International Journal of Solids and Structures xxx (2014) xxxxxx

Contents lists available at ScienceDirect

International Journal of Solids and Structures

journal homepage: www.elsevier.com/locate/ijsolstr

A nite element implementation of a coupled diffusion-deformation

theory for elastomeric gels
Shawn A. Chester a,, Claudio V. Di Leo b, Lallit Anand b
a
Department of Mechanical and Industrial Engineering, New Jersey Institute of Technology, Newark, NJ 07102, USA
b
Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, USA

a r t i c l e i n f o a b s t r a c t

Article history: The theory of Chester and Anand (2011) for uid diffusion and large deformations of elastomeric gels is
Received 11 March 2014 implemented as a user-dened element (UEL) subroutine in the commercial nite element software
Received in revised form 15 July 2014 package ABAQUS. A specialized form of the constitutive equations and the governing partial differential
Available online xxxx
equations of the theory are summarized, and the numerical implementation is described in detail. To
demonstrate the robustness of the numerical implementation a few illustrative numerical simulation
Keywords: examples for axisymmetric, plane strain, and three-dimensional geometries are shown. For educational
Elastomeric materials
purposes, and also to facilitate the numerical implementation of other coupled multiphysics theories,
Gels
Diffusion
the source code for the UEL is provided as an online supplement to this paper.
Large deformations 2014 Elsevier Ltd. All rights reserved.
Finite elements

1. Introduction
An elastomeric gel is a polymer network swollen by a uid. Gels
can be designed to swell by several hundred percent in volume,
and the amount of swelling can be controlled by varying various
stimulihumidity, temperature, and pH. Gels are ubiquitousthey
are found in foods and medicines, and they nd use in several
important and diverse applications including carriers for drug
delivery (Peppas et al., 2006), actuators and sensors in microuidic
devices (Beebe et al., 2000), tissue engineering matrices (Chan and
Mooney, 2008), as well as packers for sealing in oil wells
(Kleverlaan et al., 2005; Bhavsar et al., 2008).
Modeling elastomeric gels is interesting and challengingit
involves concurrent deformation of the polymer network and
diffusion of the solvent through the network. An early, but limited,
theory for swelling of gels is due to Tanaka and co-workers
(cf., e.g., Tanaka and Fillmore, 1979). In recent years there has been
a convergence towards a more complete coupled diffusion-
deformation theory for describing the response of gelsincluding
swelling and drying, squeezing of uid by applied mechanical
deformation, and forced permeation (cf., e.g., Doi, 2009). Within
the limits of a nonlinear eld theory in which the uidsolid mixture
is treated as a single homogenized continuum body which allows for
a mass ux of the uid, essentially similar theories (for electrically-
neutral gels) have been formulated by Hong et al. (2008), Duda et al.
Corresponding author. Tel.: +1 973 596 3658.
E-mail address: shawn.a.chester@njit.edu (S.A. Chester).
(2010) and Chester and Anand (2010, 2011). References to the vast
previous literature on gels may be found in these publications.
The past few years have also seen several publications related to
the numerical implementation of these theories for solving coupled
diffusion-deformation boundary value problems for gels. In their
early work, Suo and co-workers (cf., e.g., Hong et al., 2009;
Marcombe et al., 2010; Liu et al., 2010) using the UHYPER capabil-
ities of the commercial nite element package ABAQUS/Standard
(2013) implemented a chemical equilibrium version their theory
in which the chemical potential was presumed to be homogeneous
in the gel, and the transient diffusion and associated transient
swelling kinetics were neglected. More recently, Toh et al. (2013)
and Duan et al. (2013) have simulated the transient diffusion and
swelling kinetics of polymeric gels by drawing on an analogy
between diffusion of solvent molecules and conduction of heat in
solids, and using the built-in thermo-mechanically coupled nite-
elements and associated solution procedures in ABAQUS. While
useful for gels in which the diffusion equation has a form similar
to that for heat transfer, this methodology is not applicable to more
general multi-physics problems. In addition to UHYPER capabilities,
Zhang et al. (2009) developed a nite element method for gels
based on the theory presented in Hong et al. (2008). Lucantonio
et al. (2013)using the nite element software package COMSOL/
Multiphysicshave also recently performed simulations for
transient swelling-induced large deformations in polymeric gels.
In contrast to the work of Suo et al., Toh et al. (2013), and
Lucantonio et al. (2013), Chester and Anand (2011) implemented
their own theory for elastomeric gels by writing two- and

http://dx.doi.org/10.1016/j.ijsolstr.2014.08.015
0020-7683/ 2014 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Chester, S.A., et al. A nite element implementation of a coupled diffusion-deformation theory for elastomeric gels. Int. J.
Solids Struct. (2014), http://dx.doi.org/10.1016/j.ijsolstr.2014.08.015
2 S.A. Chester et al. / International Journal of Solids and Structures xxx (2014) xxxxxx

three-dimensional user-dened nite element subroutines (UELs) body B with the region of space it occupies in a xed reference con-
and implemented them in ABAQUS. In their approach, a new nite guration, and denote by X an arbitrary material point of B. A
element is constructed whose degrees of freedom are taken to be motion of B is then a smooth one-to-one mapping x vX; t with
the primal variables in the set of partial differential equations deformation gradient, velocity, and velocity gradient given by1
(pdes) of interest. This set of pdes need not resemble those for cou- _ 1 :
pled thermo-mechanical problems, and may be of any kind which
F rv; v v_ ; L grad v FF 2:1
are amenable to nite-element solutions methods which employ The deformed body is denoted as B.
standard C0 -continuous nite-element basis functions. The theory is based upon a multiplicative decomposition
Because detailed numerical procedures and source codes are
seldom published in scientic journal papers, the numerical imple- F Fe Fs ; with Fs ks 1; ks > 0; 2:2
mentation of a new coupled theory using the nite element of the deformation gradient F into elastic and swelling parts Fe and
method is often a challenging taskespecially for beginners in a Fs , respectively, with the swelling taken to be isotropic, where ks is
new research area. However there is a new emerging trendin the swelling stretch. With X denoting the volume of a mole of uid
order to more widely disseminate new computational methods molecules, we assume the swelling stretch is given by
and proceduresresearchers are beginning to publish papers
which address the details of their computational implementations. ks 1 XcR 1=3 ; 2:3
For example, recently Giner et al. (2009) published an ABAQUS where cR represents the uid concentration measured in moles of
implementation of the extended nite element method for linear uid per unit reference volume of the dry elastomer.
elastic fracture analysis as a UEL subroutine. Also Park and The constitutive equations of the theory are:
Paulino (2012) published an ABAQUS implementation of a cohesive
nite element as a UEL subroutine. In both cases, for educational  Free energy: A simple form of the free energy function which
purposes, the numerical implementation is discussed in detail, accounts for the combined effects of mixing, swelling, and
and the source code for the UEL was provided. elastic stretching is,
In Chester and Anand (2011) we showed the results from sev-     
eral numerical simulations of our theory which was implemented XcR 1
wR l0 cR R#cR ln v
as a user-dened element (UEL) subroutine in ABAQUS. However, 1 XcR 1 XcR
 
in that paper we did not provide any details regarding our numer- 1  2  s 1 2
G 3k  1  2 ln J J Kln J e : 2:4
ical implementation procedures and methods. Accordingly, in the 2 2
spirit of the recent papers by Giner et al. (2009) and Park and
Paulino (2012), the main purpose of this paper is to discuss the Here, l0 is a reference chemical potential for the uid, R the gas
details of the numerical implementation of our coupled diffu- constant, # is the constant temperature under consideration, v is
sion-deformation theory for non-ionic gels, and also to make avail- a dimensionless measure of the enthalpy of mixing known as
able the source code for the UEL in ABAQUS. the FloryHuggins interaction parameter, G is the shear modulus
The paper is organized as follows. In Section 2 the constitutive of the network, K is a bulk modulus of the gel, and
equations and the governing partial differential equations of the r r
k def 1 1
theory are summarized. In order to focus attention on the numer- trC 1 XcR 2=3 trCe ; 2:5
ical implementation as opposed to the details of the specic consti- 3 3
tutive functions, we use a simple specialized form of our is an effective stretch.
constitutive theory and restrict our discussion to isothermal condi-  Constitutive equation for the Cauchy stress: Corresponding to the
tions. In Section 3 we describe the numerical solution procedure in free energy (2.4), the Cauchy stress T is given by
substantial detail. Additional details regarding the user elements,  
@w h i
verication of the basic element technology, and thermo-chemo- T J 1 2Fe Re Fe> J1 G /2=3 Be  G1 J s Kln J e 1 ;
mechanical coupling is provided in three Appendices. @C
2:6
 The associated online Supplemental materials to this paper
where we have introduced the polymer volume fraction dened
include a detailed tutorial on generating an input le, and
by
instructions on running ABAQUS with our UEL. The source code
is also provided. def 1 3
/ ks : 2:7
1 XcR
In Section 4, in order to demonstrate the robustness of the numer-
Next, since
ical implementation, we show the results from illustrative example
2
problems for axisymmetric, plane strain, and three-dimensional B FF> ks Be /2=3 Be ; 2:8
geometries. We nish in Section 5 with some concluding remarks.
(2.6) reduces to

2. Summary of our theory for gels T J 1 G B  1 J s Kln J e 1: 2:9

In this section we briey summarize the basic continuum

mechanical theory for elastomeric gels under isothermal condi-
tions. Cf. Chester and Anand (2011) for complete details of the
formulation of the theory.
2.1. Kinematics. Constitutive theory
Consider a uid-free (dry) macroscopically homogeneous elas-
tomeric body. We identify such a macroscopically-homogeneous
1
Notation: We use standard notation of modern continuum mechanics (Gurtin
et al., 2010). Specically: r and Div denote the gradient and divergence with respect
to the material point X in the reference conguration; grad and div denote these
operators with respect to the point x vX; t in the deformed body; a superposed
1
dot denotes the material time-derivative. Throughout, we write Fe1 Fe ,
e> e >
F F , etc. We write trA, symA; skwA; A0 , and sym0 A respectively, for the
trace, symmetric, skew, deviatoric, and symmetric-deviatoric parts of a tensor A. Also,
the inner product of tensors A and B is denoted by A : B, and the magnitude of A by
p
jAj A : A.

Please cite this article in press as: Chester, S.A., et al. A nite element implementation of a coupled diffusion-deformation theory for elastomeric gels. Int. J.
Solids Struct. (2014), http://dx.doi.org/10.1016/j.ijsolstr.2014.08.015
 S.A. Chester et al. / International Journal of Solids and Structures xxx (2014) xxxxxx 3

 Constitutive equation for the chemical potential: The chemical with u  ; j, prescribed functions of x and t. The initial data is taken
 ; t; l
potential l is given by as
@wR 1   uX; 0 u0 X; and lX; 0 l0 X in B: 2:17
l  X J e trT l0 R# ln1  / / v/2
@cR 3
The coupled set of Eqs. (2.12) and (2.14), together with (2.15),
1 2
 XKln J e K Xln J e : 2:10 (2.16), and (2.17) yield an initial boundary value problem for the
2 displacement ux; t and the chemical potential lx; t.
 Constitutive equation for the uid ux: We assume that the In applications, for the case in which the environment consists
spatial uid ux, j, depends linearly on the spatial gradient of a pure and incompressible liquid, the boundary condition on
of the chemical potential, grad l, with the mobility tensor chemical potential l  is given by
taken to be isotropic so that
l l0 Xpa ; 2:18
j m grad l; 2:11
where l0 is a reference chemical potential for the liquid, X is the
where m is a scalar mobility coefcient, which in general is an volume of a mole of liquid molecules, and pa is the hydrostatic pres-
isotropic function of the stretch and the uid concentration. sure of the liquid. Also, if a portion of the boundary is impermeable
to the liquid, then on that portion the prescribed normal ux j
vanishes.
Remark. In our previous papers (Chester and Anand, 2010;
Chester and Anand, 2011; Chester, 2012) we had assumed a consti- 3. Numerical solution procedure
tutive equation for the uid ux of the form
jR mrl; In the absence of body forces, the strong forms of the coupled
partial differential equations of the theory are
where jR is a referential uid ux and rl is the referential gradient 8 9
of the chemical potential. However, others in the literature (c.f., e.g., < div T 0
> in B; >
>
>
>
Hong et al., 2008; Duda et al., 2010) have argued that (2.11) is the Balance of momentum u u  on S u ; >
>
>
>
: >
>
more appropriate constitutive equation for the uid ux for isotro- Tn t on S t ; >
>
>
=
pic gels. We adopt (2.11) in this paper. 8
>
> /_
> in B; >
< J X/2  div j 0
> >
>
>
Balance of fluid concentration >
>
2.2. Governing partial differential equations >
>
> l l on S l ; >
>
>
>
> >
>
: ;
The governing partial differential equations, when expressed in j  n j on S j :
the deformed body, consist of 3:1

1. The local force balance for the macroscopic Cauchy stress, Then, with w1 ; w2 denoting two weighting (or test) elds which
vanish on S u and S l , respectively, the corresponding weak forms
divT b 0; 2:12 are:
  9
with b a non-inertial body force, and T given by (2.9). R @w1 R   >
T : d v w  t da; >
>
2. The local balance for the uid concentration, B
@x St 1 >
=
!   3:2
c_ R Jdiv j; 2:13 R _
/ R @w2 R
 >
B
w2 dv  B  j dv  Sj w2j da: >
>
>
;
which using (2.7) may be written in the form J X/2 @x
S
/_ The body is approximated using nite elements, B Be , and
 div j 0; 2:14
J X/2 the trial solutions for the displacement, and chemical potential
are interpolated inside each element by
in which the uid ux j is given by (2.11), and the chemical P )
potential l is given by (2.10). u uA N A ;
P 3:3
l lA NA ;
2.3. Boundary and initial conditions
with the index A f1; 2; . . . ; Mg denoting the nodes of the element,
We also need boundary and initial conditions to complete the uA , and lA denoting nodal displacements, and chemical potentials,
theory. Let S u and S t be complementary subsurfaces of the bound- and NA the shape functions. We employ a standard Galerkin
ary @B of the body B in the sense @B S u [ S t and S u \ S t . approach, in that the weighting elds are interpolated by the same
Similarly let S l and S j be complementary subsurfaces of the shape functions, viz.
boundary: @B S l [ S j and S l \ S j . Then for a time interval P )
w1 wA1 NA ;
t 2 0; T we consider a pair of boundary conditions in which the P 3:4
displacement u is specied on S u and the surface traction on S t : w2 wA2 NA :

uu on S u  0; T; Using (3.3) and (3.4) in (3.2) yields the following element-level sys-
2:15 tem of equations:
Tn t on S t  0; T;
! 9
and a pair of boundary conditions in which the chemical potential is R @NA R
 >
>
>
Be
T dv Se NAt da; >
>
specied on S l and the uid ux on S j @x t =
) ! ! Z
 3:5
l l on Sl  0; T; R /_ R @NA >
>
2:16 Be
N A
dv  Be
 j dv  NAj da: >
>
>
;
j  n j on S j  0; T; J X/2 @x e
Sj

Please cite this article in press as: Chester, S.A., et al. A nite element implementation of a coupled diffusion-deformation theory for elastomeric gels. Int. J.
Solids Struct. (2014), http://dx.doi.org/10.1016/j.ijsolstr.2014.08.015
4 S.A. Chester et al. / International Journal of Solids and Structures xxx (2014) xxxxxx

This system of coupled equations is solved using a Newton Consider next the tangent (3.8)4,
procedure by dening the following element-level residuals for Z ! Z !
the displacement and chemical potential, AB NA @ /_ @ @NA @ l
K ll  dv m dv
Be J X @ l /2 Be @ l
B B
! 9 @xi @xi
R @NA R
A  >
> Z
> @
A
A
Ru  Be T dv Se N t da;
 >
>
@x t =  N j da;
S ej @ l
B
! ! Z
R /_ R @NA
 >
> ! !
Rl A Be N A 2
dv Be  j dv NAj da; >
>
>
;
Z
NA /_ @/ 1 @ /_
Z
@NA @NB
J X/ @x e
Sj  2 3 dv m dv
Be J X / @ lB /2 @ lB Be @xi @xi
3:6 ! !
Z Z 
@m @NA @ l A @j
which using index notation, and (2.11) for the uid ux, may be dv  N da;
written in the form Be @ lB @xi @xi S ej @ lB
! 9
A R @N A R
A  >
>
> and hence
Rui  Be T ij dv Se N t i da; >
>
@xj t = Z ! Z !
! ! AB NA NB
/_ @/ @ /_ @NA @NB
R /_ R @NA @ l R
 >
> K ll 2  dv m dv
Rl A Be N A dv  Be m dv Se NAj da: >
> Be J X/
2 / @l @l Be @xi @xi
2 >
;
JX/ @xi @xi j
Z ! Z !
@m B @NA @ l @j
3:7 N dv  NA NB da: 3:16
Be @l @xi @xi S ej @l
In addition to the residuals, the following tangents are also required
for the iterative Newton solver: Similarly, the remaining two tangents in (3.8) are given by
9 Z  
@Ru A > @NA @T ij @/ B
AB
Kuu  ;>>
> K AB
ui l N dv ; 3:17
@uB > > Be @xj @/ @ l
>
>
A >>
AB @R u >
> where
Kul  ; >
@ lB =
3:8 @T K
@Rl A > > 1;
Klu AB  ;>> @/ J e /
@uB > >
>
>
A >> and Klu AB is approximated by
@R l >
>
AB
K ll  :>; Z   B
@ lB @NA @l @N
K AB
luk  m dil dv : 3:18
Be @xi @xk @xl
First, the tangent (3.8)1, in index notation, is given by
Z Z In the solution procedure one needs to compute the polymer
@N A   @NB @ti
K AB
ui uk Aijkl dv  NA NB da: 3:9 volume fraction, /, at every increment in order to evaluate the con-
Be @xj @xl S et @uk stitutive response functions and eventually the residuals (3.6).
With TR JTF> denoting the Piola stress, the spatial tangent modu- Using J e J/ we may rewrite (2.10) in the following dimensionless
lus A is dened in terms of the referential tangent modulus AR by form

def l0  l XK 1 XK
Aijkl J 1 F jm F ln AR imkn ; 3:10 ln1  / / v/2  lnJ/ ln J/2 0;
R# R# 2 R#
where 3:19
@TR which serves as an implicit equation for /. Thus given the pair F; l
AR : 3:11
@F at any instant of time, Eq. (3.19) is solved for the corresponding
This standard result requires lengthy computations, which we give value of /. The term /_ is computed using the approximation,
in an appendix (cf. the Remark in Appendix A). For the constitutive /  /n
theory under consideration here, using the constitutive equation
/_ n1 : 3:20
Dt
(2.9) for T we obtain that _ l appearing in the tangents (3.16) are com-
Also, terms such as @ /=@
TR GF  F> J s Kln J e F> ; 3:12 puted numerically using a nite difference scheme.
In (3.6), (3.9) and (3.16) the integrals are evaluated numerically
and hence
    using Gaussian-quadrature. Since this is a standard method in the
@TR @F> @ ln J @F> nite element literature, we do not present details here; the details
AR G I J s K F>  ln J e ;
@F @F @F @F may be found in our source code and in the literature.
3:13 In its notation, ABAQUS/Standard (2013) requires certain
matrices denoted as RHS and AMATRX to be evaluated and/or
where in writing the last term in (3.13) we have used
updated by the user element subroutine UEL:
ln J e lnJ/ ln J ln /. Then using the identities
   
@F> @ ln J  The matrix RHS, as dened by the ABAQUS documentation, is
Iijkl dik djl ; F 1 1
li F jk ; F 1
lk ; 3:14
@F ijkl @F kl An array containing the contributions of this element to the
right-hand-side vectors of the overall system of equations.
the component form of the referential tangent modulus is
  Referring to (3.7), RHS is the overall elemental residual which
@TR


in matrix form is given by
AR ijkl G dik djl F 1 1
li F jk J s K F 1 1 e 1 1
ji F lk  lnJ F li F jk :
@F ijkl h i>
3:15 R R1u1 R1u2 R1l R2u1 R2u2 R2l . . . RM M M
u1 Ru2 Rl ; 3:21

Please cite this article in press as: Chester, S.A., et al. A nite element implementation of a coupled diffusion-deformation theory for elastomeric gels. Int. J.
Solids Struct. (2014), http://dx.doi.org/10.1016/j.ijsolstr.2014.08.015
 S.A. Chester et al. / International Journal of Solids and Structures xxx (2014) xxxxxx 5

in two dimensions, and (iii) a 3D 8-node linear isoparametric brick which we refer to as
h i> U3D8.
R R1u1 R1u2 R1u3 R1l R2u1 R2u2 R2u3 R2l . . . RM M M M
u1 Ru2 Ru3 Rl ;
3:22 Additional details regarding the user elements is provided
in Appendix A, and the basic element technology is veried in
in three dimensions, with M the total number of nodes per Appendix B.
element.
 The matrix AMATRX as dened by the ABAQUS documentation is  The associated online Supplemental materials to this paper
An array containing the contribution of this element to the include a detailed tutorial on generating an input le, and
Jacobian (stiffness) or other matrix of the overall system of instructions on running ABAQUS with our UEL. The source code
equations. Referring to (3.8), AMATRX is the overall tangent is also provided.
which is given by
2 11 3
K u u K 11 K 11 12 12 12
K 1M 1M 1M 4. Example problems
u u u l K u1 u1 K u1 u2 K u1 l u1 u1 K u1 u2 K u1 l
6 111 1 111 2 111 1M 7
6 K u u K u u K u l K 12 12 12
u2 u1 K u2 u2 K u2 l K 1M 1M
u2 u1 K u2 u2 K u2 l 7
7
6 2 1 2 2 2 To demonstrate the robustness of the developed user element
6 11 11 11 12 12 12 1M 1M 1M 7
6 K lu K lu K ll K lu K lu K ll K lu1 K lu2 K ll 7 subroutine (UEL), in this section we show a few illustrative numer-
6 1 2 1 2 7
6 21 7
6 K u1 u1 K 21
u u K 21
u l K 22
u u K 22
u u K 22
u l   K 2M
u u K 2M
u u K 2M
u l 7 ical simulation examples for plane strain, axisymmetric, and three-
6 1 2 1 1 1 1 2 1 1 1 1 2 1 7
6 K 21 K 21 K 21 K 22 K 22 K 22 2M 2M 2M 7
K u2 u1 K u2 u2 K u2 l 7 dimensional geometries. Table 1 lists plausible representative
6 u2 u1 u2 u2 u2 l u2 u1 u2 u2 u2 l
6
K 6 21 21 21 22 22 22 2M 7
7 values for the material properties of a polymeric gel at room
6 K lu1 K lu2 K ll K lu1 K lu2 K ll K 22 2M
lu1 K lu2 K ll 7
6 7 temperature, which we have used in our calculations.
6 .. .. .. 7
6 . . . 7 Specically, the ground state shear modulus for the polymer, G,
6 7
6 K M1 K M1 K M1 K M2 K M2 K M2 MM MM MM 7
K u1 u1 K u1 u2 K u1 l 7 is chosen to have a value 1 MPa, and the bulk modulus K is taken to
6 u1 u1 u1 u2 u1 l u1 u1 u1 u2 u1 l
6 M1 7 be two orders of magnitude larger. The volume of a solvent mole-
6K M1 M1 M2 M2 M2 MM MM MM 7
4 u2 u1 K u2 u2 K u2 l K u2 u1 K u2 u2 K u2 l   K u2 u1 K u2 u2 K u2 l 5 cule is taken as X 1:0  104 m3 /mol, and the reference chemical
K M1 M1 M1 M2
lu1 K lu2 K ll K lu1 K lu2 K ll
M2 M2
K MM MM
lu1 K lu2 K ll
MM
potential of the uid is taken as l0 0:0 J/mol. Additionally, we
3:23 have chosen a value of v 0:1 for the FloryHuggins interaction
parametera value which is favorable for a high degree of
in two dimensions. For brevity we do not list the complete
swelling.
matrix in three dimensions.
Recall that the mobility m in (2.11) is an isotropic function of
the stretch and the uid content. We assume here that the mobility
In addition to the required RHS and AMATRX as described above,
at a given temperature # is given by2
ABAQUS/Standard (2013) allows the storage and updating state
variables using SVARS, which as dened by the ABAQUS documen- Dc
m ; 4:1
tation is An array containing the values of the solution-dependent R#
variables associated with this element. Due to the term /_ appear-
where D > 0, a constant, represents a diffusion coefcient, and
ing in (3.7), and computed using (3.20), we are required to save /
c cR =J is the uid concentration measured in moles of uid per
at each integration point. For that purpose we use the
unit deformed volume. As a representative value in our numerical
SVARS(Variables) array (where the length Variables is set
simulations we take D 5  109 m2/s.
in the input le) to save / in the form

SVARS /1 /2 /3 . . . /Q ;
where Q is the total number of integration points in the element.
We have also used the PNEWDT time-step control feature of
ABAQUS. PNEWDT as dened by the ABAQUS/Standard (2013) doc-
umentation is Ratio of suggested new time increment to the time
increment currently being used. If PNEWDT is set less than 1.0,
ABAQUS will abandon the current increment and start over with
a new time increment DTIME of DTIMEPNEWDT for the next time
increment. If PNEWDT is set greater than 1.0, and the increment
converges, ABAQUS may increase the next time increment DTIME
by DTIMEPNEWDT. In this work we make extensive use of PNEWDT,
it is used for both general time incrementation as well as constitu-
tive time incrementation. An example of constitutive time incre-
mentation is limiting the local change in the polymer volume
fraction / in a given time step. Also, if a displacement increment,
or chemical potential increment is too large PNEWDT is used to
decrease the time increment. Further details may be found in the
source code.
We have implemented our theory in ABAQUS/Standard (2013)
by writing a user element subroutine (UEL) by writing three
different elements:
(i) a 2D plane-strain 4-node linear isoparametric quadrilateral
which we refer to as UPE4;
(ii) a 2D axisymmetric 4-node linear isoparametric quadrilateral
which we refer to as UAX4; and
Please cite this article in press as: Chester, S.A., et al. A nite element implementation of a coupled diffusion-deformation theory for elastomeric gels. Int. J.
Solids Struct. (2014), http://dx.doi.org/10.1016/j.ijsolstr.2014.08.015
3:24 4.1. Free-swelling of an axisymmetric cylinder followed by simple
compression between rigid platens
We rst consider free-swelling of a cylindrical sample of an
initially dry gel, followed by simple compression of the swollen
gel between rigid platens.3 We simulate compression between rigid
platens (as opposed to simple tension) in order illustrate the use of
our UEL with the contact capabilities of ABAQUS.
The initially-dry cylindrical specimen is 5 mm in diameter and
5 mm tall. In our numerical simulation we assume an axisymmet-
ric initial cylindrical geometry shown in Fig. 1. Using symmetry,
we model only half of an axisymmetric slice and approximate it
by using 175 UAX4 user-elements; the mesh is purposely chosen
to be unstructured. The compression platen (not shown) is mod-
eled as a rigid surface. Finally, to study frictional boundary effects
between the swollen gel and the rigid surface we consider two
cases: (i) the interface between the gel and the platen is frictional
with a Coulomb friction coefcient of lfric 0:05, and (ii) a fric-
tionless case with lfric 0:0.
The simulation is broken into two steps: in Step 1 the initially
dry gel is allowed to freely swell to a near equilibrium cylindrical
2
At present not much is known experimentally about the precise dependence of m
on either the stretch or the uid content, we use this simple form to describe our
numerical solution procedure.
3
Cai et al. (2010) have recently performed such experiments on an alginate
hydrogel.
6 S.A. Chester et al. / International Journal of Solids and Structures xxx (2014) xxxxxx
Table 1
Material parameters for a representative elastomeric gel at room
temperature.
Parameter Value
G 1 MPa
K (=100G) 100 MPa
X 1:0  104 m3/mol
v 0.1
l0 0.0 J/mol
D 5  109 m2/s
D l l0 with the solvent.
C  Fig. 2(a) and (b) show snapshots of the polymer volume fraction
A (d) show snapshots of the polymer volume fraction / plotted
B shots of the polymer volume fraction / plotted on the deformed
Fig. 1. Initial geometry used in the free swelling followed by simple compression
example. Here only half the geometry is shown for clarity, and the nite element
mesh superimposed. Due to symmetry, only half of an axisymmetric slice is
modeled.
shape, and in Step 2 the rigid platen is moved into the swollen gel
to compress it. The initial condition for the chemical potential of
the dry polymer is taken to be
lX; t 0 l0 14388:57 J=mole: 4:2
This initial condition is computed using Eq. (2.10), with l0
0:0 J/mole, / 0:999; # 298 K, Je 1:0, and v 0:1.4 The
boundary conditions are as follows:
 Step 1:
For the mechanical boundary conditions we prescribe ur 0
along AD, uz 0 along AB, while faces BC and CD are taken to
be traction-free. In this step, the rigid surface is sufciently
removed from the specimen and held xed.
For the chemical boundary conditions we prescribe zero ux
along AD and AB due to symmetry, while faces BC and CD are
prescribed a time dependent chemical potential
l t l0 l0 expt=td ;
4
We have used /X; t 0=0.999 rather than 1.0 to eliminate numerical difculties
with the ln1  / term in (2.10).
Please cite this article in press as: Chester, S.A., et al. A nite element implementation of a coupled diffusion-deformation theory for elastomeric gels. Int. J.
Solids Struct. (2014), http://dx.doi.org/10.1016/j.ijsolstr.2014.08.015
where l0 is the chemical potential of the surrounding solvent
and t d 300 s, so that at times t t d the prescribed chemical
potential l t approaches the value l0 0:0 J/mole of the
surrounding solvent.
The total time allowed for this swelling step is 6 h. At which
point the swelling gel has come into contact with the rigid
platen.
 Step 2:
For the mechanical boundary conditions, symmetry condi-
tions ur 0 along AD and uz 0 along AB are maintained
from the previous step, however now the rigid surface is pre-
scribed an additional downward displacement of 2.5 mm
over 300 s, and then held xed in the displaced position
for another 6 h.
For the chemical boundary conditions we maintain the zero
ux conditions along AD and AB, however now we add a no
ux boundary condition to face CD which is in contact with
the rigid surface. Face BC maintains the boundary condition
In Fig. 2,
/ plotted on the deformed geometry in the free-swelling step
after 3 h and 6 h, respectively, of swelling.
 For the case with a Coulomb friction lfric 0:05, Fig. 2(c) and
on the deformed geometry after 220 s and 6 h, respectively, in
the second compression step.
 For the frictionless case lfric 0:0, Fig. 2(e) and (f) show snap-
geometry after 220 s and 6 h, respectively, in the second com-
pression step.
The redistribution of the uid due to the compression is clearly
observed in Fig. 2an indication of the coupling between the
deformation and diffusion. Also, in the frictional case in the long
time limit Fig. 2 (d), the region of low / (more swollen) near the
edges of the gel is due to a region of tensile stress relative to the
surroundingsthe tensile stress in this region develops due to
the friction between the gel and the platen. In contrast, in the
frictionless case, Fig. 2 (f), the distribution of the polymer
volume fraction / along the direction of compression is much more
uniformas expected.
The forcedisplacement and forcetime curves computed during
Step 2 are shown in Fig. 3 for the frictionless case.5 Since the under-
lying material model for the polymer is fully elastic, the apparent
stress relaxation seen in Fig. 3(b) is due entirely to the out-ux of
uid.
This rst illustrative numerical example shows the use of our
axisymmetric user-element undergoing large deformations,
transient uid transport, and frictional contact with a rigid surface.
4.2. Constrained swelling of a bilayer in plane strain
This example problem is modeled after the recent experiments
of Yoon et al. (2010) which are similar to the experiments of
Holmes and co-workers (Holmes et al., 2011; Pandey and
Holmes, 2013). In the experiments of Yoon et al. (2010) a thin layer
of a swellable gel is bonded to a non-swellable elastomer, the gel is
allowed to swell and this causes large bending of the bilayer.
The initial dry geometry is taken to be 100 mm long and 5 mm
tall, with 2.5 mm for the swellable gel and 2.5 mm for the
5
The corresponding curves for the frictional case are not much different.
 S.A. Chester et al. / International Journal of Solids and Structures xxx (2014) xxxxxx 7

0.260 0.210
0.254 0.208
0.248 0.205
0.241 0.203
0.235 0.201
0.229 0.198
0.222 0.196
0.216 0.194
0.210 0.191
0.203 0.189
0.197 0.186
0.191 0.184
0.184 0.182

(a) (b)

0.217 0.219
0.214 0.217
0.211 0.216
0.209 0.215
0.206 0.214
0.203 0.213
0.200 0.211
0.198 0.210
0.195 0.209
0.192 0.208
0.190 0.207
0.187 0.205
0.184 0.204

(c) (d)

0.209 0.217
0.207 0.216
0.205 0.216
0.203 0.215
0.201 0.215
0.199 0.214
0.197 0.213
0.195 0.213
0.192 0.212
0.190 0.212
0.188 0.211
0.186 0.210
0.184 0.210

(e) (f)
Fig. 2. Contours of the polymer volume fraction / on the deformed body at (a) after 3 h of free swelling, (b) after 6 h of free swelling, (c) after 220 s of compression, and (d) 6 h
after the compression platen has been held xed in the deformed position with a Coulomb friction coefcient of 0.05. For comparison, considering a frictionless interface
between the platen and gel, after (e) 220 s of compression, and (f) 6 h after the compression platen has been held xed in the deformed position. For clarity the rigid surface
has been removed, and the axisymmetric mesh swept and mirrored about symmetry lines/planes. The thick black lines indicate the initial dry body.

non-swellable elastomeric substrate. Due to the symmetry of the
problem, we only model one-half of the geometry in the simula-
tion, Fig. 4; we assume that plane strain conditions prevail. The
swellable gel is approximated by 201 UPE4 plane strain user-
elements, while the non-swellable elastomer is approximated by
201 CPE4H 4-node plane strain hybrid elements (built-in ABAQUS
elements). The non-swellable elastomer is modeled as an incom-
pressible Neo-Hookean material with a shear modulus of 50 MPa.
The interface between the gel and non-swellable elastomer is
taken to be perfectly bonded.
As in the previous example, the initial condition for the
chemical potential of the dry swellable gel is taken to
be lX; t 0 l0 14388:57 J=mol. Referring to Fig. 4, for the
mechanical boundary conditions we impose symmetry along the
symmetry plane of the beam (face AD), and pin the node at point
A to avoid rigid body motion. The free end of the beam (face BC)
is constrained to remain planar, and all other faces are traction
free. For the chemical boundary conditions we prescribe no ux
along the plane of symmetry AD. The top face of the beam CD is
prescribed a uniform time dependent chemical potential of
l t l0 l0 expt=td , with td 300 s.
Fig. 5(a)(d) shows snapshots of the polymer volume fraction /
on the deformed beam after 15 min, 30 min, 1 h, and 6 h,
respectively.6
This numerical example shows the use of our plane strain user
elements in conjunction with built-in ABAQUS plane-strain
elements.
6
Note that since we model only half the geometry in the nite element simulation,
the self contact shown in Fig. 5(d) is modeled by the use of a rigid surface which is not
shown in this gure.

Please cite this article in press as: Chester, S.A., et al. A nite element implementation of a coupled diffusion-deformation theory for elastomeric gels. Int. J.
Solids Struct. (2014), http://dx.doi.org/10.1016/j.ijsolstr.2014.08.015
8 S.A. Chester et al. / International Journal of Solids and Structures xxx (2014) xxxxxx

50 50

40 40

Force (N)

Force (N)
30 30

20 20

10 10

0 0
0 0.5 1 1.5 2 2.5 0 2 4 6
Displacement (mm) Time (h)
(a) (b)
Fig. 3. (a) Force displacement response, and (b) force time response for the simple compression simulation. The small offset from the origin in the force displacement
curve is due to a small preload during the swelling step. Also, note the force relaxation due to the movement of the uid within the body.

outer diameter D0 , wall thickness t0 , and height H0 in the initial

Swellable Gel dry conguration. We consider two initial geometries:
D C
fD0 ;t 0 ;H0 g f4:636;0:206;0:6g mm and f4:636; 0:309;1:2g mm

A B for our numerical simulations. These dimensions were chosen to

Non-swellable Rubber
Fig. 4. Initial geometry used in the constrained swelling example. Here only half
the geometry is simulated due to symmetry. The swellable gel is the darker material
above and the non-swellable rubber is the lighter material below.
4.3. Swelling induced three-dimensional buckling of constrained
cylindrical tubes
This nal example problem is modeled after the recent experi-
ments conducted by Lee et al. (2012) on hydrogel tubes. Their
experiments consisted of cylindrical tubes of hydrogels which
were mechanically constrained on one end, while the other end
was placed in contact with a uid and allowed to swell and possi-
bly buckle. In their experiments Lee et al. (2012) varied the geom-
etry of their tubes and found that varying the wall thickness, radii,
and heights of their tubes resulted in the formation of different
buckled patterns on the swollen ends of their hydrogel tubes.
Due to the nature of the experimentally-observed three-dimen-
sional buckled patterns, we model the body in three dimensions.
Fig. 6 shows the basic geometry under consideration, a tube with
provide two distinct swollen shapes similar to those reported in
the recent paper of Lee et al. (2012). Guided by the (approximate)
symmetry observed in their experiments, we model only 1/4 of
the geometry. The initial meshes are shown in Fig. 6, they consist
of 2720, and 6840 U3D8 8-node brick elements with 4 and 6 ele-
ments through the thickness for the short and tall geometries,
respectively. In order to aid the initiation of buckling, all nodes on
the front face of the cylindrical body were given a small random geo-
metric imperfection on the order 102 mm in the height dimension.
As in the previous examples, the initial condition for the
chemical potential of the dry swellable gel is taken to be
lX; t 0 l0 14388:57 J=mole. For the mechanical bound-
ary conditions, referring to Fig. 6, we assume that the back face
is held xed, all symmetry planes are prescribed appropriate sym-
metry conditions, and the remaining faces are traction free. For the
chemical boundary conditions, we assume that only the front face
is in contact with the uid, and prescribe l t l0 l0 expt=t d
with t d 300 s while all other faces are ux free.
Fig. 7 shows contours of the polymer volume fraction on the
deformed body after 215 s, 250 s, and 900 s, for (a) the short-thin
tube and (b) the tall-thick tube. These shapes resemble those

0.929 0.771
0.900 0.735
0.871 0.698
0.842 0.661
0.813 0.624
0.783 0.587
0.754 0.550
0.725 0.514

(a) (b)

0.604 0.376
0.577 0.367
0.550 0.358
0.522 0.349
0.495 0.339
0.468 0.330
0.441 0.321
0.414 0.312

(c) (d)

Fig. 5. Contours of the polymer volume fraction / on the deformed body at (a) 15 min, (b) 30 min, (c) 1 h, and (d) 6 h. The initial dry geometry is shown by the thick black
lines and the non-swellable rubber is shown on the deformed mesh in white.

Please cite this article in press as: Chester, S.A., et al. A nite element implementation of a coupled diffusion-deformation theory for elastomeric gels. Int. J.
Solids Struct. (2014), http://dx.doi.org/10.1016/j.ijsolstr.2014.08.015
 S.A. Chester et al. / International Journal of Solids and Structures xxx (2014) xxxxxx 9

Fig. 6. Initial geometries used in the swell induced buckling simulations. Only one-quarter of the geometry is meshed and symmetry boundary conditions are assumed.

0.737 0.888
0.720 0.859
0.702 0.829
0.684 0.799
0.667 0.769
0.649 0.739
0.631 0.709
0.614 0.679
0.596 0.649
0.578 0.619
0.560 0.589
0.543 0.559
0.525 0.529

0.687 0.850
0.667 0.816
0.646 0.783
0.625 0.749
0.604 0.715
0.583 0.682
0.562 0.648
0.541 0.614
0.520 0.581
0.499 0.547
0.478 0.513
0.457 0.480
0.436 0.446

0.423 0.557
0.402 0.529
0.382 0.500
0.361 0.472
0.340 0.443
0.319 0.415
0.299 0.387
0.278 0.358
0.257 0.330
0.237 0.301
0.216 0.273
0.195 0.245
0.175 0.216

(a) (b)
Fig. 7. Contours of the polymer volume fraction / on the deformed body for fD0 ; t 0 ; H0 g given by (a) f4:636; 0:206; 0:6g mm and (b) f4:636; 0:361; 1:825g mm, at 215 s (top),
250 s (middle), and 900 s (bottom) of swelling. Note how the initial geometry leads to very different swollen shapes, one buckled and the other stable.

Please cite this article in press as: Chester, S.A., et al. A nite element implementation of a coupled diffusion-deformation theory for elastomeric gels. Int. J.
Solids Struct. (2014), http://dx.doi.org/10.1016/j.ijsolstr.2014.08.015
10 S.A. Chester et al. / International Journal of Solids and Structures xxx (2014) xxxxxx

Z
observed experimental by Lee et al. (2012), and we note that the
Kuu G> AG dv ; 6:2
initial geometry in Fig. 7(a) results in a buckled pattern, while Be
the initial geometry in Fig. 7(b) results in an axisymmetrically
with G the non-symmetric discrete gradient matrix, and A the
swollen tube.
matrix form of the spatial tangent modulus A. Details also to follow
These simulation results show that our three-dimensional
for the specic element types considered.
user-element is able to capture the differing deformed shapes in
the two geometries.
Remark. The lengthy, but standard, computation for the displace-
ment residual (3.9) is presented here for completeness. Restating
5. Concluding remarks
the displacement residual
Z !
The coupled theory of Chester and Anand (2011) for uid @NA
RAui  T ij dv :
permeation and large deformations of elastomeric gels is imple- Be @xj
mented as a user-dened element (UEL) subroutine in ABAQUS/
Standard (2013). The numerical solution procedure employing a Now, using the identities dv Jdv R , the denition of the Kirchhoff
user element (UEL) in ABAQUS is discussed in signicant detail, stress s JT TR F> , together with the identity F> ru grad u
and the online Supplemental materials to this paper include a for a scalar eld u, we may recast the residual in the referential form
detailed tutorial on generating an input le, and instructions on Z
@NA 1
running ABAQUS with our UEL. The source code is also provided. RAui  F sij dv R : 6:3
B e @X a aj
Further, in an Appendix, Section C, we overview a thermo-
chemo-mechanically coupled theory for thermally responsive gels. Now, using (3.8)1, and the identities
That Appendix serves as an example on how to extend the material X @N B @F 1
ji @NA @NA
shown in the main body of this paper to another theory and imple- F mn dmn uBm ; F 1 1
li F jk ; and F 1
ai ;
mentation. The source code for the thermo-chemo-mechanically @X n @F kl @xi @X a
coupled theory is also provided. we have
It is hoped that the details of the numerical implementation of !
Z
the particular coupled diffusion-deformation theory for gels pro- @RAui @N A @F aj
1
1 @ sij @F mn
K AB
ui uk  B sij F aj dv R
vided here, will also facilitate the numerical implementation of @uk Be @X a @F mn @F mn @uBk
other coupled multi-physics theories within the ABAQUS frame- Z 1
!
work via user elements. @NA @F aj 1 @ sij @NB
s ij F aj dmk dv R
Indeed, the nite element framework described here has e @X a @F mn @F mn @X n
B
already proven useful in the numerical implementation of a variety Z   B
@NA 1 @ sij @N
of other theories in ABAQUS: (a) gradient plasticity (Anand et al., F 1 F
nj am
1
sij F aj dmk dv R
2012); (b) oxide growth in thermal barrier coatings (Loeffel et al., Be @X a @F mn @X n
Z   B
2013; Al-Athela et al., 2013); (c) dielectric elastomers (Henann @NA 1 @ sij @N
F 1 1
ak F nj sij F aj dv R
et al., 2013); (d) surface tension in soft materials (Henann and Be @X a @F kn @X n
Bertoldi, 2014); (e) hydrogen transport in metals (Di Leo and Z  
@NA 1 @ sij @NB
Anand, 2013); and Lithium diffusion in electrode materials (Di F ja F 1 1
ak F nj sij F aj F ln dv R
e @xj @F kn @xl
Leo et al., 2014). B
Z  
@NA @ sij @NB
Acknowledgements djk sil F ln dv R
Be @xj @F kn @xl
Z  
The nancial support provided by a grant from the National Sci- @NA @ sij @NB
J 1 djk sil J1 F ln dv
ence Foundation (CMMI-1063626) is gratefully acknowledged. Be @xj @F kn @xl
Also, SAC gratefully acknowledges Suvrat P. Lele as a mentor who Z   B
@NA @T R;im 1 @N
provided initial guidance in programming UEL subroutines for J 1 djk sil J1 F ln F jm J F ln T R;im djk dmn dv
e @xj @F kn @xl
B
ABAQUS, and Kaspar Loeffel for nding many of the initial aws Z   B
in his code. @NA @T R;im 1 @N
J 1 djk sil J1 F ln F jm J djk sil dv
Be @xj @F kn @xl
Z  
Appendix A. Isoparametric elements @NA 1 @T R;im @N B
J F ln F jm dv
Be @xj @F kn @xl
A.1. Basic element technology
or with
Typically, in nite element formulations the volume (meaning def def @TR
Aijkl J 1 F jm F ln AR imkn ; with AR 6:4
non-surface) contribution to the displacement residual (3.7)1 is @F
evaluated in matrix form dening a spatial tangent modulus, we arrive at
Z
Z
Ru  B Tdv ;
>
6:1 @NA   @NB
Be K AB
ui uk Aijkl dv : 6:5
Be @xj @xl
and often referred to as the internal force vector. Here we have used
h
matrix notation, where Ru is the element displacement residual
vector, B the standard B-matrix (often referred to as the symmet- To accommodate both compressible and nearly incompressible
ric discrete gradient matrix), and T the Cauchy stress vector. In material behavior and mitigate volumetric locking behavior, we
the following sections specic forms for Ru ; B , and T are provided have implemented the so called F-bar method (de Souza Neto
for plane strain, axisymmetric, and three dimensional elements. et al., 1996). This method is based on replacing the deformation
Further, the corresponding displacement tangent (3.9) is often gradient suitably such that the incompressibility constraint is
evaluated in the form enforced as an approximate average throughout the element,

Please cite this article in press as: Chester, S.A., et al. A nite element implementation of a coupled diffusion-deformation theory for elastomeric gels. Int. J.
Solids Struct. (2014), http://dx.doi.org/10.1016/j.ijsolstr.2014.08.015
 S.A. Chester et al. / International Journal of Solids and Structures xxx (2014) xxxxxx 11

rather than point wise at each integration point. The method is

based on the distortional-volumetric split of the deformation
gradient
F Fdis Fvol ; 6:6
with

Fdis J 1=3 F; Fvol J 1=3 1: 6:7

To construct the modied deformation gradient at an integra-
tion point of interest, we rst determine the deformation gradient
at the centroid of the element, denoted by Fc . Then the modied
deformation gradient is constructed as
 1=3 Fig. 8. Two-dimensional linear quadrilateral, and three-dimensional linear brick
 det Fc
F F: 6:8 isoparametric master elements, with the node numbering as shown.
det F
As mentioned before, to accommodate nearly incompressible
Now when computing the stresses at the integration points, the
 is substituted in place of F. This material behavior where appropriate we use the F-bar method of
modied deformation gradient F
de Souza Neto et al. (1996). Specically in plane-strain the method
has the effect that all the integration points in the element share
reduces to
the same total volumetric deformation gradient in the element,
2 3 2 3
specically det Fc . This formulation does not change the integra- F11F12 0  1=2 F 11 F 12 0
tion point residual computation (6.1), simply that F is used to com- 6 7 F F
c;11 c;22  F F
c;12 c;21 6 7
4 F 21F22 0 5 4 F 21 F 22 0 5; 6:13
pute the constitutive response, rather than F. However, the tangent F 11 F 22  F 12 F 21
0 0 1 0 0 1
computation (6.2) must be modied to
Z Z with F c;ij the deformation gradient at the centroid of the element.
Kuu G> AG dv G> QG0  G dv ; 6:9 Correspondingly, in plane-strain, the tangent modication (6.9) is
Be Be
|{z} |{z} now given by
standard terms additional terms
1 1
with Q A : 1  1  T  1 6:14
2 2
1 2 where
Q A : 1  1  T  1: 6:10
3 3 2 3
A11 A14 0 0 0 A11 A14
We have developed a 4-node quadrilateral two-dimensional 6A A
6 21 24 0 0 0 A21 A24 7
7
plane-strain element, a 4-node quadrilateral axisymmetric ele- A : 1  1 6 7 6:15
ment, and an 8-node brick three-dimensional element. The basic 4 A31 A34 0 0 0 A31 A34 5
element technology of each is overviewed in the following A41 A44 0 0 0 A41 A44
subsections.
with
A.2. 2D elements Amn Aijkl
denoting the matrix representation of A using the following trans-
For the two-dimensional elements developed, the node order-
formation table
ing in the natural coordinates is shown in Fig. 8. Referring to
Fig. 8, the shape functions for the 4-node linear element with
m=n i=k j=l
respect to the natural coordinates are given by
1 1 1
1
N1 1  n1  g 2 2 1
4
3 1 2
1
N2 1 n1  g 4 2 2
4
3 1
N 1 n1 g
4
1 and with
4
N 1  n1 g: 2 3
4 T 11 0 0 0 T 11
6T 0 0 0 T 21 7
6 21 7
T  1 6 7 6:16
A.2.1. Plane-strain 4 T 12 0 0 0 T 12 5
For a plane-strain element we have the condition that F 33 1, T 22 0 0 0 T 22
and F 13 F 31 F 23 F 32 0. Also, the B-matrix is given by
2 3 for the plane-strain formulation. Finally, the standard non-symmet-
@N 1 @N2 @N M
0 0  0 ric discrete gradient matrix in (6.9) is given by
6 @x1 @x1 @x1
7
6 @N 1 @N2 @N M 7 2 3
B6 0 0  0 @x2 7
6:11 @N 1
0 @N 2
0  @N M
0
4 @x2 @x2 5
1 1 2 2 M M 6 @x1 @x1 @x1
7
@N @N @N @N
 @N @N 6 0 @N 1
0 @N 2
 0 7
@NM
@x2 @x1 @x2 @x1 @x2 @x1 6 @x1 @x1 7
@x1
G6
6 @N1
7
7
where M is the total number of nodes in the element, and the stress @N 2 @N M
6 @x2 0 0  0 7
vector T is given by 4 @x2 @x2 5
@N 1 @N 2 @NM
0 @x2
0 @x2
 0 @x2
>
T T 11 T 22 T 12  : 6:12
in plane strain with M the total number of nodes.

Please cite this article in press as: Chester, S.A., et al. A nite element implementation of a coupled diffusion-deformation theory for elastomeric gels. Int. J.
Solids Struct. (2014), http://dx.doi.org/10.1016/j.ijsolstr.2014.08.015
12 S.A. Chester et al. / International Journal of Solids and Structures xxx (2014) xxxxxx

A.2.2. Axisymmetric for the axisymmetric formulation. Finally, the standard non-
For an axisymmetric element we have the condition that symmetric discrete gradient matrix in (6.9) is given by
F 13 F 31 F 23 F 32 0, and F 33 R=R0 . Furthermore, the inte- 2 3
R R @N1 @N 2 @N M
gration is now modied such that Be dxdy ! Be 2prdrdz. In the
0 0  0
6 @x1 @x1 @x1
7
numerical implementation F 33 is computed as R=R0 , and numeri- 6 0 @N 1
0 @N 2
 0 @NM 7
6 @x1 @x1 @x1 7
cally this is accomplished by 6 7
6 1 @N 2 @N M 7
X G 6 @N 0 0  0 7
6 @x2 @x2 @x2 7
R NA xA1 ; 6:17 6 @N 1 @N 2
7
@NM 7
6 0 0  0
4 @x2 @x2 @x2 5
X A N1 N2 NM
R0 N X A1 ; 6:18 R
0 R
0  R
0

where xA1 are the current 1-coordinates of the nodes, and X A1 are the with M the total number of nodes.
reference 1-coordinates of the nodes. Note that this scheme auto-
matically implies that our axisymmetric element formulation A.3. Three dimensional element
assumes the radial direction is the 1-direction. Also, the B-matrix
is given by For the three-dimensional 8-node linear brick element devel-
2 3 oped the node ordering in the natural coordinates is shown in
@N 1 @N 2 @N M
0 0  0 Fig. 8. Referring to Fig. 8, the shape functions for the 8-node linear
6 @x1 @x1 @x1
7
6 @N 1
@N2
@N M7 brick element with respect to the natural coordinates are given by
6 0 0  0 7
B6 7
@x2 @x2 @x2
6 @N1 M 7
6:19 1 1
@N 1 @N 2 @N2 @N M
6 @x
4 2 @x1 @x2 @x1
 @x2
@N 7
@x1 5 N1 1  n1  g1  f; N2 1 n1  g1  f;
8 8
N1 N2 NM 1 1
0 0  0
R R R N3 1 n1 g1  f; N4 1  n1 g1  f;
8 8
where M is the total number of nodes in the element, and the stress 1 1
5 6
vector T is given by N 1  n1  g1 f; N 1 n1  g1 f;
8 8
T T 11 T 22 T 12 T 33 > : 6:20 7 1 8 1
N 1 n1 g1 f; N 1  n1 g1 f:
8 8
As mentioned before, to accommodate nearly incompressible
material behavior the F-bar method of de Souza Neto et al. Here the B-matrix is given by
(1996) is implemented, where the modied deformation gradient 2 @N1 @N 2 @N M
3
@x
0 0 @x1
0 0  @x1
0 0
is given by (6.8), viz., 6 1 7
6 0 @N 1
0 0 @N 2
0  0 @N M
0 7
 1=3 6 @x2 @x2 @x2 7
 det Fc 6 7
F F; 6:21 6 0 0 @N 1
0 0 @N2
 0 0 @NM 7
det F 6 @x3 @x3 @x3 7
B6
6 @N1 1 2 2 M
7
7 6:24
@N @N @N @N @N M
where both F 33 and F c33 are computed before applying the F-bar 6 @x2 0 0  0 7
6 @x1 @x2 @x1 @x2 @x1 7
method. The tangents need to be corrected according to (6.9) with 6 @N 1 @N 1 @N 2 @N2 @N M
7
@NM 7
6 0 0  0
2 3 4 @x3 @x2 @x3 @x2 @x3 @x2 5
A11 A14 A15 0 0 A11 A14 A15 A11 A14 A15 @N1
0 @N 1 @N 2
0 @N2
 @N M
0 @NM
6 7 @x3 @x1 @x3 @x1 @x3 @x1
6 A21 A24 A25 0 0 A21 A24 A25 A21 A24 A25 7
6 7 where M is the total number of nodes in the element, and the stress
A : 1  1 6
6 A31 A34 A35 0 0 A31 A34 A35 A31 A34 A35 7
7
6 7 vector T is given by
4 A41 A44 A45 0 0 A41 A44 A45 A41 A44 A45 5
A51 A54 A55 0 0 A51 A54 A55 A51 A54 A55 T T 11 T 22 T 33 T 12 T 23 T 13 > : 6:25
6:22 To accommodate nearly incompressible material behavior the
F-bar method of de Souza Neto et al. (1996) as described above
where
is implemented through
Amn Aijkl  1=3
 det Fc
F F; 6:26
using the following transformation table det F

m=n i=k j=l The tangents need to be corrected according to (6.9) with
A : 1  1
1 1 1 2 3
2 2 1 A11 A15 A19 0 0 0 A11 A15 A19 0 0 0 A11 A15 A19
6A A A 0 0 0 A21 A25 A29 0 0 0 A21 A25 A29 7
3 1 2 6 21 25 29 7
6 7
4 2 2 6 A31 A35 A39 0 0 0 A31 A35 A39 0 0 0 A31 A35 A39 7
6 7
5 3 3 6 7
6 A41 A45 A49 0 0 0 A41 A45 A49 0 0 0 A41 A45 A49 7
6 7
6
6 A51 A55 A59 0 0 0 A51 A55 A59 0 0 0 A51 A55 A59 7
7
6 7
and with 6 A61 A65 A69 0 0 0 A61 A65 A69 0 0 0 A61 A65 A69 7
6 7
6 A71 A75 A79 0 0 0 A71 A75 A79 0 0 0 A71 A75 A79 7
2 3 6 7
T 11 0 0 T 11 T 11 6 7
4 A81 A85 A89 0 0 0 A81 A85 A89 0 0 0 A81 A85 A89 5
6 7
6 T 21 0 0 T 21 T 21 7 A91 A95 A99 0 0 0 A91 A95 A99 0 0 0 A91 A95 A99
6 7
6 7
T  1 6 T 12 0 0 T 12 T 12 7 6:23 6:27
6 7
6T 0 0 T 22 T 22 7 where
4 22 5
T 33 0 0 T 33 T 33 Amn Aijkl

Please cite this article in press as: Chester, S.A., et al. A nite element implementation of a coupled diffusion-deformation theory for elastomeric gels. Int. J.
Solids Struct. (2014), http://dx.doi.org/10.1016/j.ijsolstr.2014.08.015
 S.A. Chester et al. / International Journal of Solids and Structures xxx (2014) xxxxxx 13

using the following transformation table (J 1) and in this case the corresponding Cauchy stress for a
neo-Hookean material is given by
m=n i=k j=l T GB  P1;
1 1 1
with P a constitutively indeterminate pressure, which is introduced
2 2 1
to satisfy the incompressibility constraint.
3 3 1
For simple compression in the e1 -direction, the analytical
4 1 2
solution for the stress-stretch behavior is given by
5 2 2
6 3 2 T 11 Gk2  k1 :
7 1 3
8 2 3 This analytical solution is compared against corresponding results
9 3 3 computed by using a single U3D8 element, and a single UAX4 ele-
ment in Fig. 9(a), in which the solid line represents the analytical
solution and the symbols the numerically calculated results.
and with In plane strain compression in the e1 -direction with no defor-
2 3 mation in the e3 -direction, the analytical solution for the stress-
T 11 0 0 0 T 11 0 0 0 T 11 stretch behavior is given by
6T 0 0 0 T 21 0 0 0 T 21 7
6 21 7
6 7 T 11 Gk2  k2 :
6 T 31 0 0 0 T 31 0 0 0 T 31 7
6 7
6 7 This analytical solution also compared in Fig. 9(a) against the corre-
6 T 12 0 0 0 T 12 0 0 0 T 12 7
6 7 sponding result computed by using a single UPE4 element; the
T  1 6
6 T 22 0 0 0 T 22 0 0 0 T 22 7
7 6:28
6 7 dashed line represents the analytical solution, and the symbols
6 T 32 0 0 0 T 32 0 0 0 T 32 7
6 7 the numerically calculated result.
6 T 13 0 0 0 T 13 0 0 0 T 13 7
6 7 Next, we examine the response in simple shear. Fig. 9(b) com-
6 7
4 T 23 0 0 0 T 23 0 0 0 T 23 5 pares the analytical results for the shear stress and normal stress
T 33 0 0 0 T 33 0 0 0 T 33 difference given by

for the three-dimensional formulation. Finally, the standard non- T 12 Gc and T 11  T 33 Gc2 ;
symmetric discrete gradient matrix in (6.9) is given by
2 3 where c is the amount of shear, against the numerical results com-
@N 1 @N2 @NM puted using a single U3D8 element. Again, the solid and dashed
@x1
0 0 @x1
0 0  @x1
0 0
6 7 lines represent the analytical solutions, and the symbols the numer-
6 0 @N 1
0 0 @N2
0  0 @N M
0 7
6 @x1 @x1 @x1 7 ically calculated result.
6 7
6 @N1 @N 2 @N M 7 The results shown in Fig. 9 verify our basic element technology
6 0 0 0 0  0 0 7
6 @x1 @x1 @x1 7
when dealing with large mechanical deformations in the absence
6 @N1 @N2 @NM
7
6 @x 0 0 0 0  0 0 7
6 2 @x2 @x2 7 of diffusion of the solvent.
6 @N 1 @N2 @N M
7
G6
6 0 @x2
0 0 @x2
0  0 @x2
0 7 7
6 7 B.2. The diffusion only problem
6 0 0 @N1
0 0 @N 2
 0 0 @N M 7
6 @x2 @x2 @x2 7
6 1 7
6 @N 0 0 @N2
0 0  @NM
0 0 7 To verify the diffusion component of our nite elements, we use
6 @x3 @x3 @x3 7
6 7 the classical analogy between the diffusion of a chemical species in
6 @N1
@N2
@NM 7
6 0 0 0 0  0 0 7 a body and the diffusion of heat in a body, and consider the energy
4 @x3 @x3 @x3 5
1 2
@N @N @N M balance in a deformable heat conductor as our governing partial
0 0 @x3
0 0 @x3
 0 0 @x3
differential equation,
with M the total number of nodes.
#_ a divgrad #;
Appendix B. Verication of the basic element technology where a is the thermal diffusivity of the material. We compare
results form a UEL implementation for this problem against the
Since typical multi-physics problems are difcult to solve corresponding widely-accepted results obtained from the built-in
analytically, in this section we briey compare a few numerical thermo-mechanical elements in ABAQUS/Standard (2013). We
solutions with those that we can compute either analytically or consider the initial geometry as shown in Fig. 10, with the initial
by using an accepted solution method for a model system. Due condition #0 300 K everywhere in the body. We take the thermal
to the complexity of the coupling, we consider it sufcient for diffusivity a to be a constant equal to 1  106 m2/s.
this paper to verify the pure deformation problem and the pure In order to examine the UEL capabilities as thoroughly as
diffusion problem, separately. possible, we consider the following two-step simulation;

B.1. The deformation only problem
In the absence of a uid, the constitutive theory reduces to that
of a compressible neo-Hookean material with the Cauchy stress
T J 1 GB  1 K lnJ1:
To approximate a nearly incompressible neo-Hookean material we
take K 103 G. We use this constitutive equation to verify our UELs
for the deformation only problem. For the corresponding analytical
solutions considered below, we assume complete incompressibility
 In the rst step, the body is deformed isothermally using
displacement boundary conditions.
 In the second step, the displacement boundary conditions are
held xed, and then the thermal boundary conditions are chan-
ged such that the transient heat equation discussed above is
solved on the deformed body.
Referring to Fig. 10, in the rst isothermal deformation step, the
nodes along face AD are xed, while nodes along face BC are pre-
scribed a displacement to deform the body, while the temperature

Please cite this article in press as: Chester, S.A., et al. A nite element implementation of a coupled diffusion-deformation theory for elastomeric gels. Int. J.
Solids Struct. (2014), http://dx.doi.org/10.1016/j.ijsolstr.2014.08.015
14 S.A. Chester et al. / International Journal of Solids and Structures xxx (2014) xxxxxx

0 1
Shear Stress
Normal Stress Difference
5 0.8 U3D8

Normalized Stress

Normalized Stress
Uniaxial
10 Plane strain 0.6
U3D8
15 UAX4 0.4
UPE4
20 0.2

25 0
0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Stretch Amount of Shear

Fig. 9. Comparison of numerical solutions against analytical solutions for the deformation only problem: (a) Normalized stress T 11 =G versus stretch k behavior in simple
compression as well as plane strain compression. (b) Normalized stress T 12 =G and normalized stress-difference T 11  T 33 =G versus amount of shear c in simple shear. In all
cases the lines represent the analytical solution and markers the nite element solution.

is still xed at 300 K everywhere. In the second step, for the

C mechanical boundary conditions the nodes along faces AD and
D 350 K over 100 s. Further, we prescribe a ux of (i) 0.01 W/m2
A B
Fig. 10. Mesh and boundary conditions used in the verication of the diffusion
component of the UEL. The temperature is prescribed on the nodes indicated on
faces AD and BC, while uxes are prescribed on the portions of faces as shown.
BC are held xed. For the thermal boundary conditions, nodes on
face AD held xed at 300 K (the open circles in Fig. 10) and nodes
on face BC (the lled circles in Fig. 10) are ramped from 300 K to
on the indicated portion of face AB; and (ii) 0.02 W/m2 on the indi-
cated portion of face CD. All thermal ux boundary conditions are
applied instantaneously at the start of the step.
The numerical simulation is performed using built-in ABAQUS
elements, as well as our user elements for both plane strain and
axisymmetric geometries. Fig. 11 shows the nodal temperature
eld on the deformed body after 500 s of simulation time with
the outline of the initial body superimposed. Fig. 12 shows the
temperature of the mid-side node on faces AB and CD.
These results verify our basic element technology when dealing
with a transient diffusion only problem for plane strain and axi-
symmetric geometries. We have performed a similar analysis using
our three dimensional elements, however for brevity, the results
are not presented here.

NT11 NT11
652.5 652.9
609.2 609.6
565.9 566.3
522.6 523.0
479.3 479.7
436.0 436.3
392.6 393.0
349.3 349.7
306.0 306.4
262.7 263.1
219.4 219.8
176.1 176.5
132.8 133.2

(a) (b)
ABAQUS CPE4T UPE4
NT11 NT11
654.3 654.5
610.8 610.9
567.2 567.4
523.7 523.9
480.2 480.4
436.7 436.8
393.2 393.3
349.7 349.8
306.2 306.2
262.7 262.7
219.2 219.2
175.7 175.7
132.2 132.1

(c) (d)
ABAQUS CAX4T UAX4

Fig. 11. Nodal temperature eld in degrees K after 500 s of simulation time for (a) an ABAQUS CPE4T, (b) a UPE4, (c) an ABAQUS CAX4T, and lastly (d) a UAX4 element.

Please cite this article in press as: Chester, S.A., et al. A nite element implementation of a coupled diffusion-deformation theory for elastomeric gels. Int. J.
Solids Struct. (2014), http://dx.doi.org/10.1016/j.ijsolstr.2014.08.015
 S.A. Chester et al. / International Journal of Solids and Structures xxx (2014) xxxxxx 15

 
550 1 1 #  #T
v# vL vH  vL  vH tanh 8:2
2 2 D
500
where #T is the transition temperature of the gel, vL is the value
450 of v below the transition temperature, vH > vL is the value of
Abaqus Plane Strain v above the transition temperature, with D the width of the
400 transition in temperatures between vL and vH . 7
(K)

UPE4
 Constitutive equation for the Cauchy stress: Corresponding to the
350 Abaqus Axisymmetric
free energy (8.1), the Cauchy stress T is given by
UAX4  
300 @w
T J 1 2Fe Re Fe> J 1 G B  1 J s Kln J e 1: 8:3
250 @C
 Constitutive equation for the chemical potential: The chemical
200
0 100 200 300 400 500 potential l is given by
Time (s)
@wR 1  
l  X J e trT l0 R# ln1  / / v/2
Fig. 12. Temperature as a function of time for the mid-side node on faces AB and @cR 3
CD. Lines indicate the simulations using the built in ABAQUS elements, while 1 2
markers indicate the UEL.  XKln J e K Xln J e : 8:4
2
 Constitutive equation for the uid ux: We assume that the
Appendix C. A thermo-chemo-mechanically coupled theory
spatial uid ux, j, depends linearly on the spatial gradient of
the chemical potential, grad l, with the mobility tensor taken
With the aim of aiding other researchers in their development
to be isotropic so that
of UEL capabilities, in this appendix we summarize an extension
of the theory presented in Section 2 aimed at modeling thermally j m grad l; 8:5
responsive gels. In this situation, one must consider not only the
where m is a scalar mobility coefcient, we assume m in the
displacement and the chemical potential elds, but also the
same form as (4.1).
temperature eld. Within the Abaqus framework, such a coupled
 Constitutive equation for the heat ux: We assume that the spa-
set of PDEs is most readily solved using the UEL capability. In
tial heat ux, q, depends linearly on the spatial gradient of the
addition to the theory summarized here, we have included
temperature, grad #, with the thermal conductivity tensor taken
the source code and relevant associated les in the online
to be isotropic so that
Supplemental materials.
q k grad #; 8:6
C.1. Summary of the theory where k is the scalar thermal conductivity, which for simplicity
is taken to be constant.
C.1.1. Constitutive equations
In what follows we neglect thermal expansion effects since they C.1.2. Governing partial differential equations
are small compared to the large volume changes due to the inow/ The governing partial differential equations, when expressed in
outow of solvent in the body. The constitutive equations of the the deformed body, consist of
thermo-chemo-mechanically coupled theory are:
1. The local force balance for the macroscopic Cauchy stress,
 Free energy: A simple form of the free energy function which
accounts for the combined effects of mixing, swelling, and divT b 0; 8:7
elastic stretching is, with b a non-inertial body force, and T given by (2.9).
     2. The local balance for the uid concentration,
0 XcR 1
wR l cR R#cR ln v
1 XcR 1 XcR /_
   div j 0; 8:8
1  2  s 1 2 J X/2
G 3k  1  2 ln J J Kln J e f #: 8:1
2 2
in which the uid ux j is given by (8.5), and the chemical poten-
Here, G# NR kB # is the temperature dependent shear modulus tial l is given by (8.4).
of the network with N R the number of polymer chains per refer- 3. The local energy balance for the temperature,
ence volume, kB the Boltzmann constant, and # the absolute
c#_ div q h 8:9
temperature. The term f # is related to the specic heat of the
material and we leave it unspecied here. where the heat ux is given by (8.6), c is the specic heat (per
Next, we specialize the temperature dependence of the chi- unit volume), and h is the contribution due to any source/sink
parameter v# in the FloryHuggins expression (8.1) for the free terms. Readers are referred to Chester and Anand (2011) for
energy of mixing. Recall that this parameter represents the dis- more details on the specic contributions to h, in what follows
afnity between the polymer and the solventif v# is here, we neglect any such contributions and set h 0.
increased, the uid molecules are expelled from the gel and
the gel shrinks, while if v# is decreased, the gel absorbs more C.1.3. Boundary and initial conditions
uid and swells. Typically, thermally-responsive gels possess a We also need boundary and initial conditions to complete the
critical transition temperature, #T , and exhibit a larger degree theory. Let S u and S t be complementary subsurfaces of the
of swelling at temperatures below #T than at temperatures
above #T (e.g., Gotoh et al., 1998; Afroze et al., 2000; Xue 7
Other models have also been developed in an effort to capture the temperature
et al., 2001; Zhang et al., 2003; Liu et al., 2008). To model such and uid concentration dependence on thermal responsiveness in the literature, (c.f.,
experimentally-observed phenomena we assume that e.g., Cai and Suo, 2011).

Please cite this article in press as: Chester, S.A., et al. A nite element implementation of a coupled diffusion-deformation theory for elastomeric gels. Int. J.
Solids Struct. (2014), http://dx.doi.org/10.1016/j.ijsolstr.2014.08.015
16 S.A. Chester et al. / International Journal of Solids and Structures xxx (2014) xxxxxx

boundary @B of the body B in the sense @B S u [ S t and with the index A f1; 2; . . . ; Mg denoting the nodes of the element,
S u \ S t . Similarly let S l and S j be complementary subsurfaces uA , lA , and #A denoting nodal displacements, chemical potentials,
of the boundary: @B S l [ S j and S l \ S j . And lastly let S # and temperatures, and NA the shape functions. We employ a stan-
and S q be complementary subsurfaces of the boundary: dard Galerkin approach, in that the weighting elds are interpo-
@B S # [ S q and S # \ S q . Then for a time interval t 2 0; T lated by the same shape functions, viz.
we consider a pair of boundary conditions in which the displace- P 9
w1 wA1 NA ; >
=
ment u is specied on S u and the surface traction on S t : P
w2 wA2 NA ; 8:17
uu on S u  0; T; P >
;
8:10 w3 wA3 NA :
Tn t on S t  0; T;
a pair of boundary conditions in which the chemical potential is This leads to the system of coupled equations is solved using a New-
specied on S l and the uid ux on S j ton procedure by dening the following element-level residuals for
) the displacement, chemical potential, and temperature,
l l on S l  0; T; ! 9
8:11 R@NA R
A  >
>
j  n j on S j  0; T; A
Rui  Be T ij dv Se N ti da;
 >
>
@xj t >
>
>
>
and a pair of boundary conditions in which the temperature is spec- ! ! >
>
R _
/ R @N A
@ l R
 =
ied on S # and the heat ux on S q A A A
Rl Be N 2
dv  B e m dv Se N j da;
) J X/ @xi @xi j >
>
>
>
# # on S #  0; T;

!

 >
>
8:12 R R @N A
@# R >
>
 on S q  0; T;
q  n q A A _
R# Be N c# dv Be k dv  Se N q A
 da: >
>
;
@xi @xi j

with u  ; j; #,
 ; t; l  and q
, prescribed functions of x and t. The initial data
8:18
is taken as
In addition to the residuals, the following tangents are also required
uX; 0 u0 X; and lX; 0 l0 X and
for the iterative Newton solver:
#X; 0 #0 X in B: 8:13
A @Rl A A 9

The coupled set of Eqs. (8.7), (8.8), and (8.9), together with (8.10), Kuu AB  @R
@uB
u
; Klu AB  @uB
; K#u AB  @R#
@uB
>
;>
>
=
(8.11), (8.12), and (8.13) yield an initial boundary value problem
A @R A @R# A
Kul AB  @R u
@ lB
; K ll AB  @ llB ; AB
K #l  @ lB ;
for the displacement ux; t, the chemical potential lx; t, and the >
>
A @Rl A @R# A ;
>
temperature #x; t. Ku# AB  @R u
@#B
; K l# AB  @#B
; AB
K ##  @#B :
8:19
C.2. Numerical solution procedure
Details of each of the tangents listed above may be found in the
In this section, since the basic procedure follows from Section 3, source code.
we only highlight and summarize the major points of modication.
The full details may be found in the source code and input les, C.3. Numerical example
which are provided with the online Supplemental materials.
In the absence of body forces, the strong forms of the coupled To demonstrate the robustness of the developed user element
partial differential equations consist of the balance of forces, as well subroutine in simulating thermo-chemo-mechanically coupled
as the balance of uid content as described in (3.1), however now problems in this section we provide an illustrative numerical sim-
we also append the balance of energy, which is the heat equation, ulation. This example problem uses the same geometry and mesh
8 as shown in Fig. 4 for the constrained bilayer swelling (bending)
_
< c# divq 0 in B;
>
used previously in Section 4.2.
Balance of energy # # on S # ; 8:14
>
: For this example a two-step simulation is implemented; the

q  n q on S q : rst step is identical to the isothermal swelling as described in Sec-
tion 4.2 and shown in Fig. 5, while in the second step the temper-
Then, with w1 ; w2 , and w3 denoting three weighting (or test) elds
ature is varied and heat is conducted into the gel causing
which vanish on S u ; S l , and S # respectively, the corresponding
deswelling. For the rst step of the simulation we keep the same
weak forms are:
  9
R @w1 R   >
B
T : dv St w1  t da; >
>
> Table 2
@x >
>
>
> Material parameters for a representative thermally responsive
!   Z >
>
R _/ R @w2
 = elastomeric gel.
w 2 dv   j d v  w 2
j da; 8:15
B 2 B Parameter Value
J X/ @x Sj >
>
>
>
  Z >
> NR 2:43  1025 /m3
R   R @w3 >
>
_ >
> G N R kB # (at 298 K) 1 MPa
B
w 3 c # d v B
 q d v w3 
q da: ;
@x Sq K 100 MPa
S X 1:0  104 m3/mol
The body is approximated using nite elements, B Be , and vL 0.1
the trial solutions for the displacement, chemical potential, and vH 0.7
temperature are interpolated inside each element by #T 320 K
D 5.0 K
P 9
u uA N A ; > l0 0.0 J/mol
P A A = D 5  109 m2/s
l l N ;> 8:16
c
P ; 4:18  106 J/(m3 K)
# #A NA ; k 0:5 W/(m K)

Please cite this article in press as: Chester, S.A., et al. A nite element implementation of a coupled diffusion-deformation theory for elastomeric gels. Int. J.
Solids Struct. (2014), http://dx.doi.org/10.1016/j.ijsolstr.2014.08.015
 S.A. Chester et al. / International Journal of Solids and Structures xxx (2014) xxxxxx 17

350 0.380
346 0.374
341 0.369
337 0.364
333 0.359
328 0.353
324 0.348
320 0.343
315 0.338
311 0.332
307 0.327
302 0.322
298 0.317
0

(a)
350 0.465
346 0.456
341 0.447
337 0.438
333 0.429
328 0.420
324 0.411
320 0.402
315 0.393
311 0.384
307 0.375
302 0.366
298 0.357
0

(b)
350 0.528
346 0.527
341 0.527
337 0.526
333 0.526
328 0.525
324 0.525
320 0.524
315 0.524
311 0.523
307 0.523
302 0.522
298 0.522
0

(c)
Fig. 13. Contours of the temperature # (left, in degrees Kelvin) and polymer volume fraction / (right) at (a) the start, (b) 4 h, and (c) 12 h in the second step of the simulation.
Here the unfurling of the bilayer is due to the thermal responsiveness of the gel as the temperature is raised to 350 K.

mechanical and chemical boundary conditions as in Section 4.2
however we now impose the initial condition on the temperature
#0 298 K over the entire body. Also, in the rst step, nodes along
face CD are prescribed # 298 K while the other faces, including
the interface between the non-swellable substrate and gel, are pre-
scribed zero heat ux. In the second step however, after six hours
of swelling in the rst step, nodes along face CD are prescribed a
temperature increase to # 350 >K over 6 hours and the simula-
tion is allowed to run for another 6 h. The other faces, including
the interface are prescribed zero heat ux. We note that the non-
swellable substrate does not conduct heat nor diffuse solvent in this
simulation.
Table 2, which is based on Table 1, lists plausible representative
values for the material properties of a thermally responsive poly-
meric gel, which we have used in the following calculation.
Fig. 13 shows contours of the temperature and polymer volume
fraction in the deforming body. From Fig. 13 the coupling between
the temperature and degree of swelling is clearly visible. It is clear
that the change in the temperature locally changes the chemistry,
i.e., v#, which drives solvent out of the gel causing an unfurling of
the bilayer. We note that since the layer is thin, and uid diffuses
much more slowly than heat, the temperature contours are nearly
constant at any given instant of time.
This numerical example shows the use of our user elements in a
thermo-chemo-mechanically coupled situation for simulating
thermally responsive gels. Along with the online Supplemental
materials this extra example should prove useful to researchers
trying to adapt UELs for their own use in other coupled problems.
Appendix D. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.ijsolstr.2014.08.
015.
References
ABAQUS/Standard, 2013. SIMULIA, Providence, RI.
Afroze, F., Nies, E., Berghmans, H., 2000. Phase transitions in the system poly(N-
isopropylacrylamide)/water and swelling behaviour of the corresponding
networks. J. Mol. Struct. 554, 5568.
Al-Athela, K., Loeffel, K., Liu, H., Anand, L., 2013. Modeling decohesion of a top-coat
from a thermally-growing oxide in a thermal barrier coating. Surf. Coat.
Technol. 222, 6878.
Anand, L., Aslan, O., Chester, S.A., 2012. A large-deformation gradient theory for
elasticplastic materials: strain softening and regularization of shear bands. Int.
J. Plast. 30, 116143.
Beebe, D.J., Moore, J.S., Auer, J.M., Yu, Q., Liu, R.H., Devadoss, C., Jo, B., 2000.
Functional hydrogel structures for autonomous ow control inside microuidic
devices. Nature 404, 588590.
Bhavsar, R., Vaidya, N.Y., Ganguly, P., Humphreys, A., Robisson, A., Tu, H., Wicks, N.,
McKinley, G.H., Pauchet, F., 2008. Intelligence in novel materials. Oileld Rev.
20, 3241.

Please cite this article in press as: Chester, S.A., et al. A nite element implementation of a coupled diffusion-deformation theory for elastomeric gels. Int. J.
Solids Struct. (2014), http://dx.doi.org/10.1016/j.ijsolstr.2014.08.015
18 S.A. Chester et al. / International Journal of Solids and Structures xxx (2014) xxxxxx
Cai, S., Hu, Y., Zhao, X., Suo, Z., 2010. Poroelasticity of a covalently crosslinked
alginate hydrogel under compression. J. Appl. Phys. 108, 113514.
Cai, S.Q., Suo, Z., 2011. Mechanics and chemical thermodynamics of phase transition
in temperature-sensitive hydrogels. J. Mech. Phys. Solids 59, 22592278.
Chan, G., Mooney, D.J., 2008. Trends Biotechnol. 26, 382392.
Chester, S.A., Anand, L., 2010. A coupled theory of uid permeation and large
deformations for elastomeric materials. J. Mech. Phys. Solids 58, 18791906.
Chester, S.A., Anand, L., 2011. A thermo-mechanically coupled theory for uid
permeation in elastomeric materials: application to thermally responsive gels. J.
Mech. Phys. Solids 59, 19782006.
Chester, S.A., 2012. A constitutive model for coupled uid permeation and large
viscoelastic deformation in polymeric gels. Soft Matter 8, 82238233.
Di Leo, C.V., Anand, L., 2013. Hydrogen in metals: a coupled theory for species
diffusion and large elastic-plastic deformations. Int. J. Plast. 43, 4269.
Di Leo, C.V., Rejovitsky, E., Anand, L., 2014. A CahnHilliard-type phase-eld theory
for species diffusion coupled with large elastic deformations: application to
phase-separating Li-ion electrode materials. J. Mech. Phys. Solids 70, 129.
Doi, M., 2009. Gel dynamics. J. Phys. Soc. Jpn. 78, 052001.
Duan, Z., Zhang, J., An, Y., Jiang, H., 2013. Simulation of the transient behavior of gels
based on an analogy between diffusion and heat transfer. J. Appl. Mech. 80,
041017.
Duda, F.P., Souza, A.C., Fried, E., 2010. A theory for species migration in nitely
strained solid with application to polymer network swelling. J. Mech. Phys.
Solids 58, 515529.
de Souza Neto, E.A., Peric, D., Dutko, M., Owen, D.R.J., 1996. Design of simple low
order nite elements for large strain analysis of nearly incompressible solids.
Int. J. Solids Struct. 33, 32773296.
Giner, E., Sukumar, N., Tarancon, J.E., Fuenmayor, F.J., 2009. An ABAQUS
implementation of the extended nite element method. Eng. Fract. Mech. 76,
347368.
Gotoh, T., Nakatani, Y., Sakohara, S., 1998. Novel synthesis of thermosensitive
porous hydrogels. J. Appl. Polym. Sci. 69, 895906.
Gurtin, M.E., Fried, E., Anand, L., 2010. The Mechanics and Thermodynamics of
Continua. Cambridge University Press, Cambridge.
Henann, D.L., Chester, S.A., Bertoldi, K., 2013. Modeling of dielectric elastomers:
design of actuators and energy harvesting devices. J. Mech. Phys. Solids 61,
20472066.
Henann, D.L., Bertoldi, K., 2014. Modeling of elasto-capillary phenomena. Soft
Matter 10, 709717.
Holmes, D.P., Roche, M., Sinha, T., Stone, H.A., 2011. Bending and twisting of soft
materials by non-homogenous swelling. Soft Matter 7, 5188.
Please cite this article in press as: Chester, S.A., et al. A nite element implementation of a coupled diffusion-deformation theory for elastomeric gels. Int. J.
Solids Struct. (2014), http://dx.doi.org/10.1016/j.ijsolstr.2014.08.015
Hong, W., Zhao, X., Zhou, J., Suo, Z., 2008. A theory of coupled diffusion and large
deformation in polymeric gel. J. Mech. Phys. Solids 56, 17791793.
Hong, W., Liu, Z., Suo, Z., 2009. Inhomogeneous swelling of a gel in equilibrium with
a solvent and mechanical load. Int. J. Solids Struct. 46, 32823289.
Kleverlaan, M., Van Hoort, R.H., Jones, I., 2005. Deployment of swelling elastomer
packers in Shell E&P. In: SPIE/IADC Drilling Conference. Amsterdam.
Lee, H., Zhang, J., Jiang, H., Fang, N.X., 2012. Prescribed pattern transformation in
swelling gel tubes by elastic instability. Phys. Rev. Lett. 108, 214304.
Liu, K., Ovaert, T.C., Mason, J.J., 2008. Preparation and mechanical characterization
of a PNIPA hydrogel composite. J. Mater. Sci. Mater. Med. 19, 18151821.
Liu, Z., Hong, W., Suo, Z., Swaddiwudhipong, S., Zhang, Y., 2010. Modeling and
simulation of buckling of polymeric membrane thin lm gel. Comput. Mater.
Sci. 49, 560564.
Loeffel, K., Anand, L., Gasem, Z., 2013. On modeling the oxidation of high-
temperature alloys. Acta Mater. 61, 399424.
Lucantonio, A., Nardinocchi, P., Teresi, L., 2013. Transient analysis of swelling-
induced large deformations in polymer gels. J. Mech. Phys. Solids 61, 205218.
Marcombe, R., Cai, S., Hong, W., Zhao, X., Lapusta, Y., Suo, Z., 2010. A theory of
constrained swelling of a pH-sensitive hydrogel. Soft Matter 6, 784793.
Pandey, A., Holmes, D.P., 2013. Swelling-induced deformations: a materials-dened
transition from macroscale to microscale deformations. Soft Matter 9, 5524.
Park, K., Paulino, G.H., 2012. Computational implementation of the PPR potential-
based cohesive model in ABAQUS: educational perspective. Eng. Fract. Mech.
93, 239262.
Peppas, N.A., Hilt, J.Z., Khademhosseini, A., Langer, R., 2006. Hydrogels in biology
and medicine: from molecular principles to bionanotechnology. Adv. Mater. 18,
13451360.
Tanaka, T., Fillmore, D.J., 1979. Kinetics of swelling of gels. J. Chem. Phys. 70, 1214
1218.
Toh, W., Liu, S., Ng, T.Y., Hong, W., 2013. Inhomogeneous large deformation kinetics
of polymeric gels. Int. J. Appl. Mech. 5, 1350001.
Xue, W., Champ, S., Huglin, M.B., 2001. New superabsorbent thermoreversable
hydrogels. Polymer 42, 22472250.
Yoon, J., Cai, S., Suo, Z., Hayward, R.C., 2010. Poroelastic swelling kinetics of thin
hydrogel layers: comparison of theory and experiment. Soft Matter 6, 6004
6012.
Zhang, X., Wang, F., Chu, C., 2003. Thermoresponsive hydrogel with rapid response
dynamics. J. Mater. Sci. Mater. Med. 14, 451455.
Zhang, J., Zhao, X., Suo, Z., Jiang, H., 2009. A nite element method for transient
analysis of concurrent large deformation and mass transport in gels. J. Appl.
Phys. 105, 093522.