for
x > 0, y > 0, x + y < 1.
This proposition is used to speed up the Monte-Carlo calculations, as inequalities (12)
contain only even degrees of wave vector components. In the Fig. 3 we can observe
a major stable area, which resembles 2D case, and three additional zones, which
make the region non-convex. After calculating coordinational numbers of deformed
lattices that form additional zones, we can conclude that they are compressed BCC
lattices.
Fig. 3 Stability region of FCC lattice in deformation space ε1 , ε2 , ε3 , Morse potential with θ = 6,
three coordinational spheres. Grey points theoretical result, black points MD result
Fig. 4 Transitions: a from

FCC (left) to BCC (right),
b from FCC to FCC
Using Bain method [17] we can write an affine transformation from equilibrium
FCC to equilibrium BCC (see Fig. 4a)
√
2 ρ BCC 2 ρ BCC
ε1 = ε2 = √ − 1, ε3 = √ −1 (13)
3 ρ FCC 3 ρ FCC
Here we need to take into account so-called “bond compression” which occurs when
more than one coordinational sphere is regarded: equilibrium distance ρ between
neighboring atoms is smaller than that of the potential.
Due to topological differences between FCC and BCC, two spheres of FCC con-
tain 18 atoms, and two spheres of BCC have only 12. Hence, if initial FCC has
equilibrium with, e.g. two spheres, stress tensor for obtained BCC will be non-zero.
This problem can be solved by cut-off interaction, e.g. [13]
F̃(r ) = k(r )F(r ), (14)
where k(r ) is shape function

⎧
⎪
⎪ 1, r ≤ b,
⎨
⎪ 2
2 2
r −b 2
k(r ) = 1− , b < r ≤ acut , (15)
⎪
⎪ 2 − b2
acut
⎪
⎩
0, r > acut
Here acut is the cut-off distance, b is the critical bond length, i.e. F (b) = 0.
Now, if we plot stress-strain diagram for cut-off smooth potential (14) on the
base of Morse potential, we will see, that equilibrium BCC may be gained from
equilibrium FCC by simple uniaxial compression, e.g. σ1 = 0, σ2 = σ3 = 0 (see
Fig. 5 for θ = 4). Due to symmetry there are all in all three equilibrium BCC
Fig. 5 Left uniaxial loading, hatching indicate stability region. Right stability region of FCC lattice
in deformation space ε1 , ε2 , ε3 , smooth cut-off Morse potential with θ = 4, acut = 10a
lattices. Moreover, equilibrium BCC is unstable if θ = 6, and Lennard-Jones does

not describe BCC-zones at all, and these results correspond to [7, 18]. If we make the
potential well wider, BCC will be stable, though more spheres should be accounted
for (see Fig. 5). Unfortunately, BCC zones do not separate from FCC, leaving the
possibility of stable FCC-BCC transition. On the other hand, stability region is non-
convex (Fig. 5), so “Bain deformation” [17], which is accomplished by strain, not
stress, will provide an unstable zone between FCC and BCC equilibria.
The next step is to include shear into consideration. To get rid of as many solid-
body rotations as possible, the following transformation is used
⎛ ⎞
◦
1 + ε1 tgϕ21 0
r ∇∼ ⎝ 0 1 + ε2 tgϕ32 ⎠ . (16)
tgϕ13 0 1 + ε3
Using Bain method [17] again, we can find six FCC lattices of the following origin
(see Fig. 4b)
√
3 2 1 1
ε1 = √ − 1, ε2 = √ − 1, ε3 = √ − 1, tgϕ21 = ± √ , (17)
2 3 2 6
which may look as if we just turned one of the axis of cubic symmetry to [1,1,1] axis.
MD simulation was carried out for triaxial strain and showed again a good agree-
ment except for the “tail” zone (see Fig. 3), which is due to different number of coor-
dinational spheres considered: the more atoms, the longer the “tail”, i.e. maximum
compression for uncut Morse potential varies from 60 % for three coordinational
spheres to 75 %. As stated before, Lennard-Jones potential is not suitable for MD
under high compression, because of infinite forces upon infinite compression.
In addition, analysis for FCC in different axes is performed, so that triangular
lattice plane problem could be accounted for. Again, two major regions in triaxial
strain space are obtained, but their cross-sections differ from 2D results, since in 2D
Fig. 6 Stability region of

triangular lattice for 2D (light
gray) and 3D (dark gray)
wave vector
study only 2D wave vectors are considered (see Fig. 6). Thus, we can conclude, that
vast stability region at compression vanishes, if minor perturbations in third direction
occur.
5 Concluding Remarks
In this work stability analysis of infinite triangular and FCC lattices without defects
is carried out. Instability is associated with exponential growth of the solution for
perturbed state. The considered approach allows to describe structural transitions on
the base of stability investigation of pre-strained crystal lattices (see Figs. 1 and 4).
FCC–BCC transition is examined, and several conclusions can be drawn. Due to
topological differences between the lattices smooth cut-off interaction force is to be
used. Lennard-Jones potential does not describe BCC zones, whereas Morse potential
is applicable if the potential well is wide enough, but this demand leads to consider-
ation of additional coordinational spheres. Equilibrium BCC may be obtained from
equilibrium FCC by simple uniaxial compression, though the whole loading path is
stable, as BCC stability zones do not separate from FCC one. On the other hand,
stability region is non-convex (Fig. 5), so “Bain deformation” [17], which is accom-
plished by strain will provide an unstable zone between FCC and BCC equilibria.
Furthermore, it is shown that stability region for triangular lattice diminishes, espe-
cially in compression zone, if 3D perturbations are imposed (Fig. 6). MD simulation
is carried out for verification of theoretical results, and they prove to be in good
agreement.
Acknowledgments Authors are deeply grateful to prof. D. A. Indeitsev, prof. E. A. Ivanova and
prof. N. F. Morozov for useful discussions. This work was supported by grants of St. Peters-
burg Government (acts No.72, 25.10.2011 and No.80, 01.11.2011) and RFBR (No.11-01-00809-a,
No.12-01-31297 mol-a).
References
1. Macmillan, N.H.: The ideal strength of solids. In: Latanision, R., Pickens, J.R. (eds.) Atomistic
of Fracture, pp. 95–164. Plenum Press, New York (1983)
2. Lurie, A.I.: Nonlinear Theory of Elasticity. North-Holland, Amsterdam (1990)
3. Krivtsov, A.M., Morozov, N.F.: On mechanical characteristics of nanocrystals. Phys. Solid
State 44(12), 2260–2265 (2002)
4. Born, M., Huang, K.: Dynamical Theory of Crystal Lattices. Clarendon, Oxford (1954)
5. Wang, J., Li, J., Yip, S., Phillpot, S., Wolf, D.: Mechanical instabilities of homogeneous crystals.
Phys. Rev. 52(17B), 12627–12635 (1995)
6. Fu, Y.B., Ogden, R.W.: Nonlinear stability analysis of pre-stressed elastic bodies. Continuum
Mech. Thermodyn. 11, 141–172 (1999)
7. Milstein, F., Rasky, D.: Theoretical study of shear-modulus instabilities in the alkali metals
under hydrostatic pressure. Phys. Rev. 54(10), 7016–7025 (1996)
8. Schraad, M., Triantafyllidis, N.: Effects of scale size on media with periodic and nearly periodic
microstructures—II failure mechanisms. J. Appl. Mech. 64, 762–771 (1997)
9. Elliott, R.S., Shaw, J.A., Triantafyllidis, N.: Stability of pressure-dependent, thermally-induced
displacive transformations in bi-atomic crystals. Int. J. Solids Struct. 39, 3845–3856 (2002)
10. Tovstik, P.E., Tovstik, T.P.: On the 2D graphite layer model. Vestnik St. Petersburg University:
Math. 3, 134–143 (2009) (in Russian)
11. Dmitriev, S.V., Baimova, Yu.A., Savin, A.V., Kivshar’, Yu.S.: Stability range for a flat graphene
sheet subjected to in-plane deformation. JETP Lett. 93(10), 571–576 (2011)
12. Wallace, D.C., Patrick, J.L.: Stability of crystal lattices. Phys. Rev. 137(1A), 152–160 (1965)
13. Krivtsov, A.M.: Deformation and Fracture of Solids with Microstructure. Fizmatlit, Moscow
(2007) (in Russian)
14. Podolskaya, E.A., Panchenko, A.Yu., Krivtsov, A.M.: Stability of 2D triangular lattice under
finite biaxial strain. Nanosyst. Phys. Chem. Math. 2(2), 84–90 (2011)
15. Podolskaya, E.A., Panchenko, A.Yu., Krivtsov, A.M., Tkachev, P.V.: Stability of ideal infinite
2D crystal lattice. Doklady Phys. 57(2), 92–95 (2012)
16. Podolskaya, E.A., Krivtsov, A.M., Panchenko, A.Yu.: Investigation of stability and structural
transition in FCC lattice under finite strain. Vestnik St. Petersburg University: Math. 3, 123–128
(2012) (in Russian)
17. Bain, E.C.: The nature of martensite. Trans. Am. Inst. Min. Metall. Eng. 70, 25–46 (1924)
18. Berinskii, I.E., Dvas, N.G., Krivtsov, A.M., Kudarova, A.M., Kuzkin, V.A., Le-Zakharov, A.A.,
Loboda, O.S., Neigebauer, I.I., Podolskaya, E.A.: Theoretical Mechanics. Elastic Properties of
Monoatomic and Diatomic Crystals. In: Krivtsov, A.M. (ed.) St. Petersburg State Polytechnical
University, St. Petersburg (2009) (in Russian)
Mathematical Modeling of Phenomena Caused
by Surface Stresses in Solids
Yuriy Povstenko
Abstract Interfacial region between two bulk phases and the transition region near
the line of contact of three media are considered as a two-dimensional and one-
dimensional continuum, respectively. A survey of works on mathematical modeling
of phenomena in such systems is presented. The equation of the linear momentum
balance for an interface generalizes the classical Laplace equation and that for a
contact line generalizes the Young equation of the capillarity theory. The influence
of nonuniform surface tension on the stress field in an infinite cylinder is investigated.
The anisotropy of wetting is discussed and explained on the basis of the generalized
Young equation taking into account the tensor character of surface stresses. Several
applications of the results in the theory of surface defects are also discussed.
1 Introduction
Near the interface between two phases there arises a transition region which state
differs from the state of contacting media owing to different conditions of material-
particle interaction. The transition region has its own physical, mechanical and chem-
ical properties, and processes occurring in it differ from those occurring in the bulk.
The properties of this region influence the course of such processes as phase tran-
sition, crack propagation, corrosion, wetting, evolution of the defect structure, etc.
Small thickness of the interface region allows us to consider it as a distinct two-
dimensional phase (see Fig. 1), to use the methods of continuum mechanics and non-
equilibrium thermodynamics, and to formulate the corresponding two-dimensional
equations for the interface. In this approach, mathematical description of processes
occurring in the bulk phase consists in formulation and solution of some system
Y. Povstenko (B)
Institute of Mathematics and Computer Science, Jan Długosz University
in Czȩstochowa, Armii Krajowej 13/15, 42-200 Czȩstochowa, Poland
e-mail: j.povstenko@ajd.czest.pl

136 Y. Povstenko
Fig. 1 Schematic presenta-

tion of an interface between
two phases (left) and a line of
contact of three phases (right)
of differential (or more complicated) equations with certain boundary conditions

being the two-dimensional analogue of the corresponding three-dimensional equa-
tions [73].
The study of the processes of catalysis, adsorption, crystal growth, wetting, spread-
ing, flotation, properties of foams, films, and edges in crystals make it possible to con-
clude that near the line of contact of three media there also arises a transition region. In
this case not only the interfaces can be considered as distinct two-dimensional media,
but also the contact line can be considered as a distinct one-dimensional medium (see
Fig. 1), which state differs from the state of contacting three-dimensional bodies and
contacting two-dimensional surfaces. The one-dimensional analogues of the corre-
sponding three-dimensional equations can be formulated for the contact line.
The surface energy, surface stresses and surface tension as well as the line energy,
line stresses and line tension are the characteristic properties of the transition regions
and are significant in many mechanical and physical processes in solids (see [2, 5,
6, 12, 18, 33, 37, 49, 57, 63, 64, 73, 96, 97, 100, 103, 109, 110], among others).
In this paper, a survey of works on mathematical modeling of surface phenomena
in solids is presented. The equation of the linear momentum balance for an interface
between two bulk phases generalizes the classical Laplace equation and that for a
contact line of three phases generalizes the Young equation of the capillarity theory.
The influence of nonuniform surface tension on the stress field in an infinite cylinder
is investigated. The anisotropy of wetting is discussed and explained on the basis of
the generalized Young equation taking into account the tensor character of surface
stresses. Several applications of the results in the theory of surface defects are also
discussed.
2 General Balance Equations
In the case of material volume consisting of two three-dimensional bulk phases and
a two-dimensional interface between them (see Fig. 1) any extensive quantity Ψ
Mathematical Modeling of Phenomena 137
characterizing this volume can be written as the sum

Ψ = ρ1 ψ1 d V + ρ2 ψ2 d V + ρΣ ψΣ dΣ, (1)
V1 V2 Σ
where ρ is the mass density, ψ is the density of a quantity Ψ . In this case ψΣ is treated
either in accordance with Gibbs’ approach as the surface excess density associated
with the dividing surface [4, 10, 11] or presentation (1) can be substantiated by the
measure theory [23, 32, 59].
The local balance equation for a surface density ψΣ can be written as [24–26]
(we do not consider phase transition and wave propagation)
dψΣ
ρΣ = ρΣ θΣ − ∇ Σ · JΣ + n1 · J1 + n2 · J2 , (2)
dt
where θΣ denotes the production density, t is time, J1 and J2 are the values of bulk
fluxes of the quantity ψ at a surface, JΣ is the surface flux of the quantity ψΣ lying
in a tangent plane to Σ, n1 and n2 are the outer unit normals to the surface Σ, ∇ Σ
is the surface gradient operator:
∂
∇ Σ = aα , α = 1, 2. (3)
∂u α
Here u α are curvilinear coordinates on a surface, aα are the basis vectors in a tangent
plane, the summation convention is assumed.
Similarly, for material volume containing three bulk phases, three two-dimensional
interfaces, and one-dimensional line of their contact (see Fig. 1) any extensive quan-
tity Ψ can be written as the sum

Ψ = ρ1 ψ1 dV + ρ2 ψ2 dV + ρ3 ψ3 dV + ρ12 ψ12 dΣ
V1 V2 V3 Σ12

+ ρ13 ψ13 dΣ + ρ23 ψ23 dΣ + ρ L ψ L dL , (4)
Σ13 Σ23 L
and the general balance equation has the following form [73, 77]
dψ L
ρL = ρ L θ L − ∇ L · J L + N12 · J12 + N13 · J13 + N23 · J23 , (5)
dt
where Ni j is the vector lying in the tangential plane to the surface Σi j and normal
to the line L , Ji j denote the surface fluxes in the corresponding surfaces (i, j =
1, 2, 3, i < j), J L stands for the line flux of the quantity ψ L lying in the tangent to
the curve L, and ∇ L is the line gradient operator:
∂
∇L = λ . (6)
∂s
138 Y. Povstenko
Here λ is the unit tangential vector of a curve, s denotes the length of a curve.
3 The Generalized Laplace Equation
With the corresponding interpretation of the quantities ψΣ and ψ L and their sources
and fluxes, the balance equations for momentum, moment of momentum, energy,
entropy, etc. follow from the general balance equations (2) and (5) [73]. Identifying
the quantity ψ with the displacement velocity vector v, the production θ with the
body force vector f and the flux J with the stress tensor σ (with the opposite sign),
from Eq. (2) we obtain the equation of motion of an interface
dvΣ
ρΣ = ρΣ fΣ + ∇ Σ · σΣ − n1 · σ1 − n2 · σ2 . (7)
dt
Equation (7) generalizes the classical Laplace equation of the theory of capillarity
(see [73, 76]). Indeed, in the case
ρΣ = 0, σΣ = γΣ a, (8)
where γΣ is the surface tension, a is the unit tensor of a surface (the first fundamental
form of a surface), we obtain from (7):
∇ Σ γΣ + 2H γΣ n1 = n1 · σ1 + n2 · σ2 . (9)
Here H is the mean curvature of an interface. The component representation of

Eq. (9) reads
(1) (2)
σnn − σnn = 2H γΣ , (10)
(1) (2)
σnα − σnα = ∇α γΣ , α = 1, 2. (11)
Finally, if γΣ = const, then we arrive at the classical Laplace equation [48]
p2 − p1 = 2H γΣ , (12)
where p1 and p2 are the hydrostatic pressures in contacting bulk phases. Equa-
tion (11) is a basis of theoretical investigation of various physical phenomena caused
by heterogeneous surface tension including wetting of heterogeneous surfaces and
interaction of surface-active melts with metals.
4 Stresses due to Nonhomogeneous Surface Tension
The stressed state of a deformable elastic solid is defined by the equation of equilib-
rium
∇ · σ = 0, (13)
the Hooke law

σ = 2μe + λ(tr e) g , (14)
and by the geometrical relation
1
e= ∇u + u∇ , (15)
2
where u is the displacement vector, e is the strain tensor, λ and μ are the material
constants, g is the unit tensor. As the boundary conditions for Eqs. (13)–(15) we
choose the two-dimensional equation of equilibrium

∇ Σ · σΣ = n · σ Σ , (16)
the generalized Hooke law
σΣ = γΣ a + 2μΣ eΣ + λΣ (tr eΣ ) a (17)
with λΣ and μΣ being the surface material constants. The designation |Σ is used for
a boundary value of bulk functions.
Rigorous analysis carried out in [30, 31, 62] shows that linearization of general
stress-strain relations for two-dimensional elastic materials in the presence of residual
stress yields the more precise constitutive equation
σΣ = γΣ a + 2 (μΣ − γΣ ) eΣ + (λΣ + γΣ ) (tr eΣ ) a + γΣ ∇ Σ uΣ . (18)
The two-dimensional geometrical relation has the following form
1
eΣ = (∇ Σ uΣ + uΣ ∇ Σ ) , (19)
2
where the symbol indicates the projection on a tangential plane. The nonseparating
condition is also assumed:
uΣ = uΣ . (20)
In the paper [86], the influence of a surface tension gradient on the stress field in
a half-space with the surface tension γΣ being a function of the radial coordinate
140 Y. Povstenko
Fig. 2 Nonuniform surface

tension at the boundary of a
half-space (left) and a cylinder
(right)

γ 2, 0 ≤ r < R,
γΣ (z) = (21)
γ 1, R < r < ∞.
We present the components of the bulk stress tensor at the surface z = 0 (neglecting
the elastic moduli of the surface):

2(γ1 − γ2 ) R 2 K(k) R 2 E(k)
σrr = ν − (1 − ν) 2 − ν + (1 − ν) 2 ,
z=0 π r r+R r r−R
(22)

2(γ1 − γ2 ) R 2 K(k) R 2 E(k)
σθθ = 1 + (1 − ν) 2 − 1 − (1 − ν) 2 ,
z=0 π r r+R r r−R
(23)
where K(k) and E(K √ ) are the complete elliptic integrals of the first and second kind,
respectively, k = 2 r R/(r + R), and ν is the Poisson ratio.
In this paper, we investigate the influence of nonuniform surface tension on the
stress field in an infinite circular cylinder of radius R with surface tension (see Fig. 2)

γ1 , −∞ < z < 0,
γΣ (z) = (24)
γ2 , 0 < z < ∞.
Such a cylinder represents one type of cylindrical Janus particles [108]. The compo-
nents of the stress tensor can be expressed in terms of biharmonic Love’s function

∂ ∂2
σrr = ν∇ 2 − L, (25)
∂z ∂r 2

∂ ∂2
σr z = (1 − ν) ∇ 2 − 2 L , (26)
∂r ∂z

∂ 1 ∂
σθθ = ν∇ 2 − L, (27)
∂z r ∂r

∂ ∂2
σzz = (2 − ν) ∇ − 2 L ,
2
(28)
∂z ∂z
where
∇ 4 L = 0. (29)
The exponential Fourier transform with respect to the spatial coordinate z leads to
[101]

L = A(ξ )I0 (r ξ ) + B(ξ )r ξ I1 (r ξ ). (30)
Here the tilde denotes the integral transform, ξ is the transform variable, In (r ) is the
modified first-kind Bessel function of order n.
In the following we neglect the surface material constants. In this case the coef-
ficients A(ξ ) and B(ξ ) are determined from the boundary conditions:
γΣ
r = R : σrr = − , (31)
R
dγΣ
r = R : σr z =− (32)
dz
or, taking into account (24),
r = R : σr z = −(γ2 − γ1 ) δ(z), (33)

γ1 γ2 − γ1
r = R : σrr = − − H (z), (34)
R R
where δ(z) is the Dirac delta function and H (z) is the Heaviside step function. We
will restrict ourselves to the additional stressed state, and instead of the boundary
condition (34) the following boundary condition
γ 2 − γ1
r = R : σrr = − H (z) (35)
R
will be considered.
The Fourier transforms of components σrr and σr z read

1
σrr = iξ
3
A(ξ ) I0 (r ξ ) − I1 (r ξ )
rξ

+ B(ξ ) [(1 − 2ν) I0 (r ξ ) + r ξ I1 (r ξ )] , (36)
σr z = ξ 3 {A(ξ )I1 (r ξ ) + B(ξ ) [r ξ I0 (r ξ ) + 2(1 − ν)I1 (r ξ )]} .

(37)
Taking into account the boundary conditions (33) and (35) and the following formula
for the exponential Fourier transform of the Heaviside function
∞
H (z) eiξ z dz = 2π δ+ (ξ ), (38)
−∞
where
1 1
δ+ (ξ ) = δ(z) − , (39)
2 2πiξ
142 Y. Povstenko
we obtain the system of equations for determining the integration constants A(ξ )
and B(ξ ):

1
A(ξ ) I0 (Rξ ) − I1 (Rξ ) + B(ξ ) [(1 − 2ν) I0 (Rξ ) + Rξ I1 (Rξ )]
Rξ (40)
γ2 − γ1
= −2π δ+ (ξ ),
i Rξ 3
γ2 − γ1
A(ξ )I1 (Rξ ) + B(ξ ) [Rξ I0 (Rξ ) + 2(1 − ν)I1 (Rξ )] = − (41)
ξ3
having the solution

γ2 − γ1
A(ξ ) = − R [(1 − 2ν) I0 (Rξ ) + Rξ I1 (Rξ )]
ξ 2 D(Rξ )

+ 2πi [Rξ I0 (Rξ ) + 2(1 − ν) I1 (Rξ )] δ+ (ξ ) , (42)
γ 2 − γ1
B(ξ ) = {[Rξ I0 (Rξ ) − I1 (Rξ )] + 2πiξ I1 (Rξ ) δ+ (ξ )} , (43)
ξ 3 D(Rξ )
where
D(Rξ ) = R 2 ξ 2 + 2(1 − ν) I12 (Rξ ) − R 2 ξ 2 I02 (Rξ ). (44)
The stress components have the following form

1 ∞ sin(zξ ) ξ 3 A(ξ ) I (r ξ ) − 1 I (r ξ )
σrr = 2π −∞ 0 rξ 1

+ξ 3 B(ξ ) [(1 − 2ν)I0 (r ξ ) + r ξ I1 (r ξ )] dξ, (45)
∞
1
σr z = cos(zξ ) ξ 3 A(ξ ) I1 (r ξ ) + ξ 3 B(ξ ) [r ξ I0 (r ξ ) + 2(1 − ν) I1 (r ξ )] dξ,
2π −∞
(46)
∞
1 1
σθθ = sin(zξ ) ξ 3 A(ξ ) I1 (r ξ ) + (1 − 2ν) ξ 3 B(ξ )I0 (r ξ ) dξ, (47)
2π −∞ rξ
∞
1
σzz = − sin(zξ ) ξ 3 A(ξ ) I0 (r ξ ) + ξ 3 B(ξ ) [2(2 − ν)I0 (r ξ ) + r ξ I1 (r ξ )] dξ.
2π −∞
(48)
Dependence of stress components σ̄i j = Rσi j /(γ2 − γ1 ) on coordinate z is depicted

in Figs. 3–6.
Solution to the similar problem with
Fig. 3 Dependence of the

stress component σrr on the
coordinate z(r = 0.9R)

stress component σr z on the

γ 2, 0 ≤ |z| < a,
γΣ (z) = (49)
γ 1, a < |z| < ∞,
can be easily obtained by superposition of two appropriate foregoing solutions.

Extensive literature testifies to the influence of heterogeneous surface tension on
various physical, mechanical and chemical processes in solids (see, for example, [20,
22, 27, 58, 72, 104, 106, 111]). In [13, 14, 70, 88, 89] glass beads were studied
144 Y. Povstenko

stress component σθ θ on the

stress component σzz on the
exhibiting hydrophilic properties on one hemisphere and hydrophobic properties on

the other and the forces acted on the “Janus Bead” placed at the water-oil interface
were analyzed. Recent investigations of properties of Janus particles are presented
in [28, 40, 90].
When surface-active melt (indium, lead, tin, gallium and so on) interacts with
metal (iron, iron-silicon alloy, copper), a zone with a high dislocation density arises
in the surface layer of the metal [43–45]. The direct observation of a fusible metal
drop spreading over a foil in the column of an electron microscope [46] and theoretical
consideration [9, 86, 87] showed that the formation of dislocations in the vicinity
of the surface active melt perimeter was connected with nonuniformity of surface
tension. This conclusion is strongly supported by the correlation between difference
in surface tension of nonwetted and wetted parts of a metal and dislocation density.
For example, the surface tension of copper is estimated to be γ1 = 1.50 N/m [116]. At
350 ◦ C lead and bismuth can form a continuous series of alloys. The surface tension
Cu/Pb
of copper wetted by melted lead is equal to γ2 = 0.39 N/m, the surface tension
Cu/Bi
of copper wetted by melted bismuth is equal to γ2 = 0.28 N/m [92]. In the first
Cu/Pb
case γ1 −γ2 = 1.11 N/m and the dislocation density in the vicinity of the wetting
perimeter is ρ = 2.8 × 1012 m−2 , in the second case γ1 − γ2
Cu/Bi
= 1.22 N/m and
the dislocation density in the vicinity of the wetting perimeter is ρ = 9.3 × 1012 m−2
[9, 87]. Similar results were obtained for various percentage of lead and bismuth in
alloy (see also [42–45] and references therein).
5 The Generalized Young Equation
Now we analyze the equation of motion of a material line dividing three volume
phases and three two-dimensional surface phases. From Eq. (5) we get [73, 75, 76]
dv L
ρL = ρ L f L + ∇ L · σ L − N12 · σ12 − N13 · σ13 − N23 · σ23 . (50)
dt
The obtained equation generalizes the classical Young equation of the theory of
capillarity. Indeed, if
ρ L = 0, σ L = γ L λ ⊗ λ, σ12 = γ12 a12 , σ13 = γ13 a13 , σ23 = γ23 a23 , (51)
then we obtain
∂γ L
N12 γ12 + N13 γ13 + N23 γ23 − λ − kγ L τ = 0, (52)
∂s
where k is the first curvature of the contact line, τ is the principle normal vector of
the wetting perimeter.
In the case of axial symmetry we get [95]
γL
γ23 cos ϑ = γ13 − γ12 − cos ϕ, (53)
R
where R is a radius of the base of a drop, ϕ is the angle of inclination of the surface
Σ23 at a contact line, ϑ is the contact angle. For ϕ = 0 we arrive at the equation
146 Y. Povstenko
γL
γ23 cos ϑ = γ13 − γ12 − (54)
R
obtained in [99, 107]. For a plane interface and γ L = 0 the classical Young equation
[115] for the contact angle ϑ follows from (54):
γ23 cos θ = γ13 − γ12 . (55)
Neglecting γ L in (52) we obtain the so-called Neumann triangle equation [69]
N12 γ12 + N13 γ13 + N23 γ23 = 0. (56)
A large number of investigations have been devoted to the experimental study

and theoretical description of the processes of wetting and spreading. The shape of
a drop lying or spreading on a solid surface can deviate from the axial symmetry
which was observed in many experiments. An anisotropy of the surface being wetted
can be created artificially by introducing various inhomogeneities. Often a drop of
the liquid has a planar elliptic shape with the longer axis oriented in the direction
of the deformation. In wetting with tin a surface consisting of ordered portions of
pyroceramic of square shape on a molybdenum base the perimeter of wetting has the
shape of an octahedron with rounded corners [66, 67]. A comprehensive survey of
early studies of wetting and spreading anisotropy can be found in [83]. This subject is
still attracting considerable interest (see [15–17, 61, 68, 71, 105, 112–114, 117] and
references therein). The anisotropy of wetting and spreading can be explained and
understood on the basis of the generalized Young equation (50) taking into account
the tensor character of surface stresses σ12 and σ13 .
6 Cosserat Surfaces and Lines
The inherent laws of development of the mechanics of continua and the extending
fields of applications of the theory have led to construction of theories with couple
stresses. Kinematics of a Cosserat continuum is described by two independent vec-
tors: a displacement vector u and a rotation vector ω [19]. A couple-stress tensor μ
appears in such media parallel with a stress tensor σ . Cosserat surfaces and lines are
used in the theories of shells, plates and rods [8, 29, 65, 73, 93] (see also the survey
[7] and references therein).
Cosserat surfaces and lines are also used in the theory of capillarity [84]. Great
attention is paid to systems in which moment effects play an important role. The first
type of such systems are microemulsions [21, 36, 39, 51, 52, 55, 56, 60, 74, 91, 94,
102]. Presence of surfactants and cosurfactants results in very low (nearly vanishing)
surface tension, hence bending effects are of particular significance. On the other
hand, bending properties are due to nonsymmetrical structure of surfactants adsorbed
at an oil-water interface and an imbalance between hydrophile-water and lipophile-
oil interaction. The second type of such systems are lipid bilayer membranes in
which the hydrophilic polar heads are pointing toward the aqueous medium and
the hydrophobic ends of the hydrocarbon chains are pointing toward the interior of
the film [34, 35, 38, 47]. The structure asymmetry of lipid bilayer due to different
lipid compositions of the two constituent monolayers, the structure asymmetry of
lipid monolayer itself, and the influence of different environments on two sides of
the bilayer lead to considerable moment effects which demand to take couples into
account.
The classical equations of the theory of capillarity—the Laplace equation and
the Young equation—represent balances of forces for a two-dimensional surface
separating two bulk phases and for one-dimensional line separating three bulk and
three surface phases. In the case of a Cosserat interface, the linear momentum balance
should be supplemented with moment-of-momentum balance equation
dwΣ
βΣ ρΣ = ρΣ mΣ − (εΣ : σΣ ) n1 + εΣ ·σΣ ·n1 + ∇ Σ ·μΣ − n1 ·μ1 −n2 ·μ2 ,
dt
(57)
whereas in the case of a Cosserat contact line, the corresponding equation reads
dw L
βL ρL = ρ L m L −λ·σ L ×λ+∇ L ·μ L −N12 ·μ12 −N13 ·μ13 −N23 ·μ23 , (58)
dt
where w is the rotation velocity vector, m denotes the body couple vector, εΣ stands
for the surface alternating tensor (the surface Levi-Civita tensor), the coefficient β
does not depend on time and is connected with the moment of inertia. The interested
reader is also referred to [1, 50, 73, 84, 98].
7 Surface Imperfections
Experimental studies show that plastic deformation of surface layers of material

begins earlier than that of the bulk [3, 41]. Many authors observed drastic changes
of dislocation density near the interface layer [3, 41]. The dislocation velocity in
surface layers exceeds that in the bulk [3, 53, 54]. Moreover, the grain boundaries
and surface layers of material have their own defect structure which differs from that
in the bulk [3, 41].
The basic equations for surface dislocations and disclinations were obtained in [78,
80–82, 85]. The surface dislocation density tensor αΣ and the surface disclination
p
density tensor θΣ are defined as incompatibility of the plastic strain tensor γΣ and
p
the plastic bend-twist tensor κΣ of a Cosserat surface
∗
p p p ×
αΣ = −∇ Σ × γΣ + εΣ · b · γΣ − κΣ × a , (59)
p p
θΣ = −∇ Σ × κΣ + εΣ · b · κΣ , (60)
148 Y. Povstenko
where a and b are the first and second fundamental forms of a surface, the asterisk
denotes the transpose of a tensor.
The surface dislocation flux JΣ and the surface disclination flux IΣ have the
following form
p
dγΣ p p ∗
JΣ = − ∇ Σ vΣ + wΣ × a , (61)
dt
p
dκΣ p
IΣ = − ∇ Σ wΣ . (62)
dt
The balance equations for densities of surface defects with taking into account the
interaction with the bulk phases read
dαΣ ∗
×
= −∇ Σ × JΣ + εΣ · b · JΣ − IΣ × a + n1 × J1 + n2 × J2 , (63)
dt
dθΣ
= −∇ Σ × IΣ + εΣ · b · IΣ + n1 × I1 + n2 × I2 . (64)
dt
Nonlinear generalization of the theory of surface imperfections was obtained in [79,
80, 82, 85] using the methods of non-Riemannian differential geometry.
References
1. Abe, S., Sheridan, J.T.: Curvature correction model of droplet profiles. Phys. Lett. A 253,
317–321 (1999)
2. Adamson, A.W., Gast, A.P.: Physical Chemistry of Surfaces, 6th edn. Wiley, New York (1997)
3. Alekhin, V.P.: Physics of Strength and Plasticity of the Surface Layers of Materials. Nauka,
Moscow (1983) (in Russian)
4. Alexander, J.I.D., Johnson, W.C.: Thermomechanical equmbrium in solid-fluid systems with
curved interfaces. J. Appl. Phys. 58, 816–824 (1985)
5. Altenbach, H., Eremeev, V.A.: On the shell theory on the nanoscale with surface stresses. Int.
J. Eng. Sci. 49, 1294–1301 (2011)
7. Altenbach, J., Altenbach, H., Eremeyev V.A.: On generalized Cosserat-type theories of plates
and shells: a short review and bibliography. Arch. Appl. Mech. 80, 73–92 (2010)
8. Antman, S.S.: Nonlinear Problems of Elasticity, 2nd edn. Springer, New York (2005)
9. Borgardt, A.A., Krishtal, M.A., Povstenko, Y.Z., Katsman, A.V.: About the mechanism of
nucleation of dislocations during spreading of fusible melts on a metal surface. Dokl. Acad.
Sci. Ukr. SSR, Ser. A 2, 22–25 (1989) (in Russian)
10. Boruvka, L., Neumann, A.W.: Generalization of the classical theory of capillarity. J. Chem.
Phys. 66, 5464–5476 (1977)
11. Buff, F.P.: Curved fluid interfaces. I. The generalized Gibbs-Kelvin equation. J. Chem. Phys.
25, 146–153 (1956)
12. Cammarata, R.C.: Surface and interface stress effects in thin films. Prog. Surf. Sci. 46, 1–38
(1994)
13. Casagrande, C., Fabre, P., Raphaël, E., Veyssié, M.: Janus beads: realization and behavior at
water/oil interfaces. Europhys. Lett. 9, 251–255 (1989)
14. Casagrande, C., Veyssié, M.: Janus beads—realization and 1st observation of interfacial prop-
erties. C. R. Acad. Sci. Paris 306, 1423–1425 (1988)
15. Chatain, D.: Anisotropy of wetting. Annu. Rev. Mater. Res. 38, 45–70 (2008)
16. Chen, Y., He, B., Lee, J., Patankar, N.A.: Anisotropy in the wetting of rough surfaces. J.
Colloid Interface Sci. 281, 458–464 (2005)
17. Chung, J.Y., Youngblood, J.P., Stafford, C.M.: Anisotropic wetting on tunable micro-wrinkled
surfaces. Soft Matter 3, 1163–1169 (2007)
18. Chizmadzhev, Y.A., Markin, V.S., Tarasevich, M.P., Chirkov, Y.G: Macrokinetics of Processes
in Porous Media. Nauka, Moscow (1971) (in Russian)
19. Cosserat, E., Cosserat, F.: Théorie des corps déformables. Paris, Hermann (1909)
20. De Coninck, J., Voué, M.: Spreading on heterogeneous substrates. Oil Gas Sci. Technol. 56,
97–104 (2001)
21. De Gennes, P.G., Taupin, C.: Microemulsions and the flexibility of oil/water interfaces. J.
Phys. Chem. 86, 2294–2304 (1982)
22. Elwing, H., Welin, S., Askendal, A., Nilsson, U., Lundström, I.: A wettability gradient method
for studies of macromolecular interaction at the liquid/solid interface. J. Colloid Interface Sci.
119, 203–210 (1987)
23. Fisher, G.M.C., Leitman, M.J.: On continuum thermodynamics with surfaces. Arch. Ration.
Mech. Anal. 30, 225–262 (1968)
24. Ghez, R.: A generalized Gibbsian surface. Surf. Sci. 4, 125–140 (1966)
25. Ghez, R.: Equilibre méchanique et de forme de petits cristaux. Helv. Phys. Acta 41, 287–309
(1968)
26. Ghez, R.: Irreversible thermodynamics of a stationary interface. Surf. Sci. 20, 326–334 (1970)
27. Good, R.J., Kvikstad, J.A., Bailey, W.O.: Hysteresis of contact angles: the effect of elongation
and roughness. Appl. Polym. Symp. 16, 153–164 (1971)
28. Granick, S., Jiang, S., Chen, Q.: Janus particles. Phys. Today 62, 68–69 (2009)
29. Green, A.E., Naghdi, P.M., Wainwright, W.L.: A general theory of a Cosseart surface. Arch.
Ration. Mech. Anal. 20, 287–308 (1965)
Mech. Anal. 57, 291–323 (1975)
31. Gurtin, M.E., Murdoch, A.I.: Addenda to our paper a continuum theory of elastic material
surfaces. Arch. Ration. Mech. Anal. 57, 389–390 (1975)
32. Gurtin, M.E., Williams, W.O.: An axiomatic foundation for continuum thermodynamics.
Arch. Ration. Mech. Anal. 26, 83–117 (1967)
33. Haiss, W.: Surface stress of clean and adsorbate-covered solids. Rep. Prog. Phys. 64, 591–648
(2001)
34. Helfrich, W.: Elastic properties of lipid bilayers: theory and possible experiments. Z. Natur-
forsch. C: Biosci. 28, 693–703 (1973)
35. Helfrich, W.: Blocked lipid exchange in bilayers and its possible influence on the shape of
vesicles. Z. Naturforsch. C: Biosci. 29, 510–515 (1974)
36. Hu, C.: Equilibrium of an microemulsion that coexists with oil or brine. Soc. Petrol. Eng. J.
23, 829–847 (1983)
37. Ibach, H.: The role of surface stress in reconstruction, epitaxial growth and stabilization of
mesoscopic structures. Surf. Sci. Rep. 29, 195–263 (1997)
38. Janmey, P.A., Kinnunen, P.K.J.: Biophysical properties of lipids and dynamic membranes.
Trends Cell Biol. 16, 538–548 (2006)
39. Kawabata, Y., Seto, H., Nagao, M., Takeda, T.: Pressure effects on bending elasticities of
surfactant monolayers in a ternary microemulsion composed of aerosol-OT/D2 O/decane. J.
Chem. Phys. 127, 044705-1–044705-9 (2007)
40. Kaufmann, T., Gokmen, M.T., Rinnen, S., Arlinghaus, H.F., Prez, F.D., Ravoo, B.J.: Bifunc-
tional Janus beads made by “sandwich” microcontact printing using click chemistry. J. Mater.
Chem. 22, 6190–6199 (2012)
41. Kramer, I.R.: Influence of the surface layer on the plastic flow. Deformation of aluminium
single crystals. Trans. Met. Soc. AIME 233, 1462–1467 (1965)
150 Y. Povstenko
42. Krishtal, M.A.: Effect of surface-active melts on the structure and properties of metals. Mater.
Sci. 24, 325–329 (1988)
43. Krishtal, M.A., Borgardt, A.A., Katsman, A.V., Loshkarev, P.V.: On the origin and develop-
ment of the dislocation structure under diffusion interaction. Phys. Chem. Mech. Mater. 5,
15–20 (1988) (in Russian)
44. Krishtal, M.A., Borgardt, A.A., Loshkarev, P.V.: Acoustic emission during interaction between
iron and its alloys and surface-active melts. Dokl. Acad. Sci. USSR 267, 626–629 (1982) (in
Russian)
45. Krishtal, M.A., Borgardt, A.A., Loshkarev, P.V.: Formation of dislocations and acoustic emis-
sion during interaction of ferro-alloys with surface-active melts. Phys. Met. Metall. Sci. 56,
587–592 (1983) (in Russian)
46. Krishtal, M.A., Loshkarev, P.V., Borgardt, A.A.: On conditions of appearance of nucleat-
ing crack during liquid-metal brittleness. In: Bekrenev, A.N. (ed.) Mechanisms of Dynamic
Deformation of Metals, pp. 131–134. Kuibyshev Polytechnical Institute Press, Kuibyshev
(1986) (in Russian)
47. Kurtisovski, E., Taulier, N., Ober, R., Waks, M., Urbach, V.: Molecular origin of model
membrane bending rigidity. Phys. Rev. Lett. 98, 258103-1–258103-4 (2007)
48. Laplace, P.S.: Mécanique céleste, Supplément au X livre. Sur l’action capillaire. Courcier,
Paris (1805)
49. Linford, R.G.: Surface thermodynamics of solids. Solid State Surf. Sci. 2, 1–152 (1973)
50. Ljunggren, S., Ericksson, J.C., Kralchevsky, P.A.: Minimization of the free energy of arbitrary
curved interfaces. J. Colloid Interface Sci. 191, 424–441 (1997)
51. Martínez, H., Chacón, E., Tarazona, P., Bresme, F.: The intrinsic interfacial structure of ionic
surfactant monolayers at water-oil and water-vapour interfaces. Proc. R. Soc. A 467, 1939–
1958 (2011)
52. Mehta, S.K., Kaur, G.: Microemulsions: thermodynamic and dynamic properties. In: Tadashi,
M. (ed.) Thermodynamics, pp. 381–406. InTech, New York (2011)
53. Milevsky, L.S., Smolsky, I.L.: Change in the mobility of dislocations during migration to a
surface in a crystal with high Peierls barrier. Phys. Solid (Fizika Tverdogo Tela) 16, 1028–1031
(1974) (in Russian)
54. Milevsky, L.S., Smolsky, I.L.: Mobility of dislocations generated by internal sources in crystals
with high Peierls barrier. In: Startsev, V.I. (ed.) Dynamics of Dislocations, pp. 30–36. Naukova
Dumka, Kiev (1975) (in Russian)
55. Miller, C.A.: Interfacial bending effects and interfacial tensions in microemulsions. J. Dis-
persion Sci. Technol. 6, 159–173 (1985)
56. Miller, C.A., Neogi, P.: Thermodynamics of microemulsions: combined effects of dispersion
entropy of drops and bending energy of surfactant films. AIChE J. 26, 212–220 (1980)
57. Miller, C.A., Neogi, P.: Interfacial Phenomena. Equilibrium and Dynamic Effects, 2nd edn.
CRC Press, Boca Raton (2008)
58. Mironova, M.L., Botvinkin, O.K.: The role of surface tension in arising of stresses under
skeletonization of two-phase natroborosilicate glasses. In: Yeremenko, V.N. (ed.) Physical
Chemistry of Surface Phenomena under High Temperatures, pp. 226–230. Naukova Dumka,
Kiev (1971) (in Russian)
59. Moeckel, G.P.: Thermodynamics of an interface. Arch. Ration. Mech. Anal. 57, 255–280
(1975)
60. Monkenbusch, M., Holderer, O., Frielinghaus, H., Byelov, D., Allgaier, J., Richter, D.: Bend-
ing moduli of microemulsions; comparison of results from small angle neutron scattering and
neutron spin-echo spectroscopy. J. Phys.: Condens. Matter 17, S2903–S2909 (2005)
61. Morita, M., Koga, T., Otsuka, H., Takahara, A.: Macroscopic-wetting anisotropy on the line-
patterned surface of fluoroalkylsilane monolayers. Langmuir 21, 911–918 (2005)
62. Murdoch, A.I.: Thermomechanical theory of elastic-material interfaces. Q. J. Mech. Appl.
Math. 29, 245–275 (1976)
63. Murr, L.E.: Interfacial Phenomena in Metals and Alloys. Addison-Wesley, London (1975)
64. Müller, P., Saúl, A.: Elastic effects on surface physics. Surf. Sci. Rep. 54, 157–258 (2004)
65. Naghdi, P.M.: The theory of plates and shells. In: Flügge, S. (ed.) Handbuch der Physik, vol.
VIa/2, pp. 425–640. Springer, Heidelberg (1972)
66. Naidich, Yu.V., Kolesnichenko, G.A., Voitovich, R.P., Kostyuk, B.D., Gavrilyuk, G.I.: Wet-
ting of inhomogeneous solid surfaces by metal melts. In: Naidich, Yu.V. (ed.) Capillary and
Adhesive Properties of Melts, pp. 18–25. Naukova Dumka, Kiev (1987) (in Russian)
67. Naidich, Y.V., Voitovich, R.P., Kolesnichenko, G.A., Kostyuk, B.D.: Wetting of inhomoge-
neous solid surfaces by metal melts for the systems with an ordered arrangement parts of
different kinds. Phys. Chem. Mech. Surf. 2, 126–132 (1988) (in Russian)
68. Neuhaus, S., Spencer, N.D., Padeste, C.: Anisotropic wetting of microstructured surfaces as
a function of surface chemistry. Appl. Mater. Interfaces 4, 123–130 (2012)
69. Neumann, F.E.: Vorlesungen über die Theorie der Kapillarität. Teubner, Leipzig (1894)
70. Ondarçuhu, T., Fabre, P., Raphaël, E., Veyssié, M.: Specific properties of amphiphilic particles
at fluid interfaces. J. Phys. (Paris) 51, 1527–1536 (1990)
71. Park, S.-G., Moon, J.H., Jeon, H.C., Yang, S.-M.: Anisotropic wetting and superhydropho-
bicity on holographically featured 3D nanostructured surfaces. Soft Matter 8, 4567–4570
(2012)
72. Paterson, A., Fermigier, M.: Wetting of heterogeneous surfaces: influence of defect interac-
tions. Phys. Fluids 9, 2210–2216 (1997)
73. Podstrigach, Y.S., Povstenko Y.Z.: Introduction to the Mechanics of Surface Phenomena in
Deformable Solids. Naukova Dumka, Kiev (1985) (in Russian)
74. Pouchelon, A., Chatenay, D., Meunier, J., Langevin, D.: Origin of low interface tensions in
systems involving microemulsion phases. J. Colloid Interface Sci. 82, 418–422 (1981)
75. Povstenko, Y.Z.: The momentum balance equation at the line of contact of three media. Dokl.
Acad. Sci. Ukr. SSR, Ser. A 10, 45–47 (1980) (in Russian)
76. Povstenko, Y.Z.: Generalized Laplace and Young conditions of mechanical contact. Math.
Methods Phys.-Mech. Fields 16, 30–32 (1982) (in Russian)
77. Povstenko, Y.Z.: The general balance equation at an interface between two media and at a
line dividing three media. Math. Methods Phys.-Mech. Fields 17, 41–43 (1983) (in Russian)
78. Povstenko, Y.Z.: Dislocations and disclinations in a two-dimwensional continuum. Theory of
incompaticle strains. Dokl. Acad. Sci. Ukr. SSR, Ser. A 4, 35–37 (1985) (in Russian)
79. Povstenko, Y.Z.: Dislocations and disclinations in a two-dimwensional continuum. Connec-
tion with non-Riemannian geometry. Dokl. Acad. Sci. Ukr. SSR, Ser. A 6, 45–48 (1985) (in
Russian)
80. Povstenko, Y.Z.: Continuous theory of dislocations and disclinations in a two-dimensional
medium. J. Appl. Math. Mech. 49, 782–786 (1985) (in Russian)
81. Povstenko, Y.Z.: Analysis of motor fields in Cosserat continua of two and one dimensions
and its applications. Z. angew. Math. Mech. 66, 505–507 (1986)
82. Povstenko, Y.Z.: The mathematical theory of derects in a Cosserat continuum. Math. Methods
Phys.-Mech. Fields 27, 34–40 (1988) (English Transl.: J. Math. Sci. 62, 2524–2530 (1992))
83. Povstenko, Y.Z.: Anisotropy of wetting and spreading. Math. Methods Phys.-Mech. Fields
31, 8–16 (1990) (English Transl.: J. Math. Sci. 64, 890–898 (1993))
84. Povstenko, Y.Z.: Generalizations of Laplace and Young equations involving couples. J. Colloid
Interafce Sci. 144, 497–506 (1991)
85. Povstenko, Y.Z.: Connection between non-metric differential geometry and mathematical
theory of imperfections. Int. J. Eng. Sci. 29, 37–46 (1991)
86. Povstenko, Y.Z.: Theoretical investigation of phenomena caused by heterogeneous surface
tension in solids. J. Mech. Phys. Solids 41, 1499–1514 (1993)
87. Povstenko, Y.: Stresses due to heterogeneous surface tension in solids. Proc. Appl. Math.
Mech. 1, 193–194 (2002)
88. Raphaël, E.: Étalement de gouttes sur une surface bigarrée. C. R. Acad. Sci. Paris Sér. II 306,
751–754 (1988)
89. Raphaël, E.: Équilibre d’un "Grain Janus" á une interface eau/huile. C. R. Acad. Sci. Paris
Sér II 307, 9–12 (1988)
152 Y. Povstenko
90. Reculusa, S., Mingotaud, C., Duguet, E., Ravaine, S.: Dissymmetrical nanoparticles. In: Con-
tescu, C.I., Putuyera, K. (eds.) Dekker Encyclopedia of Nanoscience and Nanotechnology,
2nd edn, pp. 1093–1101. CRC Press, Boca Raton (2008)
91. Rekvig, L., Hafskjold, B., Smit, B.: Simulating the effect of surfactant structure on bending
moduli of monolayers. J. Chem. Phys. 120, 4897–4905 (2004)
92. Rostoker, W., McCaughey, J.M., Markus H.: Embrittlement by Liquid Metals. Reinhold Pub-
lishing Corporation, New York (1960)
93. Rubin, M.B.: Cosserat Theories: Shells, Rods and Points. Springer, New York (2000)
94. Ruckenstein, E.: An explanation for the unusual phase behaviour of microemulsions. Chem.
Phys. Lett. 98, 573–576 (1983)
95. Rusanov, A.I.: On the theory of wetting of solids. 5. Reduction of deformation effects to linear
tension. Colloid J. 39, 704–710 (1977) (in Russian)
96. Rusanov, A.I.: Phasengleichgewichte und Grenzflächenerscheinungen. Academie-Verlag,
Berlin (1978)
97. Rusanov, A.I.: Surface thermodynamics revisited. Surf. Sci. Rep. 58, 111–239 (2005)
98. Sagis, L.M.C.: Generalized curvature expansion for the surface internal energy. Physica A246,
591–608 (1997)
99. Shcherbakov, L.M., Ryazantsev, P.P.: About influence of energy of wetting perimeter on edge
conditions. In: Deryagin, B.V. (ed.) Research in Surface Forces, pp. 26–28. Nauka, Moscow
(1964) (English Transl.: Consult. Bureau, New York (1966))
100. Slattery, J.C., Sagis., L, Oh, E.-S.: Interfacial Transport Phenomena, 2nd edn. Springer, New
York (2007)
101. Sneddon, I.N.: Fourier Transforms. McGraw-Hill, New York (1951)
102. Surabhi, K., Op, K., Atul, N., Arun, G.: Microemulsions: developmental aspects. Res. J.
Pharm. Biol. Chem. Sci. 1, 683–706 (2010)
103. Sun C.Q.: Thermomechanical behavior of low-dimensional systems: the local bond average
approach. Prog. Mater. Sci. 54, 179–307 (2009)
104. Swain, P.S., Lipowsky, R.: Contact angles on heterogeneous surfaces: a new look at Cassie’s
and Wenzel’s laws. Langmuir 14, 6772–6780 (1998)
105. Timoshenko, V., Bochenkov, V., Traskine, V., Protsenko, P.: Anisotropy of wetting and spread-
ing in binary Cu-Pb metallic system: experimental facts and MD modeling. J. Mater. Eng.
Perform. 21, 575–584 (2012)
106. Tress, H.J.: Some distinctive contours worn on alumina-silica refractory faces by different
molten glasses: surface tension and the mechanism of refractory attack. J. Soc. Glass Technol.
38, 89T–100T (1954)
107. Veselovsky, V.S., Pertsov, V.N.: Attachment of bubbles to solid surfaces. Zhurn. Fiz. Khim.
(J. Phys. Chem.) 8, 245–259 (1936) (in Russian)
108. Walther, A., Müller, A.H.E.: Janus particles. Soft Matter 4, 663–668 (2008)
109. Wang, J., Huang, Z.P., Duan, H.L., Yu, S.W., Feng, X.Q., Wang, G.F., Zhang, W.X., Wang,
T.J.: Surface stress effect in mechanics of nanostructured materials. Acta Mech. Solida Sin.
24, 51–82 (2011)
110. Wang, J., Karihaloo, B.L., Duan, H.L., Huang, Z.P.: Importance of surface/interface effect to
properties of materials at nano-scale. In: Sadowski, T. (ed.) IUTAM Symposium on Multiscale
Modelling of Damage and Fracture Processes in Composite Materials, pp. 227–234. Springer,
Dordrecht, The Netherlands (2006)
111. Wu, J., Zhang, M., Wang, X., Li, S., Wen, W.: A simple approach for local contact angle
determination on a heterogeneous surface. Langmuir 27, 2705–2708 (2011)
112. Xia, D., Brueck, S.R.J.: Strongly anisotropic wetting on one-dimensional nanopatterned sur-
faces. Nano Lett. 8, 2819–2824 (2008)
113. Xia, D., He, X., Jiang, Y.-B., López, G.P., Brueck, S.R.J.: Tailoring anisotropic wetting prop-
erties on submicrometer-scale periodic grooved surfaces. Langmuir 26, 2700–2706 (2010)
114. Xia, D., Johnson, L.M., López, G.P.: Anisotropic wetting surfaces with one-dimensional and
directional structures: fabrication approaches, wetting properties and potential applications.
Adv. Mater. 24, 1287–1302 (2012)
115. Young, T.: An essay on the cohesion of fluids. Phil. Trans. Roy. Soc. 94, 65–87 (1805)
116. Zadumkin, S.N., Kumykov, V.K., Khokonov, K.V.: Surface tension of some high-melting solid
metals. In: Tavadze, F.N. (ed.) Physical Chemistry of Melt Surface, pp. 194–200. Mecniereba,
Tbilisi (1977) (in Russian)
117. Zhao, Y., Lu, Q., Li, M., Li, X.: Anisotropic wetting characteristics on submicrometer-scale
periodic grooved surface. Langmuir 23, 6212–6217 (2007)
Buckling of a Supported Annular Plate
with a Non-Euclidean Metric
Michael Schwarzbart and Alois Steindl
Abstract To create a new surface in a solid or a liquid material, energy needs

to be spent. On the atomistic level dangling bonds are formed, which tend to be
reconstructed accompanied by an excess surface energy. For thin structures like
graphene this energy can change the global shape of the structure drastically. There
are also examples of such a behaviour of thin structures in our every day’s life.
Considering for example the stretching of pliable plastic (garbage bag) past the yield
point [9, 14], or the effect of different growth in leaves [6]. In both cases the change of
the bonding configuration leads to an expanding edge and/or to wrinkled equilibrium
configurations, which are optimal from an energetic point of view. In this work the
effect of a free edge on the global behaviour of a circular graphene patch is studied
with an atomistic and a continuum mechanics approach.
1 Atomistic Model
Graphene is modeled by classical multi-body potentials called AIREBO [15] and

REBO [3] respectively. Both potentials are implemented in the molecular dynamics
simulator called LAMMPS [10]. These force fields have been used successfully
in the past for modelling the mechanical behaviour of graphene [8] and carbon
nantotubes [17].
M. Schwarzbart (B)
Department of Applied and Numerical Mechanics, University of Applied Sciences
Wiener Neustadt, Johannes Gutenberg-Straße 3, 2700 Wiener Neustadt, Austria
e-mail: michael.schwarzbart@fhwn.ac.at
A. Steindl
Institute of Mechanics and Mechatronics, Vienna University of Technology,
Wiedner Hauptstrasse 8-10 325/2,1040 Wien, Austria
e-mail: alois.steindl@tuwien.ac.at

156 M. Schwarzbart and A. Steindl
(a) (b)
0.15
AIREBO
armchair 0.10
energy [ eV ]
AIREBO
0.04 zigzag 0.05
edge energy γ [eV / Å]
0.03 REBO
0.02 armchair 0
0.01 REBO
0 zigzag − 0.05
− 0.01
− 0.02 − 0.10
− 0.03
− 0.04 − 0.15
− 0.005 0 0.005 0.010 0.015 − 0.005 0 0.005 0.010
strain [-] strain [-]
Fig. 1 Edge energy as a function of the elastic strain, where the edge stress is the slope in the
origin (a). Energy difference of the same unit cell with periodic boundary conditions in one and
two directions (b)
1.1 Edge Energy of Graphene
An arbitrary graphene edge is geometrically formed by a combination of armchair

and zigzag parts, but also the edge energy depends on the values of these two distinct
types of edges [7]. For this reason only armchair and zigzag edges are considered for
the computation of edge energies and stresses respectively. To get the edge energy for
a given unit cell two calculations are required, one with periodic boundary conditions
in two directions and one with periodicity in only one direction. The resulting energy
difference is the quantity characterising the effect of a free edge, which is depicted in
Fig. 1b. Figure 1a shows this energy difference, in consequence of the existence of a
free edge, scaled with the length of the considered edge. By changing the length of the
unit cell perpendicular to the free edge under inspection, an elastic strain is created.
The dependency of the energy difference and the edge energy on this strain can be
seen in Fig. 1, where the curves are shifted into the origin. A detailed description of the
energy computation can be found in [5], where a ab initio formulation is presented.
Results from this DFT analysis serve as a benchmark for the results presented in this
work.
For small strains (ε 1) the edge energy as a function of elastic strain is repre-
sented by the Taylor expansion

dγ
γ (ε) = γ(ε=0) + ε + O(ε2 ) ≈ γ0 + f ε, (1)
dε ε=0
Buckling of a Supported Annular Plate with a Non-Euclidean Metric 157
Table 1 Comparison of results obtained by DFT and molecular mechanics approach

Edge energy Edge stress
γ in eV /Å f in eV /Å
armchair zigzag armchair zigzag
DFT [5] 1.243 1.533 −2.64 −2.248
Molecular mechanics (REBO) 1.317 1.247 −1.061 −1.971
Molecular mechanics (AIREBO) 1.178 1.094 −1.128 −2.035
Fig. 2 Edge energies for two

distinct edge geometries, with
respect to the applied strain.
The inset shows the energy
difference for a unit cell with
one and two pbcs. With kind
permission from American
Physical Society, Phys. Rev.
B. 78, 073405 (2009), Fig. 3
with γ0 as the edge energy of the unstrained edge, and f as the edge stress. In
other words the slope of individual curves in Fig. 1a represents the edge stresses.
In Table 1 the results obtained for the two distinct types of edges and the used
force fields are given, as well as the DFT results from [5]. At first glance the results
are in good agreement with values obtained by ab initio methods. However the
energies and stresses for armchair edges are larger than for zigzag ones, which is
inconsistent with the results obtained by an ab initio computation. A comparison
of Figs. 1 and 2 apparently shows different slopes for armchair and zigzag edges
respectively. Interestingly enough even different basis sets used in DFT methods
lead to different values of the corresponding energies [11]. Nevertheless the order of
magnitude and the sign of edge energy and edge stress are correctly computed with
the two classical potentials.
1.2 Global Structure due to an Existing Edge
As a result of the changed bonding configuration the edge wants to elongate. The
tendency of increasing the period length is constrained by parts away from the edge.
Therefore a compressive stress localised at the edge occurs, independently from the
Fig. 3 Relaxed structure of a

graphene patch (1090 atoms)
without a substrate, calculated
with LAMMPS [10]
underlying geometry of the edge. Although the mismatch of the explicit values of both
models (classical and quantum mechanical), the underlying effect of compression
at the edge is represented by the molecular mechanics approach. The advantage
of classical potentials is the size of the manageable systems, which make them
appropriate to study larger graphene patches. Figure 3 shows the relaxed structure
of a graphene patch consisting of 1090 carbon atoms. The edge is a combination
of armchair and zigzag parts. However both distinct edge types show compressive
stresses, which results in an wavy out of plane displacement. In radial direction the
displacement decays away from the edge.
To investigate the influence of a substrate on the buckled configuration, an addi-
tional fixed graphene sheet interacting via the Van-der-Waals force was considered.
The resulting relaxed structure shows a higher wavenumber in tangential direction,
but a much smaller amplitude than the configuration shown in Fig. 3. For a deeper
understanding of the phenomenon it is necessary to compute the critical edge energy,
where the buckled and the flat configuration coexist. Formulating this problem in the
framework of continuum mechanics offers the possibility of stating an appropriate
stability problem. The aim of such an approach is to obtain the stability bound-
ary in the parameter plane. With this model the influence of the substrate, and the
corresponding critical values of the edge energy can be studied.
2 Continuum Non-Euclidean Plate Model
The absent of an external load makes it necessary to think of how to achieve the
compressive stress near the edge. One way is to prescribe an appropriate eigenstress
field, which was successfully used for studying the influence of residual stresses in
rolled strips [12]. For the present problem a suitable modification of the strain field
is used, to model the changed bonding configuration at the edge. This approach has
the advantage of its purely kinematic nature.
Measuring lengths on a surface is intrinsically tied to the metric tensor. The edge
expansion is modeled as a perturbation of the metric tensor of the undeformed elastic
plate gαβ . The new metric tensor

1 0 ε̂rr ε̂r ϕ
g̃αβ = gαβ + ĝαβ = + , (2)
0 r2 ε̂r ϕ ε̂ϕϕ
is called the target metric [9, 14]. The desired expansion of the edge can be realized
with only one additional term ε̂ϕϕ = 0, which gives the length of an infinitesimal
arc in tangential direction (r = const, dr = 0)

ds = 1 + ε̂ϕϕ r dϕ. (3)
The chosen target metric term ε̂ϕϕ = βr n ensures for large values of n, that the effect
of the elongated circumference is located just near the edge, and decays very rapidly
in radial direction. With this modification of the metric term the phenomenon of an
expanding edge (Sect. 1.2) is formulated in a continuum mechanics context.
Beyond measuring of lengths on surfaces the difference between the actual and the
target metric tensor defines the strain tensor. With the common assumptions for
the derivation of the nonlinear plate equations [2, 4] and the modified strain tensor
the Föppl-von-Kármán equations for a non-Euclidean annular plate

1 2
[w, w] + ε̂ϕϕ ,rr + ε̂ϕϕ ,r = 0,
Δ2 Φ + E (4a)
2 r
6 12
12γ 1 1
K Δ w − h [Φ, w] +
2
− =0 (4b)
1+w 1+w 1+w
can be obtained. The thin plate of thickness h is described as a linear elastic material
with Young’s modulus E, Poisson’s ratio μ and bending stiffness
Eh 3
K = .
12(1 − μ2 )
In Eq. (4) the Laplacean ΔA = A,rr + r1 A,r + r12 A,ϕϕ , the Monge-Ampére operator

b,r b,ϕϕ a,r a,ϕϕ a,r ϕ a,ϕ b,r ϕ b,ϕ
[a, b] = a,rr + 2 + b,rr + 2 −2 − 2 − 2
r r r r r r r r
(5)
and the definition of the stress function
1 1 1 1
σrr = Φ,r + 2 Φ,rr ; σϕϕ = Φ,ϕϕ ; σr ϕ = 2 Φ,ϕ − Φ,r ϕ
r r r r
in polar coordinates are used. The nonlinear foundation of the plate (addend with
γ in Eq. (4b)) is modeled in the style of the Van-der-Waals interaction, in order to
make the results comparable with molecular static ones. There is no external load,
but the attempt of the edge to increase its circumferential line due to the changed
metric term (Eq. (2)).
2.1 Dimensionless Equation
The solution Φ d of the inhomogeneous disk problem (w = 0 in Eq. √ (4a)) can

be found in closed form. Using the dimensionless variables w̄ = w/ K /Eh,
β̄ = β/(K /Eh R 2 ), γ̄ = γ /(K /R 4 ) and Φ̄ = (Φ d − Φ)/K , and linearising
(4b) around the solution of the disk problem leads, on dropping the bars, to the
linearised plate equation

1+n 1 1
Δ w=β2 2
Δw(1 − r ) −
n
w,r + 2 w,ϕϕ nr − γ 4 w.
n
(6)
n(2 + n) r r
In Eq. (6) the two dimensionless parameters β and γ are taken into account to the
second and to the fourth power respectively, for a better description of the interesting
parameter range [16]. Before solving this equation a scaling analysis of the energy
terms is given.
2.2 Scaling of the Energy Terms
To investigate the stability boundary in the parameter plane qualitatively the energy
terms are estimated. For an inplane displacement w = 0 the strain energy scales
like Ud ∼ β 2 Eh R 2 . For the solution of Eqs. (6) the ansatz (8) is used. With the
wavenumber m and the amplitude of the displacement at the edge ξ , the strain energy
for bending Ub ∼ ξ 2 K m 4 /R 2 , and for the nonlinear foundation U f ∼ ξ 2 γ R 2 can
be estimated. The relation between the amplitude of the displacement ξ and the
additional metric term β can be found from geometrical considerations.
The length of the sector Fig. 4 Rϕ 2α = R(1 + εϕϕ )π/m together with the angle
in the triangle in Fig. 4b α = arccos((Rϕ − ξ )/Rϕ ) gives the important relation
ξ 2 ∼ εϕϕ R 2 /m 2 ∼ β R 2 /m 2 . With the coexistence of the trivial and the buck-
(a) (b)
π
Fig. 4 Sketch of the trivial and the buckled configuration of the circular plate (a). Sector ( m ) in
er -direction (b)
led configuration at the stability boundary the corresponding energy terms can be
equalized Ud = Ub + U f which leads to

γ4
β= m2 + . (7)
m2
As discussed before the two parameters (β, γ ) are considered to the power of two
and four respectively. For a given wavenumber of the buckled configuration Eq. (7)
represents the stability boundary in the parameter plane. A discussion of this result
is given in the next section.
2.3 Stability Analysis
The displaced configuration of graphene, shown in Fig. 3, gives reason for looking
for solutions of Eq. (6) of the form
w(r, ϕ) = g(r ) cos mϕ, m ∈ N. (8)
Inserting the ansatz (8) into the partial differential equation (6) leads to a boundary
value problem for the displacement in radial direction. Boundary conditions for the
free edge and corresponding continuity conditions in the centre of the plate are
implied. For the numerical solution of this problem COLSYS [1] was used. For the
arbitrary exponent of the metric term n = 10 was used, to concentrate the stress
at the edge of the plate. The resulting envelope of the curves (Fig. 5) for different
wavenumbers represents the stability boundary. Wavenumbers m = 0 do not arise,
because no elongation of the circumference is accompanied with this mode. A mode
with wavenumber m = 1 corresponds to a rigid body rotation of the plate. Both modes
do not reduce the strain energy, which makes them not showing up in the resulting
parameter plane. With increasing foundation stiffness a larger metric coefficient is
necessary in order to reach the stability boundary, moreover the buckling mode
shows a higher waviness. Besides the result from the numerical computation Fig. 5a,
the result form scaling considerations Eq. (7) is given in Fig. 5b. Without solving
the corresponding boundary value problem just by estimating the energy terms the
qualitative form of the stability boundary was found. The effect of the chosen value
of the exponent n is not present in the scaling analysis. With increasing values of the
exponent n the curves are shifted to higher values of the metric coefficient (Fig. 5).
In [16] similar stability charts for in-plane loaded circular plates are given. Curves
of different modes of instability cross each other in the parameter plane. For example
at point (βcrit = 9.138, γcrit = 2.307) in Fig. 5a the modes m = 2 and m = 3 may
interact. A nonlinear analysis is necessary to study this interaction, and the buckling
behaviour in the vicinity of the bifurcation.
(a)
(b)
Fig. 5 Stability boundary in the dimensionless parameter plane, for different wavenumbers m. Set
of curves a are obtained from the numerical solution, whereas the curves b are the result of scaling
considerations Eq. (7)
Nonlinear Analysis
The solution from the linear analysis in a complex form w = g2 (r )(z 2 ei2ϕ +
z̄ 2 e−i2ϕ ) + g3 (r )(z 3 ei3ϕ + z̄ 3 e−i3ϕ ) is used as a Galerkin-ansatz for the bound-
ary value problem (4) with the leading nonlinear terms taken into account. Linear
unfolding parameters for the stiffness foundation λγ and the metric parameter λβ
respectively lead to the integral

10 w,
r w,ϕϕ
Δ2 w − [Φ, w] − (βcrit
2
+ λβ ) Δw(1 − r n ) − + 2 nr n
A 120 r r

−(γcrit
4
+ λγ ) 72w − 756w 2 + 4452w 3 δw d A = 0, (9)

which gives the real bifurcation equations z 2 z 22 + E z 32 − λ = 0, z3 E z 22 + F z32
−μ) = 0 with the new parameters λ = − Aλβ + Bλγ , μ = − Cλβ + Dλγ ,
E ≈ 3.4, F ≈ 7.2. Both coordinate systems are depicted in Fig. 6, where a path
through the parameter plane is given. In Fig. 6a the bifurcation diagram for this
specific path is shown. The z 2 and the z 3 solutions bifurcate subcritically at the cor-
responding stability boundaries. In between these two boundaries a mode interaction
occurs, but no stable secondary branch exists. Therefore the resonant fifth order terms
are not investigated.
(a) (b)
Fig. 6 Bifurcation diagram (a) for the path ϕ through the dimensionless parameter plane (b). The
coordinate system of the physical unfolding parameters (λβ , λγ ) as well as for the mathematical
parameters (μ, λ) is depicted
Fig. 7 In-plane stiffness is

modeled with springs (i j),
and bending stiffness with the
relative orientation between
neighboring facets (nα nα )
2.4 Numerical Simulation of the Post Buckling Behaviour
To investigate the post buckling behaviour beyond the stability boundary a discrete
model is used. The continuous plate is discretised by means of a triangular spring
network model, which is depicted in Fig. 7. For a network of equilateral triangles the
stretching and bending energies of the plate are represented by
√
3Eh 2 Eh 3 2
Us = ri j − l 0 and Ub = √ nα − nβ ,
4
ij
12 3(1 − μ2 ) αβ
respectively [13]. The expanding edge can be defined very intuitively by increasing
the undeformed lengths l0 of the corresponding springs forming the circumference.
110
m=14 m=20
40
m=8 m=12
35
m=7 m=11
30
m=6 m=11
25
m=6
20
m=2 m=5
0 5 11
Fig. 8 Chart with different post buckling configurations, computed with a discrete spring model
described in the text
A discussion of the equivalence between continuous metric term and discrete undi-
lated spring length can be found in [9].
In accordance with the continuous formulation of the target metric (2), the unde-
formed length of every individual spring needs nto be modified depending on its
orientation according to Δl0 (r, ϕ) = Δl0 (R) Rr | sin ϕ|. To relate the continuum
plate model with the discrete one, the undeformed length as a function of the metric
parameter is obtained by equalizing the circumference for both formulations. With
the aim of linking the mechanical models with the atomistic one, the strain energy
terms need to be considered. The extra energy due to the existing edge (Sect. 1.1)
is connected with the closed form of the strain energy of the circular disc. It turns
out, that for β ≈ 110 the continuous model has the same elastic energy than the
graphene patch in Fig. 3. A foundation stiffness of γ ≈ 11 represents the existence
of a substrate in the atomistic model. LAMMPS [10] was used to find equilibrium
configurations of the discrete spring model. Figure 8 shows the resulting configura-
tions, which are beyond the stability boundary. However without a foundation γ = 0
γ= 0 m= 11 γ= 11 m= 22
Fig. 9 Buckled configurations for appropriate prescribed undeformed spring length strictly at the
circumference
the discrete model shows a m = 2 mode, which contradicts the result given in Fig. 3.
The reason is, that even for large exponents of the additional metric term, the load
is continuous, and not localized strictly at the edge like in the atomistic model. The
tendency of increasing wavenumbers of the post buckling configurations for larger
stiffness parameters of the foundation can be clearly seen in Fig. 8. With the discrete
spring model at hand, it is possible to change only the unstrained spring length at
the circumference, leaving the others unchanged. This approach is very close to the
situation in the atomistic model. Again the undilated spring length at the edge is pre-
scribed in order to represent the energy of the graphene patch considered in Sect. 1.2.
In Fig. 9 the post buckling configurations for the desired parameters are depicted.
A comparison of Fig. 9 with Fig. 3 shows good agreement, which emphasises the
localised character of the edge stress in graphene.
3 Discussion
In our study, we have addressed the effect of an edge on the global equilibrium
configuration of a circular graphene patch. The correct magnitude and sign of the
edge energy as well as the edge stress are computed with a molecular static approach.
This compressive stress at the edge results in a wavy out of plane displacement, which
decays in radial direction. To analyse this problem in a systematic way, a stability
problem in a continuum mechanics framework is derived. The expanding edge which
results in compressive stresses is modeled by an appropriate disturbance of the metric
term. The stability boundary in the parameter plane shows points of mode interaction,
where a nonlinear analysis was performed. To study the post buckling behaviour a
discrete spring model is utilised. A comparison with the atomistic model shows the
deficiency of the continuous formulation of the additional continuous metric term.
Increasing only the undeformed length of the outermost springs, results in similar
post buckling configurations, comparable to atomistic results.
References
1. Ascher, U., Christiansen, J., Russell, R.D.: A collocation solver for mixed order systems of
boundary value problems. Math. Comp. 33, 659–679 (1979)
2. Audoly, B., Pomeau, Y.: Elasticity and Geometry: From Hair Curls to the Nonlinear Response
of Shells, p. 600. Oxford University Press, Oxford (2010)
3. Brenner, D.W., Shenderova, O.A., Harrison, J.A., Stuart, S.J., Ni, B., Sinnott, S.B.: A second-
generation reactive empirical bond order (REBO) potential energy expression for hydrocarbons.
J. Phys. Condens. Matter 14, 783 (2002)
4. Fung, Y.C.: Foundations of Solid Mechanics, p. 525. Prentice Hall, Englewood Cliffs (1965)
5. Jun, S.: Density-functional study of edge stress in graphene. Phys. Rev. B. 78, 073405 (2008)
6. Liang, H., Mahadevan, L.: The shape of a long leaf. Proc. Natl. Acad. Sci. U.S.A. 106, 22049–
22054 (2009)
7. Liu, Y., Dobrinsky, A., Yakobson, B.I.: Graphene edge from armchair to zigzag: the origins of
nanotube chirality? Phys. Rev. Lett. 105, 235502 (2010)
8. Lu, Q., Huang, R.: Nonlinear mechanics of single-atomic-layer graphene sheets. IJAM 1, 443–
467 (2009)
9. Marder, M., Deegan, R.D., Sharon, E.: Crumpling, buckling, and cracking: elasticity of thin
sheets. Phys. Today 60, 33–38 (2007)
10. Plimton, S.J.: Fast parallel algorithms for short-range molecular dynamics. J. Chem. Phys. 117,
1–19 (1995). http://lammps.sandia.gov, Cited 2012
11. Qiang, L., Rui, H.: Excess energy and deformation along free edges of graphene nanoribbons.
Phys. Rev. B. 81, 155410 (2010)
12. Rammerstorfer, F.G., Fischer, F.D., Friedl, N.: Buckling of free infinite strips under residual
stresses and global tension. J. Appl. Mech. 68, 399–404 (2001)
13. Seung, H.S., Nelson, D.R.: Defects in flexible membranes with crystalline order. Phys. Rev. A
38, 1005–1018 (1988)
14. Sharon, E., Efrati, E.: The mechanics of non-Euclidean plates. Soft Matter 6, 5693–5704 (2010)
15. Stuart, S.J., Tutein, A.B., Harrison, J.A.: A reactive potential for hydrocarbons with intermole-
cular interactions. J. Chem. Phys. 112, 6472–6486 (2000)
16. Wang, C.Y.: On the buckling of a circular plate on an elastic foundation. J. Appl. Mech. 72,
795–795 (2005)
17. Yakobson, B.I., Brabec, C.J., Bernholc, J.: Nanomechanics of carbon tubes: instabilities beyond
linear response. Phys. Rev. Lett. 76, 2511–2514 (1996)
On the Modeling of Surface and Interface
Elastic Effects in Case of Eigenstrains
Konstantin B. Ustinov, Robert V. Goldstein and Valentin A. Gorodtsov
Abstract The constitutive equations of interface elasticity in case of eigenstrains are

obtained in terms of interface (surface) values defined as integrals of the excesses of
the corresponding volumetric values over the normal to the interface. The equations
are consistently linearized, which corresponds to the case of both elastic strains and
eigenstrains being small. It is shown that the obtained equations possess more general
form then Shuttleworth equations. The obtained type of equation was confirmed by
considered example: an interface formed by a thin layer of constant properties. It was
also shown that the type of energetic restrictions on the surface elastic constants may
depend essentially on the definition of the position of the surfaces.
1 Introduction
The state of matter at or near the surface or interface is generally different from the
state of the matter in the bulk. Therefore to describe behavior of finite body (liquid or
solid) the specific surface values (such as energy, stress and so on) are introduced in
addition to bulk specific values [1]. Historically efforts of studying the surface effects
were devoted to liquids rather than solids. That is because the influence of surface
effects is much more pronounced in liquids than in solids. However with reducing
the sizes of the objects under consideration the relative amount of the matter near
the surface increases, raising the role of the surface effects. The present article is
devoted to peculiarities in describing surface and interface effects in elasticity.
K. B. Ustinov (B) · R. V. Goldstein · V. A. Gorodtsov

A. Ishlinsky Institute for Problems in Mechanics RAS,
Prospect Vernadskogo 101, Moscow 119526, Russia
e-mail: ustinov@ipmnet.ru
R. V. Goldstein
e-mail: goldst@ipmnet.ru
V. A. Gorodtsov
e-mail: gorod@ipmnet.ru

168 K. B. Ustinov et al.
Fig. 1 Distribution of volu-

metric value near interface
2 Surface Values
Let us first introduce the definition of the surface value. Following [1], the surface
density g s (x, y) of any value g at any point (x0 , y0 ) of the surface z(x, y) between two
phases A and B (the interface media-vacuum forms a particular case) is understood
as the integral of the excess of the volumetric density of the corresponding value
g(z) over the normal to the surface intersecting it at the point under consideration
(Fig. 1).
zB
g (x, y) ≡
s
g(x, y, z) dz − h A g A (x, y) − h B g B (x, y),
zA
h A = (z 0 − z A ), h B = (z B − z 0 ) (1)
Here g A and g B are magnitudes of the value in question in phases A and B, respec-
tively. There remains, however, a variety in the choice of the position of the boundary,
z 0 , unless it is prescribed externally.
The variation of the density of the surface energy is
zB
δW =
s
σi j (z)δεi j (z) dz − σiAj δεiAj h A − σiBj δεiBj h B (2)
zA
It is followed directly from here that the surface energy (as well as surface elastic
moduli) should not necessarily be positive. Only the total elastic energy within the
limit of integration (the integral term in Eq. (2)) should be positive to provide the
stability of the solid. However, when considering an interface solid-vacuum (or in
some approximation solid-gas), the choice in defining the interface position is more
flexible: thus let the phase A (Fig. 1) is the vacuum; then the choice of the position
On the Modeling of Surface and Interface Elastic Effects 169
of the interface at point z 0 leads to the above description, while the choice of the
position of the interface at point z B leads to necessity for surface energy and elastic
moduli be positive.
It is convenient to present both tensors of stress, σi j , and strain, εi j , as a sum of
longitudinal and transverse parts (e.g. [2, 3]). For any second rank symmetric tensor
it holds true
⎞ ⎛ ⎛ ⊥
⎞
ω11 ω12 0 0 0 ω13
ω = ω + ω⊥ = ⎝ ω12

ω22 0 ⎠ + ⎝ 0 0 ω23 ⎠
⊥ (3)
⊥ ⊥ ⊥
ω13 ω23 ω33
0 0 0
However, such a representation makes sense in coordinate frame related to the sur-
face, only, and, hence, it is not a tensor decomposition. Application of Eq. (3) to
Eq. (2) yields
zB
A A B B
δW s = σi j (z) δεi j (z) dz − σi j δεi j h A − σi j δεi j h B
zA
zB
+ σi⊥j (z) δεi⊥j (z) dz − σi⊥A ⊥A ⊥B ⊥B
j δεi j h A − σi j δεi j h B (4)
zA
Further simplification is possible [2] if the conditions of compatibility and equi-

librium are satisfied
A B
δεi j = δεi j = δεi j , (5)
σi⊥j = σi⊥A
j = σi⊥B
j (6)
Then the expression for variation of the surface energy (4) is reduced to

δW s = σisj δεi j + σi⊥j δεisj = σisj δεi j + σi j δεisj (7)
Here
zB
A B
σisj = σi j (z) dz − σi j h A − σi j h B , (8)
z
Az B
εisj = εi⊥j (z) dz − εi⊥A ⊥B
j h A − εi j h B (9)
zA
are reasonable to call surface stress and surface strains, respectively. According to
Eq. (7) they do not form an energetic pair. Their nonzero components are
⎞⎛ ⎛ ⎞
σ11
s σs 0
12 0 0 ε13 s
σ s = ⎝ σ12
s σs 0⎠
22 , εs = ⎝ 0 0 ε23 s ⎠
(10)
0 0 0 ε13 ε23 ε33
s s s
Tensor of surface stress, as it is seen from Eq. (10) may be treated as a 2-D 2nd rank
tensor. The dimensionality of surface stress, σi j , and surface strain, εi j , are N/m, and
m, respectively. The introduced surface strains are effective normal and tangential
displacement discontinuity ε33 s =
[u 3 ] , 2ε13
s =
[u 1 ] , 2ε23
s =
[u 2 ].
The expressions for surface stresses

∂ W s ∂ W s

= σisj , = σi⊥j (11)
∂ε i j εs
∂εisj
εi j
ij
are followed directly from the appearance of the expression for the energy variation
(7), provided the accepted definition of the surface values to the material element.
The first of these expressions is the well-known Shuttleworth equation [4], written
in Lagrangian variables (e.g., [2]). Usually, the Shuttleworth equation is written in
terms of Eulerian variables
∂Ws
σαβ
s
= γ0 δαβ + ,
∂εαβ
where the term associated with the increase of the surface area is accounted (see
Appendix 6). This representation will not be used hereafter. Further we restrict our-
selves with the consideration of small deformations. If the additional conditions
δεisj = 0 ∨ σi⊥j = 0 (12)
are satisfied, the first expression of Eq. (11), as followed from Eq. (7), is written as

σisj = σis0
j + Ci jkl εkl = σi j + Ci jkl εkl , i, j, k, l = 1, 2
s s0 s
(13)
which is more familiar form of the Shuttleworth equation in Lagrangian variables.

Generally, Eq. 12 correspond to different values of stresses σis0 s
j and moduli Ci jkl .
Usually while deriving Shuttleworth equations conditions (12) are supposed to be
fulfilled a priori, which is true, or acceptable for number of cases, in particular for
free surfaces. Equation (13) are simplified for isotropic surface
σ11
s
= σ11
s0
+ (λs + 2μs )ε11 + λs ε22 ,
σ22
s
= σ22
s0
+ λs ε11 + (λs + 2μs )ε22 , (14)
σ12
s
= σ12
s0
+ 2μ ε12 s
These equations may be considered as consistently linearized form of equations

of [5, 6], which means that not only elastic strains are small, but the eigenstrains
(εi0j ∼ σis0
j /λ ), corresponding to residual stresses, are small too. However, for
s
interfaces in solids conditions (12) cease to be satisfied, and according to the first
expression of (11) the surface stresses σisj will be different for different transverse

strains εisj , i.e. they may be considered as functions of both longitudinal εi j , and
transverse εisj strains. This corresponds to generalized Eq. 13 up to the following
s⊥ ⊥
σisj = σis0
j + Ci jkl εkl + Ci jkl εkl = σi j + Ci jkl εkl + Ci jkl εkl
s s0 s s⊥ s
(15)
Similarly, for the transverse stress
σi⊥j = Cis⊥ ⊥⊥ s
jkl εkl + Ci jkl εkl (16)
It should be emphasized that the free terms in Eqs. (13)–(15) correspond to residual
stresses, or may be recalculated in terms of eigenstrains, rather than to the surface
energy. According to the used definition of surface stresses and strains coefficients
Cisjkl , Cis⊥ ⊥⊥ 2 3
jkl , Ci jkl have different dimensions, N/m, N/m , N/m , respectively.
Expressions (15), (16) may be thus considered as a generalization of Shuttleworth
equation [3]. Two usually considered models of surface effects, namely Shutleworth
an Winkler models, may be derived from the general expressions (15), (16) as par-
⊥⊥
ticular cases (the first case by dropping terms with Cis⊥ jkl , Ci jkl , the second case by
dropping terms with Cisjkl , Cis⊥ s⊥
jkl ), while due to the presence of cross-terms Ci jkl in
Eqs. (15), (16) they may not be formed [7] by combining Shutleworth and Winkler-
type [8–10] models.
For a wide range of problems the interface may be considered as transversally
isotropic, with the symmetry axis being directed along the interface normal. In that
case Eq. (15) are reduced to
σ11
s
= σ11
s0
+ C1111
s
ε11 + C1122
s
ε22 + C1133
s⊥ s
ε33 ,
σ22
s
= σ22
s0
+ C1122
s
ε11 + C1111
s
ε22 + C1133
s⊥ s
ε33 , (17)
σ12
s
= σ12
s0
+ 2C1212
s
ε12 ,
s
C1212 = (C1111 − C1122 )/2,
s s
⊥ ⊥⊥ s
σ33 = σ33 = C1133
s⊥
ε11 + C1133
s⊥
ε22 + C3333 ε33 ,
⊥ ⊥⊥ s
σ13 = σ13 = 2C1313 ε13 , (18)
⊥ ⊥⊥ s
σ23 = σ23 = 2C1313 ε23
Consider the case, where the near surface changes of all values are localized within
a narrow layer of thickness h, within which they are constant.
3 Model of the Interface as a Surface Layer

with Vanishing Thickness
Consider a layer of thickness h, embedded between two layers of other material so

that the total thickness of the structure is H > h. All three materials are supposed
elastic, transverse isotropic with the symmetry axis being directed along the interface
normal (Fig. 2). In the coordinate frame with z-axis, aligned normally to the layers the
i , Ci , Ci , Ci , Ci , Ci =
elastic properties are describe by the set of constants C11 12 13 33 44 66
(C11 − C22 )/2, where i = A for the lower layer, i = B for the upper layer, i = C for
i i
the interface layer. Consider homogeneous loading of the region of the considering
structure, containing the intermediate layer and the adjusted parts of upper and lower
layers. One-index notation will be used for stress and strain tensors
ε11
k
= ε1k , ε22
k
= ε2k , ε33
k
= ε3k , 2ε23
k
= ε4k , 2ε13
k
= ε4k , 2ε12
k
= ε6k ,
σ11
k
= σ1k , σ22
k
= σ2k , σ33
k
= σ3k , σ23
k
= σ4k , σ13
k
= σ5k , σ12
k
= σ6k ,
k = A, B, C (19)
The elastic energy of the system is
CA
H H H
h h'
C C’
CC h'=0
CB
Fig. 2 Interface as a surface layer with vanishing thickness

h C C2

U = U0 + C11 ε1 + ε2C 2 + 2 C12 ε1 ε2 + 2 C13

C C C C C
ε3 ε1C + ε2C
2

+ C33 ε3 + C66
C C2
ε6 + C44
C C2 C
ε4C 2 + ε5C 2 (20)
Here U0 is the elastic energy of the upper and lower layers.

Now, we replace the intermediate layer of thickness h with a structure consisting
of three layers: the intermediate layer of thickness h < h with elastic constants

CiCj , and two layers of thickness (h − h )/2 with elastic constants CiAj and CiBj ,

corresponding to the upper and lower layers. The constants CiCj should be chosen
so that the total elastic energy coincide for the original and effective systems. Then

directing h → 0, and changing simultaneously CiCj so that the elastic energy of the
system conserve, we obtain the equivalent structure of thickness H , consisting of two
layers of thickness H/2 separated by an interface of zero thickness but possessing
some elastic properties Ai j .
Execute the outlined procedure. The elastic energy of the equivalent system is
h − h A A 2

Ueff = U0 + C11 ε1 + ε2A 2 + 2C12 ε1 ε2 + 2 C13

A A A A A
ε3 ε1A + ε2A
4
+ C33 ε3 + C66
A A2
ε6 + C44
A A2 A
ε4A 2 + ε5A 2

+ C11
B
ε1B 2 + ε2B 2 + 2 C12 ε1 ε2 + 2 C13
B B B
ε3 ε1B + ε2B
B B
(21)

+ C33 ε3
+
B B2
ε6
B B2
C66 ε4B 2 + ε5B 2
+ C44
B
h C C2

C C C C C

+ C11 ε1 + ε2C 2 + 2C12 ε1 ε2 + 2 C13 ε3 ε1C + ε2C

2

C C2 C C2 C
+ C33 ε3 + C66 ε6 + C44 ε4C 2 + ε5C 2

Here longitudinal strains, ε1C , ε2C , ε6C , remain the same as for the initial system

ε1A = ε1B = ε1C = ε1C = ε1 , ε2A = ε2B = ε2C = ε2C = ε2 ,

ε6A = ε6B = ε6C = ε6C = ε6 (22)
while the following relations
h − h A h − h B
ε5 + ε5 + h ε5C = h ε5C ,
2 2
h − h A h − h B
ε4 + ε4 + h ε4C = h ε4C , (23)
2 2
h − h A h − h B
ε3 + ε3 + h ε3C = h ε3C
2 2
take place for the transverse strains. These relations relate the fact that the displace-
ments of the external surface of the initial and equivalent composed intermediate
layer should coincide. Besides, the strains should minimize the elastic energy (21)
C C
C13 − C13
A ε1 + ε2 + C33
C εC
3 C13 − C13
B ε1 + ε2 + C33
C εC
3
ε3A = A
, ε3B = A
,
C33 C33
(24)
C εC
C44 C εC
C44 C εC C εC
C44 C44
ε5A = A
5
, ε5B = B
5
, ε4A = A
4
, ε4B = B
4
C44 C44 C44 C44
which is equivalent to the equivalence of stresses σ3 , σ4 , σ5 in layers A, B, C.

The substitution of Eq. (23) into Eq. (21) yields the expression for the elastic
energy of the equivalent system (20), expressed in terms of the same kinematics vari-
ables ε1 , ε2 , ε6 , ε3C , ε4C , ε5C as the original system. To ensure the equivalence, equa-
tion U = Ueff must be satisfied for any kinematics variables ε1 , ε2 , ε6 , ε3C , ε4C , ε5C
and any elastic constants in all three layers Cikj (k = A, B, C). Solving the system
of six equations (corresponding to six independent strains) after performing the limit
transition h → 0 we finally obtain
A11 = A22 = A12 + 2 A66 (25)
h
A12 = A21 = A C13

B2 A
2C33 − C33C
A CB
4C33 33 − 2C33
C33 + C33
C B

+ C13
A2 B
2C33 − C33
C
+ 2 C13
C2 A
C33 + C33
B
(26)
− 4C13 C13 C33 − 4 C13
B C A
C13 C33 + 2 C13
A C B A B C
C13 C33

− 2 C12C
− C12A
− C12
B C
C33 A
C33 + C33
B
− 2C33
A B
C33 ,
B A
C13 C33 + C13 A C B CC − 2 CC C A C B
33
A13 = A23 = A31 = A32 = 33
C C A + C B − 2C A C B
13 33 33
, (27)
C33 33 33 33 33
−1
1 1 1 1
A33 = C
− A
− B
, (28)
h C33 2C33 2C33
−1
1 1 1 1
A44 = A55 = C
− A
− B
, (29)
h C44 2C44 2C44

A + CB
C66
A66 = h C66 −
C 66
(30)
2
Here

A11 = lim

h C11
C
, A12 = lim

h C12
C
, (31)
h →0 h →0

A13 = lim

C
C13 = C13
C
, (32)
h →0

A33 = lim

C
C33 / h, A44 = lim

C
C44 / h, A66 = lim

h C66
C
(33)
h →0 h →0 h →0
The obtained moduli Ai j possess different dimensions (see considerations after

Eq. (16)). They, as expected, being independent of h , however, depend on the
thickness of the initial layer h, which is natural: thus longitudinal surface moduli
A11 , A12 , A66 of a rigid intermediate layer of thickness 2h will be almost twice
higher then the moduli of the layer of thickness h. It is important that the effective
moduli Ai j are determined by the thickness h and elastic moduli of the layers through
combinations of the type of Eqs. (25)–(30).
Consider some particular cases:
• Rigid layer
Moduli CiCj have the same order and tend to infinity so that
hCiCj → const for h → 0 (34)
Fulfilling energy equivalence, at least in the sense of the leading term of

A
C66 + C66
B
/C66
C
,
require additional restrictions. In order A66 (30) be constant for h → 0 it is

necessary for C66 C to approach infinity as A / h + C A + C B /2. Similar rela-
66 66 66
tions should be satisfied for C11C , C C . As for moduli A , A , A , in order to
12 33 44 55
prevent infinite contribution to the elastic energy to prohibit the associated kine-
matic values.The considered limit corresponds to the classical surface elasticity.
Non-vanishing coefficients for this case are obtained by expanding (25)–(30) over
small CiAj /CiCj , CiBj /CiCj

C2
C13
A11 = A22 = A12 +2 A66 , A12 = A21 = h C12
C
− C
, A66 = hC66
C
(35)
C33
For isotropic case these formulae coincide with the one of [11].
• Soft layer
Moduli CiCj have the same order and tend to zero so that
h CiCj −1 → const for h → 0,

i.e. compliance rather than rigidity increases with the layer thickness decreasing. In
this case only terms with A33 , A44 contribute essentially to elastic energy. The other
terms either vanish (terms with A11 , A12 , A66 ), or (term with A13 ) is multiplied by
the vanishing value. The obtained model is the model of springs [8–10]. Condition
of the model applicability shows that the model makes sense for soft interfaces.
Non-zero coefficients of the model are obtained by expanding Eqs. (25)–(30) over
small CiCj /CiAj , CiCj /CiBj
C
C33 C
C44
A33 = , A44 = A55 = (36)
h h
However, none of the considered particular cases allows conserving the elastic
energy for the arbitrary elastic moduli and arbitrary deformation. The conserva-
C −1 C −1
tion takes place when conditions h C33 → const, h C44 → const, h C11C →
C → const, C C → const for h → 0 are satisfied, which corresponds to
const, h C12 13
(31)–(33) and could not be obtained by a superposition of cases 1 and 2 due to the
presence of cross-term with A13 .
4 Generalization of the Model in Case of Eigenstrains
If eigenstrains occur in at least one of the phases A or B (Fig. 2), then it seems
inconsistent to neglect the interface eigentrains. Indeed, the reasonable question
arises: from which state should one calculate the strains of the interface. The answer
is obvious: since for the interface there should be no preference between the left
and right sides the interface may possess eigenstrains equal to the eigenstrains of left
s0 , ε 0 .
layer, the one of the right layer, or, the most natural case, its own eigenstrains εkl kl
Therefore the constitutive equations for the interface elasticity may be written for
the considered linear case in the following form

σisj = Cisjkl εkl − εkl

0
+ Cis⊥
jkl ε s
kl − εkl ,
s0
(37)

σi⊥j = Cis⊥
jkl εkl − ε 0
kl + C ⊥⊥
i jkl ε s
kl − ε s0
kl
It is seen that Eq. (37) may be considered as another form of Eqs. (15), (16), which
means that at least in linear case the descriptions in terms of surface eigenstrains and
surface residual stresses are equivalent.
A generalization of the described problem may be performed by assuming that
each layer, including the interface layer, may possess eigenstrains, and that the total
strains (which may be assumed to be the sum of eigenstrains and elastic strains) is
compatible. Repeating the described procedure for comparing the elastic energy of
the original and equivalent system one obtains the system of equation for unknown
parameters, which in addition to the elastic moduli will contain parameters relating
stress components with eigenstrains. Such a system was formed and solved in [3],
however the solution is too awkward to be written here.
Importance of accounting the surface eigenstrains was demonstrated in [12],
where the solution of a generalized Eshelby problem for a spherical inclusion is
given accounting eigenstrains and surface eigenstrains was obtained. In particular it
was shown that depending on the surface eigenstrains the additional terms, relate to
the surface effect may have opposite signs.
5 Conclusions
The constitutive equations of interface elasticity are obtained in terms of interface

values defined as integrals of the excesses of the corresponding volumetric values
over the normal to the interface. It is shown that the obtained equations possess
more general form then Shuttleworth equations [4]. In general case the constitutive
equations include both in-plane surface stresses and strains and stresses and strains
with components normal to the interface surface. Therefore tensor of surface elastic
moduli becomes 3-D [3] rather then 2-D as it was in the classical Shuttleworth
equations.
The two classical types of constitutive equation for surface elasticity (Shuttle-
worth equations, where strains are supposed to be continuous through the interface,
while normal components of stresses are allowed to possess jumps, and the model
of linear springs, e.g. [9], where, on the contrary, stresses are supposed to be con-
tinuous through the interface, while normal components of strains are allowed to
possess jumps) are obtained as particular cases. The ranges of parameters, for which
application of the particular cases is justified, are estimated.
The influence of the choice of type kinematics description (Lagrangian or
Eulerian) on the form of the obtained equations is also discussed.
A particular case of a medium with interface is considered, for which all near-
interface changes in elastic properties are localized within a narrow layer, where
they are constant, An equivalent medium with a layer of zero thickness is introduces,
which effective surface elastic properties is chosen so that elastic energy of the
initial and equivalent samples be equal for any externally imposed kinematics. It is
shown that the constitutive equations in this case are in the form of the proposed
type, and that they may be reduced to one of the two mentioned classical forms
only for particular combinations of elastic constants of the phases. The full set of
effective surface elastic constants is obtained for the case of transversal isotropy of
two adjusted phases and the interface layer.
The model is generalized to the case of the presence of eigenstrains within the
interface layer as well as within the adjusted phases.
6 Appendix: On Various Forms of Shuttleworth Equation
Following [4] consider deformation of a crystal surface, when the initially square
element of area A0 transforms to a rectangle with the sides parallel to the initial
square. The process require the work
dW = σ1 d A1 + σ2 d A2 , (38)
where σ1 , σ2 are components of surface stresses, normal to the square sides; A1 , A2

are the increases in areas in that directions. For isotropic surfaces (which takes place,
for example, for crystallographic planes with symmetry axes of the 3-D and 6th
order) and hydrostatic stresses σ1 = σ2 = σ the expression for work is reduced to
dW = σ d A (39)
Here d A = d A1 + d A2 . In case of elastic process this work is transformed to the

increase in elastic energy U , which may be represented as a product of the specific
elastic energy and the area. However, the definition of the specific energy is not
unique: as the ratio of the energy to the current an initial area, respectively.
γ = U/A, (40)
γ0 = U/A0 (41)
Usually the first form is used, for which the differential of the work is
σ d A = d (γ A) (42)
From where:
dγ
σ =γ +A (43)
dA
If the strain increment is defined as
dA
dε = ,
A
Equation (43) reads
dγ
σ =γ + , (44)
dε
which is the most customary form of the Shuttleworth equation.
Using Eq. (41), as a definition of the specific energy, instead of Eq. (42) we have
σ d A = d (γ0 A0 ) (45)
Since A0 is a constant, it may be carried out of the integral. As the result the analog
of Eq. (43) is
dγ0
σ = A0 (46)
dA
According to Eqs. (40), (41) γ0 = γ A/A0 . Substitution of this expression into
Eq. (46) gives Eq. (43). Defining the strain increment as
dA
dε0 = ,
A0
leads to Eq. (46) reducing to

dγ0
σ = (47)
dε0
For liquids, since the quantity γ , determined by Eq. (40), does not depend on A
(the number of atoms at the surface is proportional to the current surface area), it is
preferable to use representation (44), followed (due to vanishing the derivative in the
right hand side of Eq. (43)) by identity of the surface energy and the surface stress
σ = γ.
For solids, from the point of view of rational mechanics, both representations
may be used. However if the new surface is not form (the number of surface atoms
remains the same, while the distance between them increases), which seems to be the
case of surface elasticity, it is natural to refer the elastic energy to the initial surface,
i.e. to use representation (41), (47). Note that in both cases the introduced specific
energy corresponds to the energy of a particular number of atoms.
Acknowledgments The work was done under financial support of the Program of the Executive
Committee of RAS Nr. 24.
References
1. Ibach, H.: The role of surface stress in reconstruction, epitaxial growth and stabilization of
mesoscopic structures. Surf. Sci. Rep. 29, 195–263 (1997)
2. Podstrigach, Y.S., Povstenko, Y.Z.: Introduction to Mechanics of Surface Effects in Solids (in
Russ.). Naukova Dumka, Kiev (1985)
3. Müller, P., Saul, A.: Elastic effects on surface physics. Surf. Sci. Rep. 54, 157–258 (2004)
4. Shuttleworth, R.: The surface tension of solids. Proc. Phys. Soc. A63, 444–457 (1950)
Mech. Anal. 57(4), 291–323 (1975)
Mech. Anal. 59, 389–390 (1975)
7. Ustinov, K.B.: On constructing a theory of surface elasticity for an internal plane interface in
case of eigenstrains. Preprint IPMechRAS Nr. 987, 2011 (in Russ.)
8. Hashin, Z.: Thermoelastic properties of fiber composites with imperfect interface. Mech. Mater.
8, 333–348 (1990)
9. Duan, H.L., Wang, J., Huang, Z.P., Karihaloo, B.L.: Eshelby formalism for nano-
inhomogeneities. Proc. R. Soc. A461, 3335–3353 (2005)
10. Duan, H.L., Wang, J., Huang, Z.P., Karihaloo, B.L.: Size-dependent effective elastic constants
of solids containing nano-inhomogeneities with interface stress. J. Mech. Phys. Solids 53,
1574–1596 (2005)
surface stresses taken into account. Mech. Solids 45(3), 331–342 (2010)
12. Goldstein, R.V., Gorodtsov, V.A., Ustinov, K.B.: Effect of residual stresses and surface elasticity
on deformation of nanometer spherical inclusions in elastic matrix. Phys. Mesomech. 13(5–6),
318–328 (2010)
On Kinetics of Chemical Reaction Fronts in
Elastic Solids
Elena N. Vilchevskaya and Alexander B. Freidin
Abstract A chemical reaction front where an oxidation reaction is localized is

considered as an internal surface dividing two solid deformable constituents. The
reaction is sustained and controlled by the diffusion of the gas constituent through
the oxide layer. The transformation strains produced by the chemical reaction lead to
internal stresses which in turn affect the chemical reactions front kinetics. Analitical
solution of axially-symmetric mechano-chemistry problems in a case of small strain
approach are obtained. We examine how stress state affects the reaction front kinetics
and demonstrate reaction locking effects due to internal stresses. We also study how
the reaction rate depends on the chemical reaction front curvature.
1 Introduction
Interconnections between kinetics of chemical reaction and deformation and fracture

processes in a body are of significant interest for both fundamental and applied engi-
neering science. Chemo-mechanical problems have received a new attention in recent
years due to miniaturization of structure elements. For example, fracture processes
in micron-scale parts of MEMS made of polycrystalline silicon thin films involve
sequential oxidation of polysilicon and environmentally-assisted crack growth inside
an oxide layer. In turn, the kinetics of the oxide growth is determined by mechani-
cal stresses produced by the crack. The catastrophic failure happens when the crack
reaches the reaction front. Thus, major events which determine the life time of MEMS
are related with coupling between stresses and chemical reactions (see details in [17]).
E. N. Vilchevskaya (B) · A. B. Freidin

Institute for Problems in Mechanical Engineering, Russian Academy of Sciences,
V.O. Bolshoy pr. 61, Saint-Petersburg 199178, Russia
e-mail: vilchevska@gmail.com
B. Freidin
e-mail: alexander.freidin@gmail.com

182 E. N. Vilchevskaya and A. B. Freidin
Reactions similar to silicon oxidation also take place in hydrogen absorption in the
process of metal hydride formation used in hydrogen storage applications [7, 8].
Both reaction are sustained by the diffusion of the gas constituent through the trans-
formed material and the chemical reaction takes place at the chemical reaction front
that divides two solid constituents. The density change and lattice misfit produced
by the chemical reactions lead to internal stresses which in turn affect the chemical
reaction front kinetics.
Many models of the reaction kinetics in metal/hydrogen or silicon/silica systems
arise to pioneering papers by Deal and Grove [1]. For example in the work [7] a
model to describe moving planar interface growth kinetics for hydride formation is
developed. To construct the model they track the chemical potential of a hydrogen
as it traverses from the gas phase to the hydride phase. However neither external
loading nor internal stresses were taken into account.
One of the first attempts to obtain an expressions of the chemical potential in a
multicomponent solid under stress was made by Larche and Cahn [11–14]. They
considered diffusing solids and showed that the chemical potential depends the trace
of the stress tensor. This result was further developed in [21], where it was justified
that the chemical potential is proportional to the trace of the Eshelby stress tensor.
In the present paper we consider a chemical reaction in an elastic body within
the frameworks of configuration forces concept. Basing on the expression for the
chemical affinity tensor obtained in [3] as a combination of Eshelby stress tensors
of the solid constituents and a chemical potential of a gas constituent, we formulate
a kinetic equation that relates the reaction front velocity and the normal component
of the chemical affinity tensor. As a result the front velocity depends on chemical
parameters and temperature as well as on stresses. In turn, stresses depend on the
position of the reaction front.
Then we introduce the notion of the equilibrium concentration of the gas con-
stituent. We say that the gas concentration is equilibrium at the reaction front if,
given temperature, front position and stresses, the chemical affinity is equal to zero.
This allows us to formulate a kinetic relationship for the reaction front velocity
in terms of current gas concentration at the reaction front and the equilibrium gas
concentration that depends on stresses at the front. We note that if the equilibrium
concentration at the reaction front is greater than the concentration at the outer sur-
face of a body then the reaction front propagation is impossible. This observation
leads to the possibility of locking effects related with reaction blocking by internal
or/and external stresses.
Finally we give solutions of two simplest axially-symmetric mechano-chemical
problems considering oxidation of infinite body with a cylindrical hole and oxidation
of a cylinder. We study in detail how the oxidation kinetics and locking effect depend
on stresses and the hole or cylinder radii and emphasize differences between the
reaction kinetics in the cases of a hole and a cylinder.
On Kinetics of Chemical Reaction Fronts in Elastic Solids 183
dX
V V Fe
0
G
υ υ
dXg Fe t
Vg
g
Fig. 1 Configurations resulting from chemical transformation and deformation
2 Statement of the Problem
Having in mind a silicon oxidation Si + O2 → SiO2 we consider a solid body in

which a chemical reaction
ν− B− + ν∗ B∗ → ν+ B+ (1)
takes place, where B− , B+ and B∗ are chemical formulae of the reacting constituents,
ν− , ν∗ and ν+ are stoichiometric coefficients, B− (the initial material) and B+
(“oxide”) are solid constituents, and B∗ is a gas constituent (oxygen).
Direct experimental observations [6, 18, 20] show that the bulk of the oxide does
not incorporate oxygen during oxidation and that the new oxide essentially grows at
the Si–SiO2 interface due to the diffusion of the oxidant species through the oxide.
Thus we suppose that the reaction is localized at the front that divides the solid
constituents and sustained and controlled by the diffusion of the gas constituent
through the chemically transformed material B+ .
For simplicity sake we ignore thermal effects of the reaction and assume that
temperature T is a given parameter. We also neglect the stress relaxation effects due
to inelastic deformations and consider constituents B− and B+ as elastic materials.
2.1 Strains and Stresses: Chemical Transformation Strain Tensor
Let υt = υ− ∪ υ+ is a current configuration of a body at time t, and volumes υ− and

υ+ are occupied by solid materials B− and B+ , respectively (Fig. 1).
Along with the reference configuration V0 = V− ∪ V+ of the material before the
reaction we introduce an intermediate reference configuration Vg of the body that
corresponds to a stress-free oxide B+ (Fig. 1). Material segments dx− and dx+ from
υ− and υ+ can be related with corresponding segments dX− and dXg from V− and
Vg by elastic deformation gradients:
du − ρ0
dx− = F−
e
· dX− , det F−
e
= = ,
d V− ρ−
du + ρg
dx+ = F+
e
· dXg , det F+
e
= = ,
d Vg ρ+
where ρ0 , ρ− , ρ+ and ρg are densities of initial material and oxide in corresponding

configurations.
Although material points in V+ and Vg+ present different materials, they are related
by the chemical reaction (1), and linear segments dX and dXg which consist of
corresponding points of B− and B+ can be related by the chemical transformation
strain tensor G as
dXg = G · dX, (2)
where G is determined basing on chemical reaction formula and mass balance as

follows [3]. If M− and M+ are the molar masses of B− and B+ respectively then
due to the chemical reaction (1) the volume element d V0 = ν− M− /ρ0 transforms
into the volume element d Vg = ν+ M+ /ρg . Thus
d Vg ν+ M+ ρ0
det G = = (3)
d V0 ν− M− ρg
Note that det G = ρ0 /ρg . Further we assume that the transformation tensor is
isotropic:

ν+ M+ ρ0 1/3
G = gE, g = , (4)
ν− M− ρg
where E is the unit tensor.

One may also consider a transformation
dx+ = F+ · dX+ , F+ = F+
e
· G, dx+ ⊂ v+ , dX+ ⊂ V+ (5)
Representation (5) would be Lee’s decomposition if G was a plastic strain tensor

[15]. A representation of this kind have been also used in mechanics of biological
growth with G as a growth (“transplant”) tensor (see, e.g. [2, 5, 8, 9, 16, 19] and
reference therein).
For simplicity sake we accept a solid skeleton approach. It means that gas diffusion
does not affect strains in the oxide, oxide stresses depend only on its deformation
and skeleton deformation does not affect gas pressure. Under these assumptions the
Piola-Kirchhoff stress tensors S− and S+ referred to V− and Vg+ are given by the
g
constitutive equations
∂ f− g ∂ f+
S− = ρ0 , S+ = ρg e , (6)
∂F−e ∂F+
where f − = f − (F−
e , T ), f = f (Fe , T ) are free energies of B and B per unit
+ + + − +
mass, T is a temperature.
2.2 Chemical Affinity Tensor
The entropy production P[S] due to stress-assist chemical reaction front quasi-static
propagation in a case of constitutive equations (6) is given by [3]

ρ0 Γ ρ0
TP[S] = N · A · V dΓ = A N VNΓ dΓ ≥ 0, (7)
ν− M− ν− M−
Γ Γ
A N = N · A · N, VNΓ = VΓ · N,
where VΓ and N are a reaction front velocity and a normal to the reaction front with
respect to the reference configuration, A is a chemical affinity tensor,
A = ν− M− b− + ν∗ M∗ M∗ − ν+ M+ b+ (8)
1 g 1 T
b+ = f + E − (S+ )T · F+
e
, b− = f − E − S− · F−
e
ρg ρ0
M∗ is the molar mass of the gas, b− , b+ are Eshelby stress tensors determined with
respect to V0 and Vg configurations, M∗ = μ∗ (c, T )E, μ∗ is a chemical potential
density of a gas constituent, c = ρ∗ /ρg is a relative gas concentration, ρ∗ is a gas
density.
Tensorial character of the chemical affinity reflects the fact that the reaction front
velocity depends on the orientation of the front with respect to the stress tensor.
Note the similarity between the tensorial expression (8) and the expression of
a chemical
affinity in classical chemical thermodynamic given by the formula
A = νk Mk μk , where μk is the chemical potential per unit mass of the k-th
constituent, stoichiometric coefficient νk is taken with the sign “−” if the k-th con-
stituent is produced due to the chemical reaction and taken with the sign “+” in other
case [4].
Note also that in a case of martensite phase transformation, due to mass balance
g 3 ≡ d Vg /d V0 = ρ0 /ρg , M∗ = 0, and Eq. (7) becomes a known formula for entropy
production due to interface boundary propagation [10].
2.3 The Equilibrium Gas Constituent Concentration and Kinetic

Equation
Basing on the expression of the entropy production (7) one can represent the kinetic
equation that determines the normal component of the reaction front velocity as a
function of the normal component of the affinity tensor:
VNΓ = V (A N ), VNΓ V (A N ) ≥ 0. (9)
Note that the transformation strains and changes in elastic modules lead to internal
stresses which in turn affect the chemical reactions front kinetics.
Following [4] and taking into account the fact that the reaction is localized at the
reaction front we accept that the reaction rate ω equals to the number of oxide moles
produced in unit time per unit area is given by:

AN
ω = ω̂ 1 − exp − , (10)
RT
where ω̂ = k∗ c− c k∗ c is a rate of the direct chemical reaction, k∗ is a kinetic

constant, c− and c are partial concentrations of reacting constituents, and since the
“−” constituent is solid, one can put c− = 1.
From the mass balance it follows that
ν+ M+ ω = ρg VNΓ (11)
Then from Eq. (10) to (11) it follows that the kinetic equation (9) can be taken in the
form
ν+ M+ AN
VNΓ = k∗ c 1 − exp − (12)
ρg RT
In the case of chemical equilibrium A N = 0 and the chemical reaction rate equals
to zero. One can introduce a notion of an equilibrium gas concentration at the reaction
front as the concentration c = ceq such that
A N (c = ceq ) = ν− M− μ− + ν∗ M∗ μ∗ (ceq , T ) − ν+ M+ μ+ = 0, (13)

μ− = N · b− · N, μ+ = N · b+ · N
Under accepted assumptions the chemical affinity tensor depends on the gas con-
centration only through the dependence μ∗ (ceq , T ). Then Eq. (13) determines the
equilibrium concentration in dependence on stresses at the reaction front, tempera-
ture and material parameters.
If the equilibrium concentration is found then the current value of the normal
component of the chemical affinity tensor can be calculated as:

A N = ν∗ μ∗ (c(Γ ), T ) − μ∗ (ceq , T ) (14)
As a result the reaction front velocity can be calculated by formulae (12), (14),
where the equilibrium concentration ceq is to be found from the Eq. (13). Note that
if stresses depend on the front position then ceq also depends on the front position.
Not far from chemical equilibrium

∂μ∗
A N = ν∗ (c(Γ ) − ceq ), (15)

∂c
ceq

ν+ M+ AN ν+ ν∗ M+ ∂μ∗
VNΓ = k∗ ceq = k∗ ceq (c(Γ ) − ceq ) (16)

ρg RT ρg RT ∂c
ceq
Further we assume that μ∗ is an ideal gas chemical potential

M∗ μ∗ = η(T ) + RT ln(c)
Then not far from the chemical equilibrium

k∗ ν∗ ν+ M+
VNΓ = (c(Γ ) − ceq ) (17)
ρg
Note that the direct chemical transformation dominates and, thus, VNΓ > 0, only
if c(Γ ) > ceq . Formally, from Eq. (17) it follows that if c(Γ ) < ceq then VNΓ < 0,
i.e. the reverse chemical transformation dominates. It is known that direct chemical
reaction is accompanied by heat release and the reverse transformation would demand
an additional heat supply. We do not take into account these processes and further
just assume that the reaction is blocked, i.e. VNΓ = 0, if c(Γ ) ≤ ceq .
From Eq. (17) it follows that to describe the chemical reaction front kinetics one
has to find the equilibrium concentration ceq from Eq. (13), taking into account the
dependencies of the stresses on the chemical reaction front current position, and to
find the current concentration c(Γ ) on the chemical reaction front from the solution
of the diffusion problem.
3 Axially Symmetric Problem in Small Strain Approach
In a case of small strains free energy densities of solid constituents are represented
by quadratic dependencies
1
ρ0 f − = ρ0 f −0 (T ) + ε : C− : ε, (18)
2
1
ρg f + = ρg f +0 (T ) + ε − εch : C+ : ε − ε ch , ε ch = ε ch E, εch = g − 1
2
(19)
c0
0 R h
p R h
c0
Fig. 2 Axially-symmetric problems
Then the constitutive equations take the form
σ − = C− : ε − , σ + = C+ : (ε+ − εch ) (20)
We assume that the solid constituents are isotropic. Then the elasticity tensor:

1
C± = K ± E ⊗ E + 2μ± I− E⊗E , (21)
3
where K ± are bulk moduli, μ± are shear moduli, I is the symmetric forth rank unit
(isotropic) tensor.
As an example we consider two axially-symmetric problems: a linear-elastic cylin-
der of the radius R under external stress σ0 and an infinite linear-elastic medium with
a cylindrical hole of the radius R under internal pressure p and external stress σ0
(Fig. 2). Note that comparison of the results allows us to understand the role of the
surface curvature sign.
The oxygen surrounds the cylinder and also contains inside the cylindrical hole
(c0 is the concentration of the gas). Note that for a simplicity sake we don’t associate
the concentration of the gas inside the hole with the pressure p.
We suppose that the oxide forms a cylindrical layer of a thickness h from the body
surface. In the cylindrical coordinates (r, ϕ, z) the radial displacements in the initial
material and the oxide are given by Lame formula:
u ± = A±r + D± /r, (22)
where A± , D± are found from the boundary conditions and displacement and traction
continuity conditions on the reaction front. Strains and stresses are given by
εr± = A± − D± /r 2 , εϕ± = A± + D± /r 2
σr± = 2K ± (A± − ε±
ch
) − 2μ± D± /r 2 , (23)
σϕ± = 2K ± (A± − ε±
ch
) + 2μ± D± /r 2 ,
where ε− ch = 0, ε ch = ε ch . In the case of the cylinder u is to be finite at r = 0,

+
therefore D− = 0.
The concentration of the gas constituent at the chemical reaction front is deter-
mined by a diffusion equation. We assume that the diffusion process to be steady
state, then the diffusion equation in the cylindrical coordinate takes the form
∂ 2 c 1 ∂c
+ =0 (24)
∂r 2 r ∂r
The boundary conditions are:
c = c0 at r = R, (25)
∂c
D + k∗ ν∗ (c − ceq ) = 0 at r = R − h (cylinder), r = R + h (hole),
∂r
where h is the oxide layer thickness, D is a the diffusion coefficient.
The last condition is a mass balance at the reaction front between mass flux due
to the diffusion and the mass sink due to the chemical reaction. It is assumed that the
chemical reaction rate is proportional to the deviation of the concentration from the
equilibrium value.
4 Results
The dependencies of the oxide layer thickness on time and the reaction front velocity
on the layer thickness in the cases of the cylinder and hole are presented in Figs. 3
and 4, respectively. It is seen that the reaction front spreads faster in the case of the
cylinder than in the case of the hole of the same radius. The curves for the cylinder
and the hole are divided by the curve for the plane reaction front obtained as limit
cases of the fronts in the cases of the cylinder and the hole at infinite radius.
It should be noted that in the case of the cylinder the oxide layer thickness is
restricted by the cylinder radius. The front velocity increases drastically when the
front approaches to the center of the cylinder. This reflects the convergence of the
diffusion fluxes to the central point.
If the reaction front moves then stresses at the front change and, thus, the equilib-
rium concentration changes. Figure 5 shows the dependencies of the equilibrium con-
centration on the layer thickness at various external stresses. In the case of the cylinder
the equilibrium concentration decreases if the oxide layer thickness increases. This
in turn promotes the reaction front propagation.
Fig. 3 The dependencies of

the oxide layer thickness on
time (1) the cylinder Rc = 2,
(2) the cylinder Rc = 1, (3)
the hole Rc = 1, (4) the hole
Rc = 0.5, dashed line the case
of the plane reaction front
(oxidation of the half-space)
180
t
Fig. 4 The dependencies of

the reaction front velocity on
the oxide layer thickness for
the cylinder of the unit radius
(1) and the hole of the unit
radius (2)
V∗
In contrast, the oxide layer growth in the case of the hole leads to the increase of the
equilibrium concentration, and this may block the reaction if the gas concentration
c(Γ ) at the chemical reaction front becomes less then ceq . Also it is obviously that
c(Γ ) can not be greater than the concentration of the gas at the hole surface. Thus the
(a) c eq (b) c eq
c0 c0
h cr h h cr h
Fig. 5 The dependence of the equilibrium concentration on the oxide layer thickness in the case
of the cylinder (a) and the hole (b) at various external stresses: (1) σ0 = 0, (2) σ0 = 0.25,
(3) σ0 = −0.25
inequality ceq /c0 < 1 is the necessary condition of the reaction front propagation.
From Fig. 5b it is seen that internal stresses produced by the chemical reaction can
block the reaction if the oxide layer thickness is more than a critical value h cr . Note
that the value of the critical thickness depends on the external stress and increases
at compression and decreases at tension. Thus, the sign of the applied stress affects
the chemical reactions.
If the external stress is zero then, given the concentration c0 and the energy
parameter γ = ν∗ η + ν− M− f −0 − ν+ M+ f +0 , internal stresses can block the reaction
in the case of the cylinder if the layer thickness is less than a critical value. In this
case the equilibrium concentration at the front ceq > c0 . Thus, an initial oxide layer
on the cylinder surface may play a protective role, but its damage leads to further
chemical reactions.
The thickness of the critical layer depends on the energy parameter (Fig. 6). It is
seen that γ increasing leads to the critical layer thickness decreasing. Figure 7 reflects
the influence of the external loading on h cr . Similar to the case of the hole in the
medium, the value of the critical thickness increases at compression and decreases at
tension. Thus, given initial thickness, tension may initiate further chemical reaction.
On the other hand, compression may decelerate and even block the front propagation.
Note that the dependence ceq (σ0 ) is not monotonic. Figure 8 shows the dependence
of the equilibrium concentration ceq at the oxide layer thickness h = 0.1. One
can see that the external pressure changes the value ceq thus that for some value
σ01 ≤ σ0 ≤ σ02 the condition c0 > ceq does not hold. It means that external pressure
can block the chemical reaction. Note that without external loading the equilibrium
concentration in the case of the cylinder is less than in the case of the hole. Thus the
critical layer thickness in less in the case of the cylinder.
Fig. 6 Dependence of the hcr

critical layer thickness on γ
Fig. 7 Critical layer thickness hcr

versus the external loading
(γ = 0.0145)
0
Fig. 8 Dependencies of ceq

the ceq /c0 of the gas con- c0
stituent on the external loading
σ0 .(h = 0.1) (1) the hole, (2)
the cylinder
(1)
(2)
5 Summary
We developed a model for describing stress-assist chemical reaction front

propagation. The model includes the expressions of the chemical transformations
strains and chemical affinity tensor, and kinetic relationship that can be represented
in terms of equilibrium and current concentrations of the gas constituents. Solutions
of axially-symmetric mechano-chemical problems considering oxidation of infinite
body with a cylindrical hole and oxidation of a cylinder are presented. We study in
detail how internal and external stresses affect the reaction kinetics and predict the
reaction locking effect.
Acknowledgments This work was supported by Russian Foundation for Basic Research (Grant
10-01-00670), Sandia National Laboratories and RAS Programs for Fundamental Research.
References
1. Deal, E., Grove, A.S.: General relationship for the thermal oxidation of silicon. J. Appl. Phys.
36, 3770–3778 (1965)
2. Epstein, M., Maugin, G.A.: Thermomechanics of volumetric growth in uniform bodies. Int. J.
Plast. 16, 951–978 (2000)
3. Freidin, A.B.: On chemical reaction fronts in nonlinear elastic solids. In: Indeitsev, D.A.,
Krivtsov, A.M. (eds.) Proceedings of XXXVI International Summer School-Conference APM’
2009, pp. 231–237. Institute for problems in, mechanical Engineering, St.Petersburg (2009)
4. Glansdorff, P., Prigogine, I.: Thermodynamic theory of structure, stability and fluctuations.
Wiley, New York (1971)
5. Guillou, A., Ogden, R.W.: Growth in soft biological tissue and residual stress development.
In: Holzapfel, G.A., Ogden, R.W. (eds.) Mechanics of Biological Tissue. Springer, Heidelberg
(2006)
6. Gusev, E., Lu, H., Gustafsson, T., Garfunkel, E.: Growth mechanism of thin silicon oxide films
on Si(100) studied by medium-energy ion scattering. Phys. Rev. B52, 1759–1775 (1995)
7. Kelly, S.T., Clemens, B.M.: Moving interface hydride formation in multilayered metal thin
films. J. Appl. Phys. 108(1), 013521 (2010)
8. Kikkinides, E.S.: Design and optimization of hydrogen storage units using advanced solid
materials: General mathematical framework and recent developments. Comput. Chem. Eng.
35, 1923–1936 (2011)
9. Klisch, S.M., VanDyke, T.J.: A theory of volumetric growth for compressible elastic biological
materials. Math. Mech. Solids 6, 551–575 (2001)
10. Knowles, J.K.: On the dissipation associated with equilibrium shocks in finite elasticity. J.
Elast. 9, 131–158 (1979)
11. Larche, F., Cahn, J.W.: A linear theory of thermochemical equilibrium of solids under stress.
Acta Metall. 21, 1051–1063 (1973)
12. Larche, F., Cahn, J.W.: A nonlinear theory of thermochemical equilibrium of solids under
stress. Acta Metall. 26, 53–60 (1978)
13. Larche, F., Cahn, J.W.: Thermochemical equilibrium of multiphase solids under stress. Acta
Metall. 26, 1579–1589 (1978)
14. Larche, F., Cahn, J.W.: The effect of self-stress on diffusion in solids. Acta Metall. 30, 1835–
1845 (1982)
15. Lee, E.H.: Elastic-plastic deformation at finite strains. ASME J. Appl. Mech. 36, 1–8 (1969)
16. Lubarda, V.A.: Constitutive theories based on the multiplicative decomposition of deformation
gradient: Thermoelasticity, elastoplasticity, and biomechanics. Appl. Mech. Rev. 57(2), 95–108
(2004)
17. Muhlstein, C.L., Stach, E.A., Ritchie, R.O.: A reaction-layer mechanism for the delayed failure
of micron-scale polycrystalline silicon structural films subjected to high-cycle fatigue loading.
J. Acta Mater. 50, 3579–3595 (2002)
18. Rosencher, E., Straboni, A., Rigo, S., Amsel, G.: An 18 O study of the thermal oxidation of
silicon in oxygen. Appl. Phys. Lett. 34, 254–257 (1979)
19. Taber, L.A.: Biomechanics of growth, remodeling and morphogenesis. ASME Appl. Mech.
Rev. 48, 487–545 (1995)
20. Trimaille, I., Rigo, S.: Use of 18 O isotopic labeling to study the thermal dry oxidation of silicon
as a function of temperature and pressure. Appl. Surf. Sci. 39, 65–80 (1989)
21. Wu, C.H.: The role of Eshelby stress in composition-generated and stress-assisted diffusion.
J. Mech. Phys. Solids 49, 1771–1794 (2001)

Surface Effects in Solid Mechanics PDF

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Surface Effects in Solid Mechanics PDF

Hochgeladen von

Copyright:

Verfügbare Formate

Advanced Structured Materials

For further volumes:

Surface Effects in Solid

ISSN 1869-8433 ISSN 1869-8441 (electronic)

Library of Congress Control Number: 2013931510

Springer-Verlag Berlin Heidelberg 2013

Printed on acid-free paper

Springer is part of Springer Science?Business Media (www.springer.com)

The collection of papers ‘‘Surface Effects in Solid Mechanics: Models, Simulations

Magdeburg, St. Petersburg, August 2012 Holm Altenbach

Mathematical Study of Boundary-Value Problems of Linear

On the Influence of Residual Surface Stresses on the Properties

On the Isotropic Elastic Properties of Graphene Crystal Lattice . . . . . 33

A Comparison of Atomistic and Surface Enhanced

Surface Mechanics and Full-Field Measurements: Investigation

Effect of a Type of Loading on Stresses at a Planar Boundary

Surface Stress in an Elastic Plane with a Nearly Circular Hole. . . . . . 81

Glass Spheres: Functionalization, Surface Modification

Spectral Properties of Piezoelectric Bodies with Surface Effects . . . . . 105

Stability and Structural Transitions in Crystal Lattices . . . . . . . . . . . 123

Mathematical Modeling of Phenomena Caused by Surface Stresses

Buckling of a Supported Annular Plate

On the Modeling of Surface and Interface Elastic Effects

On Kinetics of Chemical Reaction Fronts in Elastic Solids . . . . . . . . . 181

Holm Altenbach Lehrstuhl für Technische Mechanik, Institut für Mechanik,

Valentin A. Gorodtsov A. Ishlinsky Institute for Problems in Mechanics RAS,

Artem Panchenko St. Petersburg State Polytechnical University, 29 Polite-

Holm Altenbach, Victor A. Eremeyev and Leonid P. Lebedev

Abstract Following [1, 2] a mathematical investigation of initial-boundary and

H. Altenbach (B) · V. A. Eremeyev

H. Altenbach and N. F. Morozov (eds.), Surface Effects in Solid Mechanics, 1

2 Basic Relations of Linear Elasticity with Surface Stresses

First we consider the problems with mixed boundary conditions. So

Fig. 1 Elastic body with t t0 tS

For simplicity, we restrict ourselves to an isotropic material. The constitutive

In equilibrium, the dynamic equation (1) changes to

u|Ωu = 0, n · σ |Ωt = t0 , (n · σ − ∇ S · τ )|Ω S = t0 , (7)

If Ω S = Ω then parts Ωu and Ωt are absent. In what follows for simplicity we

W (ε) ≥ c1 ε : ε, U (ε̃ε ) ≥ c2ε̃ε : ε̃ε , ∀ε, ε̃ε , c1 , c2 > 0 (11)

3λ + 2μ > 0, μ > 0, λ S + μ S > 0, μ S > 0. (12)

The admissible range of surface material parameters is also studied in [11].

3 Existence Theorems of Weak Solutions

By Korn’s inequality on C(2)

uW 1,2 (V ) ≤ C1 ue

Let Ω S consists of a finite number of sufficiently smooth surfaces with smooth

uW 1,2 (Ω S ) ≤ C2 ue

with some constant C2 that does not depend on u ∈ E. Here

The element u of E is an element of W 1,2 (V )3 and so by the trace theorem,

where u0 is a uniquely defined element of E. So Eq. (9) reduces to the equation

3.1.1 Existence Theorem: Free Boundary

3.1.2 On the Finite Element Method

Let us consider an important case of dynamical problems: the eigenoscillation prob-

holds true. Moreover,

Applying the Gauss–Ostrogradsky theorem we get

Using these definitions we prove the following theorem:

4 Spectrum and Stiffness of an Elastic Body

∇ · σ = −ρω2 w, x ∈ V ; w|Ωu = 0, n · σ |Ω S = 0; (23)

Let us note that we may consider E as a subset of H, and H0 as a subset of E

4.1 Rayleigh Variational Principle

On the orthogonality of the eigensolutions established in Theorem 1.5 the Rayleigh

R(w) is called Rayleigh’s quotient. Conversely, a stationary point of R(w) is an

4.2 The Least Eigenfrequency

The squared least eigenfrequency ωmin is determined as the infimum of R(w):

uW 1,2 (V ) ≤ C1 ue

uW 1,2 (Ω S ) ≤ C2 ue

d f [v1 , v2 , . . . , vk−1 ] = inf w2H ,

w2E ≥ inf w2E .