Comput Mech (2018) 61:687–697

https://doi.org/10.1007/s00466-017-1474-4

ORIGINAL PAPER

A methodology for modeling surface effects on stiff and soft solids

Jin He1,2 · Harold S. Park2

Received: 4 May 2017 / Accepted: 17 August 2017 / Published online: 2 September 2017
© Springer-Verlag GmbH Germany 2017

Abstract We present a computational method that can 1 Introduction

be applied to capture surface stress and surface tension-
driven effects in both stiff, crystalline nanostructures, like The surfaces of solids exhibit different mechanical behavior
size-dependent mechanical properties, and soft solids, like as compared to their bulk. For both soft and hard (crystalline)
elastocapillary effects. We show that the method is equiva- solids, this is due to the fact that surface atoms have a dif-
lent to the classical Young–Laplace model. The method is ferent bonding environment, and specifically fewer bonding
based on converting surface tension and surface elasticity on neighbors than do atoms that lie within the material bulk
a zero-thickness surface to an initial stress and corresponding [1]. These surface effects, which are typically negligible for
elastic properties on a finite thickness shell, where the con- macroscopic solids, manifest themselves in different ways
sideration of geometric nonlinearity enables capturing the and at different length scales depending on whether the solid
out-of-plane component of the surface tension that results is soft or stiff.
for curved surfaces through evaluation of the surface stress In crystalline (stiff) solids, surface stress effects have
in the deformed configuration. In doing so, we are able to been shown over the past two decades to lead to interesting,
use commercially available finite element technology, and non-bulk mechanical properties in nanostructures, including
thus do not require consideration and implementation of the size-dependent mechanical properties [2–5], unique multi-
classical Young–Laplace equation. Several examples are pre- functionality, like phase transformations [6], shape memory
sented to demonstrate the capability of the methodology for and pseudoelasticity [7,8] and non-bulk plastic deformation
modeling surface stress in both soft solids and crystalline mechanisms [9,10]. Much of the work that has been done to
nanostructures. capture surface effects on nanomaterials either computation-
ally [11–15] or analytically [16–20] is based on the Young–
Keywords Surface tension · Finite element · Soft tissue · Laplace (Y–L) model for incorporating the effect of the sur-
Bending · Buckling face tension in solids, though alternative computational meth-
ods based on decomposing the surface and bulk energies [21–
26] have also been proposed. Named after Thomas Young and
Electronic supplementary material The online version of this
Pierre Simon Laplace, the Y–L equation was originally devel-
article (https://doi.org/10.1007/s00466-017-1474-4) contains
supplementary material, which is available to authorized users. oped in the early nineteenth century [27,28], and describes
the pressure difference across a curved interface between two
B Jin He fluids due to the surface tension. The Y–L equation has also
jinhe@njtech.edu.cn
been demonstrated to be effective in modeling surface effects
Harold S. Park on not only fluids, but also crystalline solids [29–31].
parkhs@bu.edu
Similarly, there has been significant recent interest in so-
1 School of Mechanical and Power Engineering, Nanjing Tech called elastocapillary mechanics, where surface tension due
University, Nanjing 211816, China to fluid-structure interactions has been used to change the
2 Department of Mechanical Engineering, Boston University, mechanical behavior and properties of soft solids, like gels
Boston, MA 02215, USA [32,33]. While the best-known example of surface tension in

123
688 Comput Mech (2018) 61:687–697

fluid mechanics is likely that of deforming liquid droplets, respect to the deformed configuration of the shell through
there has been interest in using it to deform solid structures geometric nonlinear analysis.
and the relevant reviews are in Refs. [34,35]. This interest We present a range of numerical examples demonstrat-
in using elastocapillary forces to deform soft structures has ing the ability of the method to capture surface effects on
emerged since for these systems the elastocapillary number, both stiff and soft solids. We additionally demonstrate that
which is defined as τ 0 /(μl), where τ 0 is the surface tension, time-dependent material behavior, through viscoelasticity,
μ is the shear modulus and l is a characteristic length, is close and strain-dependent surface stresses, which are critical for
to unity, implying that elastocapillary effects can be substan- stiff nanomaterials, and have recently become of interest in
tial for these soft materials. For example, recent work has the mechanics of soft solids [33], can be easily accounted
highlighted the important role surface tension plays in the for through the standard options in ANSYS and COMSOL.
contact mechanics and adhesion of soft solids [36–38], iden- Overall, directly incorporating surface tension-related effects
tifying its effect on wetting, large deformations, and phase into a commercial FE code should enhance the ability of
separation [39], using liquid inclusions to stiffen soft solids scientists and engineers to model, design, study, and under-
[40], and inducing Rayleigh-Plateau elastocapillary instabil- stand mechanical problems involving surface effects. All
ities in soft solids [41]. input files with detailed modeling procedures, geometric and
Finite element (FE) models of surface tension in the con- mesh sizes, as well as material properties are provided in the
text of soft solids have also emerged within the past decade. electronic supplemental material.
Examples include the works of Saksono and Peric [42], Javili
et al. [25], Henann and Bertoldi [43], Seifi and Park [44,45], 2 Method overview
and Wang and Henann [46]. As in the case of crystalline
solids, all of these works have captured surface tension effects In this section, we first present an overview of our proposed
using the Y–L model. Recently, a method without using the methodology. After doing so, we also present a compari-
Y–L model based on the commercial FE software ANSYS son between the proposed approach and the Y–L equation,
[47] was proposed [48], in which the surface tension is taken to establish that the proposed approach reproduces the Y–L
into account by utilizing a special feature of the structural equation.
surface element type SURF153 or SURF154 provided by To consider the effects of surface stress, we show in Fig. 1a
ANSYS, where the surface tension induced in-plane force is a solid with a surface, in which the surface stress acts on
applied to the surface nodes by assuming the surface nodes the zero-thickness film enveloping the core. This kinematic
are coplanar. However, this surface element does not account model (solid with a zero thickness surface) has been used
for surface curvature [47], and furthermore, such surface ele- previously in both the Y–L model as well as the well-known
ments are generally not used in other commercial FE codes. Gurtin-Murdoch theory of surface elasticities [30,50]. How-
Our objective in the present work is to present a FE model ever, the notion of a zero thickness surface presents some
that captures surface effects, and can be applied to either stiff challenges for formulating a core-shell surface stress model,
or soft solids that undergo arbitrary deformations, including as shown in Fig. 1b, and therefore our method is based on
bending and surface curvature, using commercially used FE converting surface tension and surface elasticity on a zero-
codes. One motivation for this work is the realization that the thickness surface to an initial stress and the corresponding
out-of-plane force induced from the in-plane surface stress on elastic properties on an equivalent finite thickness shell.
a curved interface of a solid, described by the Y–L equation, The constitutive equation for the surface stress tensor τi j
can alternatively be captured using geometric nonlinearity. in Fig. 1a for the zero-thickness surface, in the first order
In doing so, we arrive at a formulation that can exploit con- approximation, can be written as [51,52]
ventional FE methods. We additionally demonstrate in 2D
that the expression of the out-of-plane force induced from τi j = τi0j δi j + Si jkl kl (i, j, k, l = 1, 2), (1)
the in-plane surface stress without using the Y–L equation
is identical to that obtained using the Y–L equation. Any FE
where indicial notation is used, τi0j is the surface tension, δi j
code with shell and solid type elements, as well as the fea-
is the Kronecker delta, Si jkl is the surface elasticity, and kl
tures of initial stress and geometric nonlinearity, can be used
is the Green strain. The Green strain is written as
to capture surface effects on stiff and soft solids, which are
demonstrated by using the commercially-available FE codes 1 T
ANSYS [47] and COMSOL [49]. Without explicitly using = (F F − I), (2)
2
the Y–L equation, the methodology implicitly mimics the
out-of-plane force induced from the in-plane surface stress where F is deformation gradient and I is the identity matrix.
through formulating the surface stress induced force with The Green strain is introduced here since the following com-
putations are based on finite deformation theory.

123
Comput Mech (2018) 61:687–697 689

(a) ⎧ 0
(b) ⎪ σi j = τi0j δi j /t (i, j = 1, 2) (4a)
⎪
⎪
⎪
⎪
τ22 σ22 ⎪
⎪ E 1 = (S1111 S2222 − S1122 2
)/(S2222 t) (4b)
⎪
⎪
σ21 ⎨
τ21 E 2 = (S1111 S2222 − S1122 )/(S1111 t)
2
(4c)
t
x2 τ12 x2 σ12 ⎪
⎪ G 12 = S1212 /t (4d)
⎪
⎪
x1 τ11 σ11 ⎪
⎪
x1 ⎪ ν12 = S1122 /S2222
⎪ (4e)
⎪
⎩
Film Core Shell Core
ν21 = S1122 /S1111 . (4f)

(c) τ Enlarged view of Δs

Since plane stress is assumed, t needs to be far less than the
smallest dimension of the solid. When the surface is isotropic,
Flat τ(tA,nA) Δs
τ11
0 = τ 0 = τ 0 and Eq. 1 is simplified as
22
A
τ t B
τ(tB,nB) τi j = τ 0 δi j + λs kk + 2μs i j (i, j, k = 1, 2), (5)
Curved Δθ R
n Δs where λs and μs are referred as surface Lamé constants [50],
Δs→ds, Δθ→dθ which can be obtained by imposing the symmetries of the
Q=κτ τ
π-dθ isotropic surface [52,53]
The Y-L
dθ
equation τ

based λs = S1122 (6a)

τdθ=τ(dθ/ds)ds=κτds
methods μs = (S1111 − S1122 )/2. (6b)
Fig. 1 The solid with the surface. a Core-film model; b Core-shell
Substituting Eq. 6 into Eq. 4 leads to the Young’s modu-
model; c Difference between the Y–L equation-based methods and the
present methodology lus E and Poisson’s ratio ν of the equivalent isotropic shell
expressed with the surface Lamé constants [48]

On the right hand side of Eq. 1, the first term corresponds E = E s /t (7a)
to the strain-independent part of the surface stress and the ν = λs /(λs + 2μs ), (7b)
second term corresponds to the strain-dependent part of the
surface stress, which represents the first order approximation where the surface Young’s modulus E s is
of the Shuttleworth effect.
When Si jkl is orthotropic and Si jkl = Skli j , Eq. 1 is E s = 4μs (λs + μs )/(λs + 2μs ). (8)
analogous to Eq. 3 below, which is written in the current
coordinate system and describes the stress-strain relation- The core of the solid conforms to the laws of classical solid
ship of an orthotropic, finite thickness shell in Fig. 1b under mechanics and can be modeled as appropriate for the material
plane stress with initial stress σi0j system under consideration, i.e. linear elastic, hyperelastic,
viscoelastic, etc. The whole shell-core structure is analyzed
⎡ ⎤ ⎡ 0 ⎤ as a conventional solid, which can be subject to arbitrary
σ11 σ11
⎣ σ22 ⎦ = ⎣ σ 0 ⎦ loads and deformations, including bending.
22
The methodology described above is sufficiently simple
σ12 0
⎡ ⎤⎡ ⎤ that it can be implemented in any commercially-available
E 1 E 2 ν12 0 11 FE package that contains both solid (for the bulk) and shell
1 ⎣ E 1 ν21 E 2 ⎦ ⎣ 22 ⎦ ,
+ 0 elements (for the surface). We use both ANSYS [47] and
1 − ν12 ν21
0 0 (1 − ν12 ν21 )G 12 12 COMSOL [49] in the present work to demonstrate the ver-
(3) satility and ease of implementation of the proposed method.

where E 1 , E 2 are the Young’s moduli in the x1 , x2 directions 2.1 Comparison with Y–L equation
respectively, ν12 , ν21 are the major, minor Poisson’s ratio
in the x1 x2 plane respectively, G 12 is the shear modulus in We now present an analysis of the proposed approach, and
the x1 x2 plane, and σi j (i, j = 1, 2) is the second Piola- demonstrate its equivalence with the Y–L model. The Y–L
Kirchhoff stress, which is work conjugate to the Green strain. model has the feature that when the surface is flat, the sur-
The relationship between the surface stress and the strain face tension results in an in-plane component only. However,
in Eq. 1 is alternatively expressed by dividing by the shell when the surface is curved, the surface tension has compo-
thickness t, which leads to nents both in and out of the plane of the surface. According

123
690 Comput Mech (2018) 61:687–697

to the Y–L equation, the surface stress induced out-of-plane surface stress effects on the mechanical properties of crys-
force per area Q i j for curved surfaces is [31,51] talline nanostructures. As previously mentioned, we use both
ANSYS [47] and COMSOL [49] to demonstrate the ease
Q i j n i n j = κi j τi j (i, j = 1, 2), (9) of use and implementation in widely used commercial FE
packages. The input files with detailed modeling procedures,
where n i is the normal to the surface and κi j is the curvature geometric and mesh sizes, as well as material properties are
tensor. When Eq. 1 is formulated in the coordinate system provided in the Online Supplementary Materials.
for the deformed configuration, the out-of-plane force com-
ponent described by the Y–L equation is implicitly realized 3.1 Rayleigh-plateau instability
by formulating the surfaces stresses in the deformed con-
figuration. Explanations are presented in a 2D schematic One of the most interesting findings with regards to elas-
illustration (Fig. 1c), where the surface of a solid domain tocapillary effects on the deformation of soft solids was
is initially flat and then curved by an external force. In 2D, the recent experimental report of surface-tension-induced
Eq. 1 becomes Rayleigh-Plateau instability in soft gels [41]. In this exper-
iment, researchers increased the elastocapillary number on
a soft gel until the gel broke up in a manner similar to the
τ = τ 0 + S, (10)
classical Rayleigh-Plateau instability in fluids. Subsequent
researchers have investigated this effect further [54–58].
where S is surface elasticity and τ , τ 0 ,  are the surface
We simulated this problem using ANSYS, by considering
stress, the surface tension, the strain in t direction lying in
a soft, incompressible rod with shear modulus μ = 12 Pa and
the surface respectively. t and n are the directions tangential
radius R = 150 µm. Both ends of the rod are clamped and the
and normal to the surface respectively. In 2D, Eq. 9 becomes
rod has an aspect ratio of 200 in the ANSYS model, where
8-node element SOLID185 and 4-node element SHELL181
Q = κτ, (11) are used for modeling the core and the shell respectively. The
material model of the core is modeled as Neo-Hookean and
where Q is the out-of-plane force per area in n direction and the shell is linear elastic. The initial shell-curvature effect is
κ is curvature as shown in the bottom left of Fig. 1c. considered by setting KEYOPT(5) of the SHELL181 element
Consider a small segment of the curved surface Δs shown to 1 in order to improve the accuracy in modeling the strain in
in the middle left of Fig. 1c and the enlarged view of the the cross-sectional direction. The numbers of elements along
segment on the right side of Fig. 1c. The segment is subject the length and cross-sectional directions are 1000 and 28
to surface stress τ (t A , nA ) and τ (t B , nB ) at points A and respectively. Surface stress resides in a few atomic layers of
B respectively. Since the segment is small, the segment can a crystal according to molecular dynamic simulations [59]. In
be approximated as an arc of a circle with radius R. The the following computations, the shell thickness t is arbitrarily
central angle of the arc is Δθ = Δs/R and the curvature chosen to be 0.1 nm. Choosing different values of t does not
of the arc is κ = 1/R = Δθ/Δs. As the segment becomes lead to discernible results as long as t is far less than the
infinitely small, Δs → ds, Δθ → θ , point A approaches critical dimension of the material. The critical dimension of
point B, and the strain at A has the same magnitude as the this problem is 150 µm.
strain at B, leading to the same magnitude of the two surface We show in Fig. 2a the sinusoidal variation of the displace-
stresses τ at one point. The angle of the two surface stresses ment field along the axial direction, which is characteristic
is π − dθ . The net force on the the segment ds is τ dθ = of the elastocapillary Rayleigh-Plateau instability [58] of the
τ (dθ/ds)ds = κτ ds, in direction of n. The net force divided rod, if the surface tension is 11 mN/m. In contrast, if the sur-
by ds is the out-of-plane force per area κτ , identical to the face tension is slightly smaller, i.e. 10 mN/m as in Fig. 2a, no
2D Y–L equation shown in Eq. 11. This net force is captured Rayleigh-Plateau instability occurs. Figs. 2a and b present
in the present work through geometric nonlinearity, which about half of the rod including the contours of the displace-
ensures that the surface stress calculations are performed in ment magnitude, where the deformations are magnified 105
the deformed configuration. times for better illustration. In accordance with previous ana-
lytic studies of the experiments of Mora et al. [55–58], the
Shuttleworth effect is neglected in ANSYS, which is ensured
3 Numerical examples by making the shell Young’s modulus a very small value.
The wavelength associated with the instability is plotted
We now present numerical examples demonstrating the per- as a function of the surface tension in Fig. 2c, where the
formance of the proposed approach for problems relevant to curve obtained from ANSYS agrees well with the curves
soft solids, involving elastocapillary phenomena, as well as from previous analytic studies [41,58]. The wavelengths

123
Comput Mech (2018) 61:687–697 691

(a) from ANSYS are obtained by performing Fourier transform

(FT) of the radial displacements along the surface. As an
example, the method for determining the wavelength with
One of the
the surface tension of 12 mN/m is shown in Fig. 2d. The
clamped ends
amplitude of the FT of the radial displacement reaches a peak
when the inverse of the length is about 0.33 mm−1 , as shown
0 6.8 nm in the inset of Fig. 2d, and thus the wavelength is determined
(b) to be 1/0.33 ≈ 3 mm. The surface tension threshold calcu-
lated from ANSYS is between 10.6 and 10.8 mN/m, which
One of the matches the theoretical threshold 6μR = 10.8 mN/m quite
clamped ends closely [41,58].

0 2.9 nm 3.2 Stiffening soft solids with liquid inclusions

(c)
12
Another interesting elastocapillary effect that has recently
ANSYS been reported is that a soft solid embedded with liquid inclu-
10
Theory, from Ref. [41] sions may be stiffer than its counterpart without the inclusions
Theory, from Ref. [58] due to the resulting surface tension that acts on the solid
Wavelength (mm)

8 Theoretical threshold, from Refs. [41,58]

[40,46]. The 3D model of a droplet in a large solid domain
under far-field strains is shown in Fig. 3a, where 1/8 of the
6 entire domain is studied due to symmetry. The major and
minor diameters of the deformed droplet in Fig. 3b are l and
4 w respectively, and the liquid is assumed to be incompress-
ible. The Young’s modulus of the incompressible soft solid
2 is 1.7 kPa and the surface tension is τ 0 = 3.6 mN/m [40].
The 3D Solid Mechanics interface and Shell interface are
0 used to model the core and the shell respectively in COM-
10 10.8 12 14 16 18 20 SOL since the incompressible fluid can be implemented more
Surface tension (mN/m) easily in COMSOL than in ANSYS. In modeling the large
(d) domain, the ratio of the edge size of the cube to the radius
1 of the sphere is chosen to be 10. Approximately 1.2 × 105
FT amplitude
(arb. unit)

quadratic tetrahedral elements are generated in the simula-

2
tion. Default settings are used for the 3D Solid Mechanics
Radial displacement (arb. unit)

0 interface and Shell interface.

1
0.3 1.5 The aspect ratios l/w under the far-field strains 11 =
Length-1 (mm-1)
5.6% and 22 = −1.5% calculated from COMSOL are com-
pared with the experimental and theoretical results [40] in
0 Fig. 3c, where we initially neglect the Shuttleworth effect in
COMSOL. The present results are consistent with the exper-
iment and the theory. Both the results from COMSOL and
-1
the theory predict that the aspect ratios of the liquid inclu-
sion become smaller when the surface tension is considered,
-2 which means that the solid is stiffened by the surface ten-
0 5 10 15 20 25 30 sion. We also verify that if the surface tension τ 0 = 0, the
Axial location (mm) COMSOL results match the analytic theory.
The scattered experimental data in Fig. 3c are due to
Fig. 2 Rayleigh-Plateau instabilities in soft rods. a The instability the experimental complexities involving coating silicone
occurs when the surface tension is 11 mN/m. b The instability does
not occur when the surface tension is 10 mN/m. c Wavelength due to gel embedded with liquid drops on a stretchable sheet and
the instability versus surface tension; d The radial displacement along attaching fluorescent nanoparticles for displacement field
the surface and the inset shows the corresponding FT when the surface measurements [40]. In Ref. [40], the experiment is performed
tension is 12 mN/m under different combinations of 11 and 22 , which shows that

123
692 Comput Mech (2018) 61:687–697

The Shuttleworth effect, whereby the surface stress is

(a) (b)
strain-dependent, has been used widely in the mechanics
of crystalline nanostructures, but its effect on the mechan-
ics of soft solids subject to surface tension has rarely been
considered [33]. Here, we consider the influence of the Shut-
tleworth effect by assuming different combinations of surface
Young’s modulus and surface Poisson’s ratio (E s , ν), with
the COMSOL results shown in Fig. 3d. The combination
(0,0.01) corresponds to the case without the Shuttleworth
(c) effect, which is also shown in Fig. 3c. The Shuttleworth effect
increases the stiffening effect of the soft solid for all the com-
binations: (5τ 0 ,0.01), (5τ 0 ,0.49), (10τ 0 ,0.01), (10τ 0 ,0.49),
though the change in inclusion aspect ratios due to the Shut-
tleworth effect appears to be smaller than 5%.

3.3 Surface tension effects on relaxation of a viscoelastic

half-space subject to a point force

Most of the studies related to elastocapillary effects on soft

solids have not considered the time-dependent deformation
of the solid, or viscoelastic effects [33]. Viscoelasticity can
have a significant effect as most soft solids exhibit varying
degrees of rate-dependent material behavior. Furthermore,
(d) consideration of such effects is a useful demonstration of
the different types of material behavior that can be easily
considered by implementing the surface tension model into
a commercial FE package.
Therefore, we consider a computational model following
the theoretical solution obtained by Hui and Jagota for the
effect of the surface tension on the relaxation of an incom-
pressible viscoelastic half-space excited by a point force [61].
We model the problem using a two-dimensional axisymmet-
ric (2DA) approach, as illustrated in Fig. 4a, b.
The short and long time shear modulus of the half-space
are μ0 = 50 Pa and μ∞ = 200 Pa respectively. The relax-
ation time is tr = 1 ms and the surface tension is τ 0 =
1 mN/m. The elastocapillary lengths are τ 0 /μ0 = 20 µm
and τ 0 /μ∞ = 5 µm. A point force F = 1 nN in the z direc-
Fig. 3 The soft solid stiffened by a liquid inclusion. a 1/8 of the 3D tion is initially applied at location r = 0 for a sufficient
model; b Deformed configuration in X 1 X 2 plane; c Size-dependent
time until the half-space is fully relaxed (Fig. 4a). After the
aspect ratio due to the surface tension; d The shuttleworth effect
force is suddenly removed, the z-directional displacements
along the surface gradually relax to zero (Fig. 4b). The 2DA
Solid Mechanics interface and Membrane interface are used
the curves of the size-dependent aspect ratios agree reason- to model the core and the shell respectively in COMSOL.
ably with the theory. The maximum relative difference of the Default settings are used for the 2DA Solid Mechanics inter-
aspect ratios in Fig. 3c between the COMSOL and the theory face and Membrane interface. In modeling the half-space, the
is less than 0.9%. Further mesh refinement of the COMSOL radius of the semicircle is chosen to be 100τ 0 /μ0 , which is
model yields to less than 0.2% difference in the aspect ratio. 100 times the elastocapillary length. Approximately 1.2×105
One of the possible causes of the slight difference between triangular elements are generated in the simulation. The anal-
COMSOL and the theory is that small strain is assumed in ysis is performed in two steps – a stationary analysis with the
the theory [60] while nonlinear kinematics are considered in the point force applied followed by a time dependent analysis
the COMSOL model. with the point force removed.

123
Comput Mech (2018) 61:687–697 693

(a) (b) (a) (b)

z F z
r r

Fb Fb
(c)
Case 2; ANSYS
Axis Axis 0.0 Case 3; ANSYS
Case 2; COMSOL
(c)
Case 3; COMSOL

Transverse displacement (nm)

150
COMSOL, t=tr Cases 2, 3; theory,
0.2
COMSOL, t=3tr from Ref. [62]
120
z-directional displacement (nm)

Theory, t=tr, from Ref. [61]

Theory, t=3tr, from Ref. [61] 0.4 Case 0; ANSYS
90 Case 1; ANSYS
150
(0,0.01) Case 0; COMSOL
(10τ0,0.01) 0.6 Case 1; COMSOL
60 (10τ0,0.49) Cases 0, 1; theory, from Ref. [62]

0.8
30 0 0 50 100 150 200
0 100
Axial location (nm)
0 (d)
0 20 40 60 80 100 Case 0; theory, from Ref. [17]
Radial location (μm) Case 1; theory, from Ref. [17]
0.0 Case 2; theory, from Ref. [17]
Fig. 4 The relaxation of the viscoelastic half-space excited by the point Case 3; theory, from Ref. [17]
Transverse displacement (nm)

force. a The 2DA model with the point force applied; b The 2DA model
with the point force removed; c z-directional displacements along the
0.4
surface at t = tr and 3tr . The inset shows the shuttleworth effect at
t = tr

0.8
The z-direction displacements at z = 0 calculated from
COMSOL agree with the theory [61] at time t = tr and 3tr Case 0; ANSYS Case 0; COMSOL
Case 1; ANSYS Case 1; COMSOL
in Fig. 4c, where the Shuttleworth effect is neglected. The 1.2 Case 2; ANSYS Case 2; COMSOL
Shuttleworth effect calculated from COMSOL is presented in Case 3; ANSYS Case 3; COMSOL
the inset of Fig. 4c, in which the z-directional displacements 0 50 100 150 200
at t = tr with (E s , ν) equal to (10τ 0 ,0.01) and (10τ 0 ,0.49) Axial location (nm)
are almost the same as those without the Shuttleworth effect.
Fig. 5 The surface effects on static bending NWs. a CF boundary con-
ditions; b CC boundary conditions; c Transverse displacements of the
3.4 Surface stress effects on crystalline nanowires in CF NW; d Transverse displacements of the CC NW. Case 0: no surface
bending effects; Case 1: effect of the surface tension only; Case 2: effect of the
surface elasticities only; Case 3: both effects
We now focus on validating the proposed approach in captur-
ing the effects of surface stress on the mechanical behavior examples. The number of elements in length direction is 100
of stiff crystalline nanostructures, or nanowires (NWs). For both in ANSYS and COMSOL. The numbers of elements
NWs, it has been well-established that surface stress effects in diameter direction are 16 and 8 in ANSYS and COM-
may influence the bending behavior of static NWs, which has SOL respectively. The absolute values of the surface tension
been shown experimentally, theoretically, and computation- and the surface Young’s modulus of face-center cubic crystal
ally [2–4,17,62,63]. surfaces are on the order of 10−1 to 101 N/m [52]. In Fig. 5,
The bending NWs in clamped-free (CF) and clamped- τ 0 = 1 N/m and E s = 10 N/m are used for illustration pur-
clamped (CC) boundary conditions are in Fig. 5a, b respec- poses. The Young’s modulus and the Poisson’s ratio of the
tively, which are modeled by using ANSYS and COMSOL NW core, same as the bulk material properties, are assumed
in 3D. The cross section of the NWs is circular and the diam- to be E m = 78 GPa and νm = 0.42 respectively. The surface
eter is D = 10 nm. The ratio of length to diameter is 20. tension only (Case 1), the surface elasticities only (Case 2),
The element types are the same as those in the previous 3D both effects (Case 3), and neither effects (Case 0) are studied.

123
694 Comput Mech (2018) 61:687–697

Under a point transverse force Fb = 0.01 nN applied at (a) (b)

x
the free end, the transverse displacements of the CF NW y
along the NW axial direction are shown in Fig. 5c, where d v
the displacement curves are categorized into two groups. h Fa
The transverse displacements calculated from ANSYS and L Fa+2τd
2τd(d2v/dx2)
COMSOL coincide with the theory [62] if the effective flex-
ural rigidity of the circular NW is formulated as (E I )∗ = (c) CF (d) CC
π E m D 4 /64 + π E s D 3 /8 [17]. Fa Fa
The surface tension does not have a significant influence
on the bending behavior of the CF NW and the detailed expla- (e) 12
nations are in Ref. [62]. The CF NW appears stiffer under the

Axial force (nN)

influence of the surface elasticities since the surface surface 8
Young’s modulus is positive. Under Fb = 1 nN applied at the
CF
midspan of the CC NW, the transverse displacements of the 4 Without surface effects
CC NW are in Fig. 5d. Good agreements are found between With surface effects
the results from ANSYS, COMSOL, and the theory [17], 0
0 2 4 6 8 10
where both the surface tension and the surface elasticities
Transverse displacement at the tip (nm)
increase the stiffness of the CC NW.
(f)
160

Axial force (nN)

3.5 Surface stress effects on crystalline nanowires in 120
buckling 80
CC
40 Without surface effects
Our final example examines surface stress effects on the
0 With surface effects
buckling of crystalline nanowires. Before presenting the
results calculated by using the proposed methodology, we 0 2 4 6 8 10
show theoretical derivations of the critical loads of axially Transverse displacement at the midspan (nm)
buckling NWs under the influence of the surface effects based (g)
40
Critical buckling load (nN)

on a simplified 2D model. A CF NW, with height, depth, and

CC Theory, this work
length denoted as h, d, and L respectively, shown in Fig. 6a, 30 Atomistic simulations, from Ref. [67]
is under an external compressive force Fa . The dimensions Theroy, from Ref. [18]
satisfy L  d  h so that the bending in x y plane can be 20
considered as a 2D plane strain problem as shown in Fig. 6b. 10
In equilibrium, the total axial compressive load on the NW
is the summation of the external force Fa and the counteract- 0
20 30 40 50
ing force balancing the in-plane surface stress induced force Aspect ratio L/a
2τ d. Denote the transverse displacement as v. When the NW
is under small bending, the out-of-plane force component, Fig. 6 The surface effects on buckling NWs. a CF NW satisfying L 
derived from the Y–L equation, is 2τ d(d2 v/dx 2 ) [16–18], d  h; b 2D plane strain model; c CF NW with square cross section;
d CC NW with square cross section; e Axial force versus transverse
which can be treated as a distributed transverse force per displacement at the tip of the CF NW with a = 10 nm and L/a = 10;
length. Thus, the differential equation of the axis of the bend- f Axial force versus transverse displacement at the midspan of the CC
ing NW under the total axial compressive load Fa + 2τ d and NW with a = 10 nm and L/a = 10; g Critical buckling loads of the
the distributed transverse force per length 2τ d(d2 v/dx 2 ) is CC NW with a = 5.71 nm
[64]
Eq. 12 can be simplified as
d4 v d2 v d2 v
(E I )∗ 4 + (Fa + 2τ d) 2 = 2τ d 2 , (12)
dx dx dx d4 v d2 v
(E I )∗ 4
+ Fa 2 = 0. (14)
dx dx
where the effective flexural rigidity (E I )∗ of the rectangular
As can be seen from Eq. 14, the surface tension does not
NW is [16–18]
influence the critical load of the buckling CF NW if the short-
ening of the NW due to the surface tension is neglected. The
(E I )∗ = E m dh 3 /12 + E s dh 2 /2 + E s h 3 /6. (13) reason is that the contribution from the out-of-plane force

123
Comput Mech (2018) 61:687–697 695

component due to the surface tension is completely compen- Table 1 The critical buckling loads of the CF and CC NWs
sated by the contribution from the in-plane axial force due Without surface effects With surface effects
to the same surface tension. We tentatively infer that Eq. 14
COMSOL Theory Diff. COMSOL Theory Diff.
is suitable for buckling CF NWs with any cross-sectional
(nN) (nN) % (nN) (nN) %
geometry whenever the effective flexural rigidity is formu-
lated accordingly, such as the buckling CF NW with a square CF 10.1 10.1 0.0 9.4 9.3 1.1
cross section in Fig. 6c. In order to make the NW buckle, at CC 143.9 146.9 −2.0 134.2 136.2 −1.5
least one end of the NW can move axially regardless of the
The parameters are a = 10 nm, L/a = 10, τ 0 = 1.41 N/m, E s =
boundary conditions, and therefore the counteracting force −4.96 N/m, E m = 49.5 Gpa, and νm = 0.455
balancing the in-plane surface stress induced force always
exists, which implies that Eq. 14 is valid for all the boundary
conditions, such as the CC NW in Fig. 6d. The critical load of a = 5.71 nm. Good agreement is found between the theory
the axially buckling NW FcrE B , based on the Euler-Bernoulli and the atomistic simulations. For comparison, the incor-
beam theory, is the solution of Eq. 14 rect critical buckling loads obtained from Ref. [18] are also
shown in Fig. 6g, which occurs if the Y–L equation is directly
FcrE B = π 2 (E I )∗ /(K L)2 , (15) applied for buckling without accounting for the axial force
resulting from the surface tension.
where K = 2.0 for the CF NW and K = 0.5 for the CC
NW [64]. The transverse shear deformation is neglected in
Eq. 15, which is nontrivial for stubby NWs. Based on previ- 4 Conclusion
ous discussions and the Timoshenko beam theory [64], the
critical load of the axially buckling NW FcrT is similar to the We have presented a methodology for modeling surface stress
classical one except that the flexural rigidity is substituted by effects on both stiff and soft solids. In contrast to most pre-
the effective flexural rigidity vious approaches for modeling surface stress, the proposed
methodology leverages the vast resources available in widely
π 2 (E I )∗ /(K L)2 used commercial FE software packages like ANSYS and
FcrT = , (16) COMSOL. The present methodology is also different in that
1 + π 2 (E I )∗ /[βμm A(K L)2 ]
the Young–Laplace equation is not the starting point for
where the shear coefficient is β = 5(1 + νm )/(6 + 5νm ) for capturing surface stress effects. Instead, we model the out-
rectangular cross sections [65], μm is shear modulus, and A of-plane force component induced by the surface tension in
is cross-sectional area. curved surfaces through geometric nonlinear analysis. The
The calculations in studying the surface effects on the essence of the methodology is to convert the surface ten-
NWs in Buckling are based on gold in [001] growth direc- sion and the surface elasticities in the zero-thickness surface
tion with (001) crystal face. For the (001) crystal face of to an initial stress and corresponding elastic properties in
gold, the surface tension is τ 0 = 1.41 N/m and the surface the equivalent finite-thickness shell. The results obtained by
Young’s modulus is E s = (S1111 S2222 − S1122 2 )/S using the methodology are consistent with previous experi-
2222 =
−4.96 N/m [52]. The Young’s modulus and Poisson’s ratio mental and analytical work where surface tension impacts
of macroscopic single-crystal gold in [001] direction are both stiff and soft solids. The methodology offers exten-
E m = 49.5 Gpa and νm = 0.455 respectively [66]. The sive opportunities for performing complex multidisciplinary
CF NW (Fig. 6c) and the CC NW (Fig. 6d) have a square modeling work such as electromechanical coupling, fluid-
cross section with a side dimension of a. The curves of the structural coupling, and thermomechanical coupling in the
axial force Fa versus the transverse displacement v of the solid systems by exploiting the built-in functionality of com-
CF and CC NWs calculated from COMSOL are in Fig. 6e mercial FE codes under the influence of the surface tension
and f respectively, where a is 10 nm and the aspect ratio L/a or surface stress.
is 10. In the COMSOL model, small forces in x direction
Acknowledgements This work is supported by the National Natu-
contributing to approximately v = a/100 are applied at the ral Science Foundation of China (Grant No. 11504170). Both authors
tip and the midspan of the CF and CC NWs respectively for acknowledge the support of the Department of Mechanical Engineering
introducing the perturbation. Increasing Fa until Fa changes at Boston University.
slowly with respect to v leads to the critical buckling loads,
which agree well with the theoretical values calculated from References
Eq. 16 in Table 1. The theoretical values are further compared
to the results from atomistic simulations for [001]/(001) gold 1. Cammarata RC, Sieradzki K (1994) Surface and interface stresses.
[67] in Fig. 6g, where the boundary conditions are CC and Annu Rev Mater Sci 24:215

123
696 Comput Mech (2018) 61:687–697

2. Park HS, Cai W, Espinosa HD, Huang H (2009) Mechanics of 26. Yang Q, To AC (2017) Multiresolution molecular mechanics: sur-
crystalline nanowires. MRS Bull 34(3):178 face effects in nanoscale materials. J Comput Phys 336:212
3. Cuenot S, Fretigny C, Demoustier-Champagne S, Nysten B (2004) 27. Young T (1805) An essay on the cohesion of fluids. Philos Trans
Surface tension effect on the mechanical properties of nano- R Soc Lond 95:65
materials measured by atomic force microscopy. Phys Rev B 28. Laplace PS (1805) Traité de Mécanique Céleste, vol 4. Gauthier-
69(16):165410 Villars, Paris
4. Jing GY, Duan HL, Sun XM, Zhang ZS, Xu J, Li YD, Wang JX, Yu 29. Shuttleworth R (1950) The surface tension of solids. Proc Phys Soc
DP (2006) Surface effects on elastic properties of silver nanowires: Lond Sect A 63(5):444
contact atomic-force microscopy. Phys Rev B 73(23):235409 30. Gurtin ME, Murdoch AI (1975) A continuum theory of elastic
5. Chen CQ, Shi Y, Zhang YS, Zhu J, Yan YJ (2006) Size depen- material surfaces. Arch Ration Mech Anal 57(4):291
dence of Young’s modulus in ZnO nanowires. Phys Rev Lett 31. Chen TY, Chiu MS, Weng CN (2006) Size dependence of young’s
96(7):075505 modulus in ZnO nanowires. J Appl Phys 100(7):5
6. Diao J, Gall K, Dunn ML (2003) Surface-stress-induced phase 32. Andreotti B, Baumchen O, Boulogne F, Daniels KE, Dufresne ER,
transformation in metal nanowires. Nat Mater 2(10):656 Perrin H, Salez T, Snoeijer JH, Style RW (2016) Solid capillarity:
7. Park HS, Gall K, Zimmerman JA (2005) Shape memory and pseu- when and how does surface tension deform soft solids? Soft Matter
doelasticity in metal nanowires. Phys Rev Lett 95:255504 12:2993
8. Liang W, Zhou M, Ke F (2005) Shape memory effect in Cu 33. Style RW, Jagota A, Hui C-Y, Dufresne ER (2016) Elastocapillar-
nanowires. Nano Lett 5(10):2039 ity: surface tension and the mechanics of soft solids. Annu Rev
9. Park HS, Gall K, Zimmerman JA (2006) Deformation of FCC Condens Matter Phys 8:99–118
nanowires by twinning and slip. J Mech Phys Solids 54(9):1862 34. Roman B, Bico J (2010) Elasto-capillarity: deforming an elastic
10. Weinberger CR, Cai W (2012) Plasticity of metal nanowires. J structure with a liquid droplet. J Phys: Condens Matter 22:493101
Mater Chem 22(8):3277 35. Liu JL, Feng XQ (2012) On elastocapillarity: a review. Acta Mech
11. Gao W, Yu SW, Huang GY (2006) Finite element characteriza- Sin 28(4):928
tion of the size-dependent mechanical behaviour in nanosystems. 36. Style RW, Dufresne ER (2012) Static wetting on deformable sub-
Nanotechnology 17(4):1118 strates, from liquids to soft solids. Soft Matter 8(27):7177
12. He J, Lilley CM (2009) The finite element absolute nodal coordi- 37. Style RW, Boltyanskiy R, Che Y, Wettlaufer J, Wilen LA, Dufresne
nate formulation incorporated with surface stress effect to model ER (2013) Universal deformation of soft substrates near a contact
elastic bending nanowires in large deformation. Comput Mech line and the direct measurement of solid surface stresses. Phys Rev
44(3):395 Lett 110(6):066103
13. Yvonnet J, Mitrushchenkov A, Chambaud G, He QC (2011) Finite 38. Style RW, Hyland C, Boltyanskiy R, Wettlaufer JS, Dufresne ER
element model of ionic nanowires with size-dependent mechanical (2013) Surface tension and contact with soft elastic solids. Nat
properties determined by ab initio calculations. Comput Methods Commun 4:2728
Appl Mech Eng 200(5–8):614 39. Jensen KE, Sarfati R, Style RW, Boltyanskiy R, Chakrabarti A,
14. Farsad M, Vernerey FJ, Park HS (2010) An extended finite ele- Chaudhury MK, Dufresne ER (2015) Wetting and phase separation
ment/level set method to study surface effects on the mechanical in soft adhesion. Proc Nat Acad Sci 112(47):14490
behavior and properties of nanomaterials. Int J Numer Methods 40. Style RW, Boltyanskiy R, Allen B, Jensen KE, Foote HP, Wettlaufer
Eng 84(12):1466 JS, Dufresne ER (2015) Stiffening solids with liquid inclusions. Nat
15. Yvonnet J, Quang HL, He QC (2008) An XFEM/level set approach Phys 11(1):82
to modelling surface/interface effects and to computing the size- 41. Mora S, Phou T, Fromental JM, Pismen LM, Pomeau Y (2010)
dependent effective properties of nanocomposites. Comput Mech Capillarity driven instability of a soft solid. Phys Rev Lett
42:119 105(21):214301
16. Wang GF, Feng XQ (2007) Effects of surface elasticity and residual 42. Saksono PH, Peric D (2006) On finite element modelling of surface
surface tension on the natural frequency of microbeams. Appl Phys tension. Variational formulation and applications—part I: qua-
Lett 90(23):231904 sistatic problems. Comput Mech 38:265
17. He J, Lilley CM (2008) Surface effect on the elastic behavior of 43. Henann DL, Bertoldi K (2014) Modeling of elasto-capillary phe-
static bending nanowires. Nano Lett 8(7):1798 nomena. Soft Matter 10:709
18. Wang GF, Feng XQ (2009) Surface effects on buckling of 44. Seifi S, Park HS (2016) Computational modeling of electro-elasto-
nanowires under uniaxial compression. Appl Phys Lett 94:141913 capillary phenomena in dielectric elastomers. Int J Solids Struct
19. Sharma P, Ganti S, Bhate N (2003) Effect of surfaces on the size- 87:236
dependent elastic state of nano-inhomogeneities. Appl Phys Lett 45. Seifi S, Park HS (2017) Electro-elastocapillary rayleigh-plateau
82(4):535 instability in dielectric elastomer films. Soft Matter 13:4305
20. He J, Lilley CM (2008) Surface stress effect on bending resonance 46. Wang Y, Henann DL (2016) Finite-element modeling of soft solids
of nanowires with different boundary conditions. Appl Phys Lett with liquid inclusions. Extreme Mech Lett 9:147
93(26):263108 47. ANSYS (2016) Mechanical APDL 17.0
21. Park HS, Klein PA, Wagner GJ (2006) A surface Cauchy–Born 48. He J (2015) Surface stress on the effective Young’s modulus and
model for nanoscale materials. Int J Numer Methods Eng 68:1072 Poisson’s ratio of isotropic nanowires under tensile load. AIP Adv
22. Park HS, Klein PA (2007) Surface Cauchy–Born analysis of surface 5(11):117206
stress effects on metallic nanowires. Phys Rev B 75:085408 49. COMSOL (2016) COMSOL multiphysics user’s guide, version
23. Park HS, Klein PA (2008) A surface Cauchy–Born model for sili- 5.2a
con nanostructures. Comput Methods Appl Mech Eng 197:3249 50. Gurtin ME, Murdoch AI (1978) Surface stress in solids. Int J Solids
24. Javili A, Steinmann P (2009) A finite element framework for con- Struct 14(6):431
tinua with boundary energies. Part I: the two-dimensional case. 51. Miller RE, Shenoy VB (2000) Size-dependent elastic properties of
Comput Methods Appl Mech Eng 198:2198 nanosized structural elements. Nanotechnology 11(3):139
25. Javili A, Steinmann P (2010) A finite element framework for con- 52. Shenoy VB (2005) Atomistic calculations of elastic properties of
tinua with boundary energies. Part II: the three-dimensional case. metallic fcc crystal surfaces. Phys Rev B 71(9):094104
Comput Methods Appl Mech Eng 199:755

123
Comput Mech (2018) 61:687–697 697

53. Shenoy VB (2006) Erratum: Atomistic calculations of elastic prop- 62. Song F, Huang GL, Park HS, Liu XN (2011) A continuum model
erties of metallic fcc crystal surfaces [Phys. Rev. B 71, 094104 for the mechanical behavior of nanowires including surface and
(2005)]. Phys Rev B 74(14):149901 surface-induced initial stresses. Int J Solids Struct 48(14–15):2154
54. Barriere B, Sekimoto K, Leibler L (1996) Peristaltic instability of 63. Yun G, Park HS (2009) Surface stress effects on the bending prop-
cylindrical gels. J Chem Phys 22:1735 erties of fcc metal nanowires. Phys Rev B 79:195421
55. Cialetta P, Amar MB (2012) Peristaltic patterns for swelling and 64. Timoshenko SP, M GJ (1961) Theory of elastic stability, 2nd edn.
shrinking of soft cylindrical gels. Soft Matter 8:1760 McGraw-Hill, NY
56. Mora S, Phou T, Fromental JM, Pismen LM, Pomeau Y (2010) Cap- 65. Hutchinson JR (2001) Sheer coefficients for Timoshenko beam
illarity driven instability of a soft solid. Phys Rev Lett 105:214301 theory. J Appl Mech-Trans ASME 68(1):87
57. Taffetani M, Ciarletta P (2015) Beading instability in soft cylin- 66. Gan Y, Sun Z, Chen Z (2015) Extensional vibration and size-
drical gels with capillary energy: weakly non-linear analysis and dependent mechanical properties of single-crystal gold nanorods.
numerical simulations. J Mech Phys Solids 81:91 J Appl Phys 118(16):164304
58. Xuan C, Biggins J (2016) Finite-wavelength surface-tension-driven 67. Olsson PAT, Park HS (2012) On the importance of surface elastic
instabilities in soft solids, including instability in a cylindrical chan- contributions to the flexural rigidity of nanowires. J Mech Phys
nel through an elastic solid. Phys Rev E 94(2):023107 Solids 60(12):2064
59. Muller P, Saul A (2004) Elastic effects on surface physics. Surf Sci
Rep 54(5–8):157
60. Duan HL, Wang J, Huang ZP, Karihaloo BL (2005) Eshelby for-
malism for nano-inhomogeneities. Proc R Soc Math Phys Eng Sci
461(2062):3335
61. Hui CY, Jagota A (2016) Effect of surface tension on the relaxation
of a viscoelastic half-space perturbed by a point load. J Polym Sci
Part B-Polym Phys 54(2):274

123