View Article Online / Journal Homepage / Table of Contents for this issue

Soft Matter Dynamic Article Links < C

Cite this: Soft Matter, 2011, 7, 9474

www.rsc.org/softmatter PAPER
Propagation of a brittle fracture in a viscoelastic fluid
Herv
e Tabuteau,†ab Serge Mora,ab Matteo Ciccotti,abc Chung-Yuen Huid and Christian Ligoure*ab
Received 1st June 2011, Accepted 21st July 2011
DOI: 10.1039/c1sm06024d

During pendant drop experiments, a model physical gel made from oil in water microemulsion droplets
reversibly linked together by triblock copolymers, exhibit a very peculiar filament rupture
Published on 30 August 2011. Downloaded on 9/24/2018 8:45:22 AM.

corresponding to highly brittle failure of a viscoelastic fluid. The fracture propagation has been tracked
by high speed videomicroscopy. Analysis of the time evolution of the fracture profile shows that the
fracture is purely elastic and reversible without any significant bulk and interfacial viscous dissipation.
However, since the elastic moduli of such complex fluids are low, hyper elastic corrections have to be
taken into account for a quantitative analysis of the fracture profile. This brittle behavior is well
explained by a hyperelastic generalization of the viscoelastic trumpet model of de Gennes. The velocity
of the fracture’s propagation is measured and compared to the predictions of a simple microscopic
model.

1 Introduction experiments were essentially limited to the measurements of

extensional viscosities. On the other hand, quantitative analysis
Although fracture mechanics were originally based on an equi- of the crack propagation and morphology in viscoelastic solids in
librium transformation between the mechanical energy stored in planar geometry have been performed.4,5,17 However, to the best
an elastic body and the surface energy that is necessary to of our knowledge, the systematic analysis of morphology and
produce a fracture in the body,1 in most practical cases the high propagation of a crack in a viscoelastic fluid does not have been
concentration of stress in the crack tip region implies the acti- yet reported.
vation of dissipative processes that adsorb a large amount of Here we present a system where the Griffith theory is appli-
energy, which is generally dominant in the energy balance of cable with an excellent approximation, consisting in a soft
fracture propagation. This energy dissipation implies an irre- transient network gel in which nanodroplets of an oil-in water
versibility of fracture and a velocity dependence of the fracture microemulsion are reversibly linked by telechelic polymers18
energy G(V).2,3 In his famous seminal work Griffith1 tested its (Fig. 1). This system is a model Maxwell fluid. Indeed, its linear
energy balance criterion on a very brittle solid such as glass where rheological properties can be characterized by a shear plateau
energy dissipation is considered to be minimal. In less brittle modulus mN and a single dominant relaxation time s.18 More-
materials such as polymers or metals, the fracture energy can be over, it does not exhibit any shear thinning, nor shear thickening
hundreds to thousands of times larger than the surface tension nor shear banding until it breaks, as shown from flow
due to the high amount of plastic deformation in the neighbor- curves.11,12,14 This is in contrast to the behaviour of one of the
hood of crack tips. more popular Maxwell fluids, i.e. solutions of entangled worm-
Fracture in reversible physical gels is less documented and like micelles.
understood than in solid materials but attracts a great deal of
interest in recent years.4–16 Filament stretching extensional rhe-
ometry8–10 or capillary breakup rheometry experiments9,10 have
been used to study the fracture of networks of associative poly-
mers or solutions of entangled wormlike micelles, but these

a
Universit

e Montpellier 2, Laboratoire Charles Coulomb UMR 5221,
F-34095, Montpellier, France. E-mail: [email protected]
b
CNRS, Laboratoire Charles Coulomb UMR 5221, F-34095, Montpellier,
France
c
Fig. 1 A schematic of a bridged microemulsion. The telechelic polymers
Laboratoire PPMD-SIMM, UMR CNRS 7615, ESPCI, 10 rue
can either link two oil droplets or loop on a single one. (Left) Before the
Vauquelin, 75005 Paris, France
d
Department of Theoretical and Applied Mechanics, 322 Thurston Hall, crack nucleation (bold dashed line) polymers can bridge oil droplets on
Cornell University, Ithaca, NY 14853, USA both sides of the bold dashed line. (Right) When the crack occurs, the
† Present address: Institut de Physique de Rennes, UMR UR1-CNRS same polymers cannot cross the bold dashed line any more and form
6251, Universit
e de Rennes I, B^at. 11A, 35042 Rennes Cedex, France. bridges in the other directions or loops.

9474 | Soft Matter, 2011, 7, 9474–9483 This journal is ª The Royal Society of Chemistry 2011
 View Article Online

In a previous paper,12 some of us have investigated the fracture its tentative extension to the hyperelastic case. In section 5,
initiation of this Maxwell fluid using a pendant drop experiment, a simple theoretical approach is proposed to understand
which corresponds to a pure elongational stress condition due to the velocity of fracture propagation. Finally, in section 6, the
the lack of contact with solid interfaces near the fracture region. experimental results are discussed in the framework of the
We have shown that the fracture initiation process is governed by models developed in sections 4 and 5.
the thermally activated nucleation of a critical crack in the
polymeric network. In this approach, the rupture stress was
2 Materials and methods
predicted to be on the order of the shear modulus, in very good
agreement with experimental data. One of the key points to 2.1 Description of the transient network gel
understand this mechanism of fracture is the bond reversibility
and the corresponding relevant ultra low interfacial energy The system we used is composed of an oil-in-water droplet
needed to nucleate the crack. This interfacial tension results from microemulsion to which telechelic polymers are added (Fig. 1).
the loss of conformational entropy of polymeric bonds near This system was previously described by Filali et al.23 The o/w
a crack interface and is typically on the order of few mN m
1.19 microemulsion involves a cationic surfactant, cetyl-pyridinium
The aim of this paper is the rationalization and the quantifi- chloride CPCl, and a cosurfactant n-octanol. The droplets are
Published on 30 August 2011. Downloaded on 9/24/2018 8:45:22 AM.

cation of the crack propagation for such a Maxwell fluid in this swollen with decane and dispersed in 0.2 M NaCl brine. The
droplets are spheres of radius b ¼ 62 A  and were found to be
failure geometry, by a time-resolved analysis of the shape of the
growing crack. We want to address the following questions: robust to variations of both the microemulsion concentration
What is the fracture energy? What is the crack propagation and of the amount of added polymer.23 The volume fraction in oil
speed? We will show that the fracture profile has a parabolic droplets is fixed to f ¼ 10%. The polymer chains (Poly-ethylene
shape that can be rationalized using a finite elasticity theory20,21 oxide) of molecular weight 10 kDa are grafted at both ends with
that needs to be used, because of the large deformations aliphatic chains of eighteen CH2 groups. After modification, the
exhibited in this soft material. The only ingredients affecting the degrees of substitution of the hydroxyl groups were determined
energy balance in the fracture mechanism of a brittle material are by NMR and were found to be larger than 98%. These hydro-
the bulk elastic modulus mN and the surface tension g. According phobic end groups (stickers) anchor reversibly into the micro-
to such a description, the fracture energy G is substantially emulsion droplets. The polymer amount is represented by the
independent of fracture velocity (once the velocity is enough for apparent connectivity R, i.e. the average number of hydrophobic
having a purely elastic material response), and the crack velocity stickers per droplet. Far above the percolation threshold this
should only be limited to the Rayleigh waves propagation model system behaves as an elastic network with a shear modulus
velocity.22 However, for the complex fluid we consider, the mN ¼ nkBT (n is the number density of linking chains, kB the
propagation of the crack takes place at a constant velocity of the Boltzman’s constant, T the temperature).18 In this regime the
order of few mm s
1, which is much smaller than the Rayleigh connectivity is higher than 5. In this study we use two fluids with
waves velocity in the gel cR x 1 m s
1. the connectivity equal to 6 and 12. In the following, we named
The origin for this characteristic velocity must be sought at the those fluids R6 and R12.
scale of the microscopic mechanisms of crack propagation, i.e.
the debonding mechanisms of the polymer chains grafted on the 2.2 Rheological properties
oil droplets. The energetics of polymer debonding are very weak
The linear properties of all the samples are Maxwellian18 (Fig. 2).
compared to the surface tension of the gel. This is constituted by
The elastic shear modulus mN is controlled by the density of the
a reversible term gpol associated to the free energy of purely
entropic origin required to ungraft the polymers along a unit
interface and an irreversible term associated to the Stokes dissi-
pation caused by the motion of the oil droplet in the viscous fluid
(water) that is necessary to propagate the stresses to the new
crack tip. These two terms describe the wet fracture of the
polymer network inside the solvent and our claim is that they
determine the time scale for fracture propagation. However, they
are both energetically very small in relation to the energy to
create the two new interfaces between the gel and the air, and
they are thus not easily accessible by experimentally measuring
the fracture energy. The high degree of reversibility of the frac-
ture is thus related to a significant decoupling between the
mechanisms of creation of new gel/air interfaces (which dominate
the energetics) and the mechanisms of network bond breaking on
the polymer scale (which determine the time scales for crack
propagation). Fig. 2 Frequency sweep experiments (strain amplitude: 10%). Storage
The paper is organized as follow. Section 2 is devoted to modulus, G0 , (filled symbols) and loss modulus, G0 , (unfilled symbols) as
materials and methods. The time resolved analysis of the fracture a function of the frequency u for the fluids R6 (circles) and R12 (trian-
profiles is reported in section 3. Section 4 presents a review of the gles). Solid lines correspond to fits by a Maxwell model which give us the
viscoelastic trumpet model of de Gennes for a Maxwell fluid and elastic shear modulus and the relaxation time of each fluid.

This journal is ª The Royal Society of Chemistry 2011 Soft Matter, 2011, 7, 9474–9483 | 9475
 View Article Online

active polymer and the terminal time s is determined by the

average residence time of the hydrophobic stickers in the oil
droplets. Since the material is incompressible, which implies that
its Poisson ratio is n ¼ 0.5, the Young modulus is EN ¼ 3mN. All
the rheological measurements were performed with an RFS III
controlled-strain rheometer at 23  C. The shear modulus and the
relaxation time are, respectively, equal to 330 Pa and 0.6 s for R6,
and 2400 Pa and 2 s for R12.

2.3 Fracture experiments

We carried out pendant drop experiments with our fluid. This
very simple test allows us to get a pure elongational flow, without
the influence of a solid surface on the flow properties as is the case
Fig. 3 A sequence of images of the fall of the R6 fluid. Regime 1
in the gap of an extensional rheometer. A syringe pump (KDS
Published on 30 August 2011. Downloaded on 9/24/2018 8:45:22 AM.

corresponds to images a and b and regime 2 corresponds to images c to g.

200 from KD Scientific, USA) was used to form the drops with
At the beginning of regime 2 we define the radius of the filament R* ¼
a fixed volume of 50 ml at a constant rate of 2 ml h
1 for both D*/2 and L0, the length of the drop. The white scale bar corresponds to
fluids. The material, initially in the syringe, flows through a lower 2.5 mm.
plastic tube of a diameter of 2.596 mm and a drop emerges at the
tube outlet that is enclosed in a glass box to reduce air currents. flow results from the balance between the surface tension, the
These conditions ensure that no elastic instability modifies the viscous force and the weight (Fig. 3a and b). In a second regime
extensional flow while the drop starts to form underneath the (Fig. 3 images c to g), the filament is stretched by the falling drop,
tube. All the experiments were performed at room temperature whose shape remains approximately constant and is parabolic.
of 23  C. We use regular image analysis to determine the length In this regime, the drop has an initial length L0 and the initial
and the radius of the drop. We also use a fast camera (Photron radius of the filament is R* (Fig. 3c).
FastCam PCI CCD camera) equipped with a macro lens to study In the following, we suppose that inertia is negligible because
the fracture propagation with a pixel size resolution of 8  8 mm2. we work with very viscous fluids (Reynolds number is of the
The images were analyzed using NIH ImageJ (NIH, freely order of 0.01). The filament is assumed to be a cylinder in an
available for download at the ImageJ website URL: http://rsb. extensional flow, assuming that there is no flow from the filament
info.nih.gov/ij/). The fracture movies have been selected in such into the drop. The evolution of the radius R, normalized by R*,
a way that the fracture propagates in a plane perpendicular to the with time in regime 2 and the corresponding sequence of images
camera direction. However, in most of the experiments the crack of the fall are very similar to those predicted by the inertialess
may nucleate everywhere around the filament and then the slender-drop theory and finite element computation of Stokes
fracture propagates in a plane which is not orthogonal to the et al.24 (cf. Fig. 4). The filament thinning is rather slow and its
camera direction. Approximately 50 experiments were per- dynamics can be described by a simple balance between the
formed for each fluid in order to meet the orthogonality condi- viscous and gravitational forces, the surface tension
tion in a couple of cases where the fracture profile was then being neglected. The variation of the radius filament is given by
analyzed. We got the critical stress at the rupture of the drop sf in
the following way. We measured the diameter D0 of the drop
where the fracture occurs and we weighted the mass of the falling
part with the help of a laboratory balance of accuracy 1 mg (TP
303 Denver Instrument, Germany) placed underneath the injec-
tion set-up. In this condition the stress is measured with a preci-
sion of 5%. Finally the stress at the rupture is equal to sf ¼
mg/(pR2rupture)
2gs/D0 with gs/D0 the radial stress corresponding
to the Laplace pressure and mg/(pR2rupture) is the tensile stress
corresponding to the weight of the failing part of the drop at the
rupture.

3 Results
The sequence of images in Fig. 3 shows the evolution of the drop
Fig. 4 The minimum radius R along the filament versus time for the R6
under gravity from its formation to the final break-up for the R6
fluid (crosses). The letters on the graph corresponds to the labels on the
fluid. After the injection of the fluid there is the formation of
images of Fig. 3. The black line corresponds to the fit of the data by the
a drop followed by the stretching of the filament by the falling formula defined by Stokes for the purely viscous fall,24 with L0 ¼ 5.25,
drop. The flow can be separated in two regimes. In the first tv ¼ 15.57 s and R* ¼ 0.84 mm. The variation of the deformation rate
regime, the drop of fluid begins to fall when its weight exceeds the against the time left to break-up (tv
t) deduced from this formula is
surface tension retaining force. This balance of forces determines plotted in the inset. The dashed cross corresponds to the moment where
the length scale of the drop, which is about 2 mm. The resulting the fracture occurs.

9476 | Soft Matter, 2011, 7, 9474–9483 This journal is ª The Royal Society of Chemistry 2011
 View Article Online

R(t) ¼ R*(1
t/2tv)1/2 with tv ¼ 3h/(rgL0). L0 is the initial length (thermally activated crack) of a microcrack within the oil
of the drop in the second regime, r is the mass density and g droplet/telechelic polymers network, the microcrack being filled
gravity’s acceleration (cf. Fig. 4). Most of the thinning process of up with solvent. In the following we discuss in detail the second
the filament takes place during this viscous fall for R* down to step, which corresponds to the destabilization of the capillary
0.2R*. For smaller radii there is a fast rupture of the filament bridge and the propagation of a dry fracture through the
(Fig. 3g), described in detail later on, corresponding to the material.
fracture propagation. From the above expression of the time It is worth noting that from the beginning of the propagation
evolution of the radius we calculate the deformation rate, 3_ ¼ up to the complete fracture (i.e., when the sample is separated in

2/R(dR/dt)(see inset of Fig. 4). In this way, we get the defor- two parts) the fracture profile exhibits a parabolic shape (Fig. 5a)
mation rate when the fracture starts to propagate, which is equal as expected for an elastic solid breaking under tension. These
to 1 s
1 and 0.06 s
1 for R6 and R12 fluids, respectively. It is observations were confirmed by quantitative analysis of the
interesting to notice that elastic effects do not affect the time fracture profile u(x) on the overall crack propagation across the
variation of the radius of the filament for t > s25 and so are not sample. Different fracture profiles measured in the fracture
visible in Fig.4 because s ¼ 0.6 s. The total duration of the fall moving frame and a parabolic fit
before the opening of a crack is typically on the order of 50 to pffiffiffi
uðxÞ ¼ a x
Published on 30 August 2011. Downloaded on 9/24/2018 8:45:22 AM.

(1)
200 s, much larger than the relaxation time of the Maxwell
fluid s ( 1 s. The tensile stress increases because of the decrease are represented in Fig. 5b.
of R(t), until a crack nucleates at the surface of the filament for We could be tempted to interpret this parabolic shape
a critical stress sf x 0.5E.12 Then the fracture propagates across according to the linear elastic fracture mechanics solutions:26
the sample and eventually leads to the rupture of the drop rffiffiffiffiffiffi
(Fig. 5a). In the experiments presented here, the fracture mech- KI 8x
uðxÞ ¼ 0
anism is a two step process. It has been shown in a previous E p
paper,12 that the first step consists of a spontaneous nucleation where the (local) stress intensity factor KI can be bound to the
strain energy release rate by G ¼ K2I/E(1
n2), but in soft solids
the crack tip region exhibits very large deformations that require
the use of finite elasticity theories.27 The finite elasticity
formalism for crack tip stress and displacement fields in a Neo-
Hookean solid is developed in the Appendix 1 for a 2D plane
strain problem. The nonlinear character of the problem requires
the fracture energy to be estimated by the J-integral method28
(the method is summarized in Appendix 1) and it can be related
to the parameter a of the parabolic crack opening profile by:
pmN a2
J¼ (2)
4
One of the main aims of the present work will be to show that
finite elasticity fully describes the behavior of our gel during this
rapid fracture experiment (cf. sections 4 and 6). This local esti-
mation of the J-integral thus also provides an estimate of the
strain energy release rate G ¼ J.
We stress the point that the use of eqn (2) essentially assumes
a 2D symmetry of the crack profile, which is not realistic in the
case of a cylindrical filament. However, since the movie has been
selected in order to present an excellent orthogonality between
the camera orientation and the direction of crack propagation,
the observed opening profile provides a good local estimate
of the fracture energy, the measurement being more accurate for
shorter cracks.
The variation of the estimated fracture energy G and the
Fig. 5 (a) Pictures of the propagation of the fracture across sample R6 measured length of fracture L as a function of the time to rupture
from the right to the left. The time left to achieve complete fracture of the trupt
t are represented in Fig. 6. Two distinct regimes are clearly
filament corresponding to each picture labelled with a letter is: evident depending on the length of the crack:
a (8.50 ms), b (5.16 ms), c (2.67 ms), d (2.00 ms), e (1.00 ms), f (0.33 ms), g
(i) for L < 0.1D0, both the fracture energy and the crack speed
(0.17 ms) and h (0.ms). The last picture on the right shows almost all of
remain almost constant with values roughly equal to G z
the elongated drop, with the crack being well developed. The white scale
bar corresponds to 0.1 mm. (b) Fracture profiles u(x) for different times
90 mJ m
2 and V ¼ 4 mm s
1. It turns out that the estimated value
before the break-up in the fracture moving frame, corresponding to the for G is roughly twice the surface tension gs x 45 mJ m
2 of the
pictures of the part a of this figure. The black line is a parabolic fit cor- solvent, i.e. the stabilised oil-in-water droplet microemulsion
responding to eqns (1) and (2) with J ¼ 2gs. We report only the profile for without telechelic polymers. We remark firstly that the surface
L < 0.1D0. energy needed to pull-out the hydrophobic stickers from the oil

This journal is ª The Royal Society of Chemistry 2011 Soft Matter, 2011, 7, 9474–9483 | 9477
 View Article Online

viscoelastic trumpet,17,29,30 which presents a qualitative theoret-

ical analysis of the dissipative processes during the bulk fracture
in a viscoelastic material. This model was initially conceived to
explain at the level of scaling laws the remarkable relation
between the fracture energy per unit area G(V) at crack velocity
V and G0, the limiting value of the fracture energy at zero crack
rate that was reported for elastomeric materials:2,3

G(V) ¼ G0 + GV(V) ¼ G0(1 + f(aTV)) (3)

where aT is the temperature shift factor given by the Williams-

Landel-Ferry equation.31
A similar argument can be used to express the fracture energy
for a fracture of length L(t) moving at velocity V. The trumpet
model predicts the following scaling form for the dissipated
Published on 30 August 2011. Downloaded on 9/24/2018 8:45:22 AM.

energy term GV(V):

Z Z  
T S_ 1 ​ 1 ​ s g_ 
GV ðV Þx x dxdysg_ ¼ dxdyRe
V V V 2
Z  
1 ​ s 20 um00 ðuÞ 

¼ rdr  (4)
V 2 m0 ðuÞ2 þm00 ðuÞ2 u¼V=r

where the complex strain g is related to the complex stress s

through g ¼ s =mðuÞ and the distance to the crack tip r is related
to the frequencies of the excitation u by the scaling relation
u(r) ¼ V/r. T is the temperature, S_ is the entropy creation rate
and s0 ¼ js j.
The complex modulus m(u) of a Maxwell fluid as function of
frequency is the following:
00 ius
mðuÞ ¼ m0 ðuÞ þ im ðuÞ ¼ mN (5)
Fig. 6 (a) Fracture energy versus the time before complete fracture 1 þ ius
trupt
t for R6 (filled circles) and R12 (unfilled circles) fluids. The letters
still correspond to the picture in Fig. 5a associated with the R6 fluid At low frequency (us  1), the modulus is purely imaginary (m ¼
fracture. The dotted line corresponds to G ¼ 2gs. (b) Evolution of the iumNs) and the material behaves as a liquid of viscosity h ¼ mNs.
total length of the fracture L (D0
L in the graph) with the time before At high frequency (us [ 1) we are dealing with an elastic solid
complete fracture trupt
t. The legend is the same as for part a. The of elastic modulus m z mN. So from eqns (4) and (5), one gets
diameter of the filament when the crack occurs at the surface of the
m00 ðuÞ 1
material D0 is equal to 0.584 and 0.407 mm for R6 and R12 fluids, ¼ (6)
respectively. The dotted lines correspond to the linear fit of the data m0 ðuÞ2 þm00 ðuÞ2 mN us
which gives a constant crack speed V ¼ dL/dt ¼ (4.00  0.02) mm s
1.
Putting eqn (6) into (4) one gets
ð umax
V du
droplet gpol, of the order of 10 mN m
1 12 is completely negligible. GV ðV Þ ¼ s20 (7)
2mN s umin u3
Secondly, the relation G z 2gs strongly supports the proposal
that there is no significant dissipative contribution in the crack where the limiting values umin ¼ V/L and umax ¼ V/l define the
tip region. In this regime the crack speed is also constant and is range of frequency over which the material is excited, l being the
roughly equal to 4 mm s
1 (Fig. 6b) length of a small microscopic nonlinear zone, typically on the
(ii) for L > 0.1D0, the estimated fracture energy and the crack 17 (see Fig. 7).
order of 100 A
speed increase when we go closer to the complete break-up of the
filament. It’s worth noting that the results concerning G must be
interpreted with caution as the finite size effect may be important.
It is then questionable whether the use of eqn (2), valid for a 2D 4.1 Linear elastic case
plane strain geometry, is still justified. de Gennes analysis is founded on the consideration that for
a viscoelastic medium, the scaling law for the stress s as a func-
tion of the distance r from the crack tip is still equivalent to that
4 Hyperelastic correction of the trumpet model for
which we have for a steadily growing mode I interface in plane
the fracture in a Maxwell fluid stress or plane strain in an elastic medium
pffiffi pffiffiffiffiffiffiffiffiffiffi
In order to provide a rationale for the presented behaviour, we s x KI = r z KI u=V , where KI is the applied stress intensity
propose here a development of the de Gennes model of the factor, and G0 ¼ K2I/mN is the fracture energy that would be

9478 | Soft Matter, 2011, 7, 9474–9483 This journal is ª The Royal Society of Chemistry 2011
 View Article Online

regime takes place at short propagation times t* < s and can

constitute the whole crack propagation if there is some maximal
cut-off for the fracture length Lmax < Vs, as is the case for the
filament rupture experiment reported in this paper, where Lmax ¼
D0 z 600 mm is the diameter of the filament just before failure. In
this purely elastic regime of the fracture propagation, where the
viscous region does not appear because of the finite size of the
sample, the profile u(x) of the fracture should be parabolic,28 and
should not depend on the rate of propagation of the fracture.
(iii) Viscous dissipation will occur only if Lmax > Vs. In this
regime, eqn (8) shows that G(V) > G0, provided that l remains
small with respect to the sample dimension. In this zone (Vs <
x < L(t)), de Gennes has shown that the scaling form s  x
1/2 for
the stress components remains valid for a viscoelastic medium
leading to the expected profile u(x)  x3/2. The sign change in the
Published on 30 August 2011. Downloaded on 9/24/2018 8:45:22 AM.

Fig. 7 A schematic representation of the space and time scales associ- concavity of the fracture profile at x  Vs is at the origin of the
ated to a crack of length L moving with velocity V in a Maxwell fluid name ‘‘trumpet’’ for this model. Such a trumpet profile has been
according to the trumpet model.17 A small microscopic nonlinear zone of experimentally observed for adhesive fractures in polymer
length l is represented in black. The behaviour of the material is solid like melts.17
at scales smaller than Vs, then fluid like at larger scales.

4.2 Hyperelastic case

present in an elastic medium under the same conditions. Eqn (7) The crack tip stress singularity mathematically implies very large
for the dissipated energy thus reads: strains in a region close to the crack tip with a characteristic size
  Rtip  G0/mN that questions the application of linear elasticity to
K2 1 1 L
‘ L
GV ðV Þ ¼ I
¼ G0 x G0 (8) fracture mechanics.27 For most materials this region is smaller
mN s umin umax Vs Vs
than the fracture process zone and is thus not relevant. However,
And so, if L < Vs, then (G(V)
G0)/G0 x L/Vs < 1, and the bulk for very soft materials this region can become significant and
viscous dissipation is negligible. It is convenient to express this must be taken into account. Four our gels it should be in the 100
characteristic ratio as: mm range, which is large compared to the small cohesive zone l
L t
 and it is of the same order of magnitude as the characteristic
¼ (9) diameter of the breaking gel filament. We discuss in this section
Vs s
the generalization of the de Gennes argument to the estimation
where t* ¼ L/V is the characteristic time of propagation of of the dissipated energy in the hyperelastic case.
a crack of length L at the velocity V. The condition for neglecting The crack tip solutions for plane strain finite elasticity are
dissipation just reads t* < s: the characteristic time needed to discussed in Appendix 1, as well as the computation of the J
create a crack of length L is smaller than the relaxation time of integral. When considering the scaling of the true Cauchy stress
the viscoelastic fluid. s(C) (eqn (34)), which is the most physically relevant description
For a simple Maxwell fluid exhibiting a crack propagation we of stress, we remark that the stress components present different
can distinguish three spatial regions with different properties scaling relations as a function of the distance r from the crack tip,
corresponding to three regimes of frequencies: notably r0, r
1/2and r
1. The second one is the same as in the linear
(i) 0 < x < l: directly ahead of the crack tip there is a small elastic case. At the level of the scaling law analysis of the trumpet
microscopic nonlinear zone of length l, independent of the model, and omitting all numerical prefactors, we can estimate the
separation rate, where the fracture process leads to the term G0 of contributions of each component of the Cauchy stress tensor to
the fracture energy in eqn (3). the dissipated energy from eqns (7) and (34).
(ii) l < x < Vs: in this region, the complex modulus is essen- The r0 component s(C) (11)
11 gives the contribution GV (V) to the
tially real (m(u) z mN), the viscous dissipation is negligible and viscous dissipated energy:
the material can be considered as an elastic solid.
ð
(iii) Vs < x < L(t): in this region, the complex modulus is ð11Þ V umax  ðCÞ 2 du L2 mN C 4
essentially imaginary (m(u) z iumNs) and the material can be GV ðV Þ  s x (10)
2mN s umin 11 u3 Vs
viewed as a Newtonian liquid of viscosity h ¼ mNs. Only this
region contributes to the bulk viscoelastic dissipation processes. where, as discussed in Appendix 1, C z 1and the scaling of the J
From this model several interesting features appear: integral is J x mNa2 (eqn (2)). We can thus rewrite eqn (10) as:
(i) The characteristic value G0 of the fracture energy is gov- ð11Þ
GV ðV Þ L L L t
erned by the small nonlinear zone of size l near the crack tip and  2 ¼ (11)
J a V s Rtip s
equals the Dupr
e work G0 ¼ 2gs needed to cut the material in air,
gs being the surface tension of the fluid. where Rtip  J/mN x a2 is the radius of curvature of the crack tip
(ii) If the fracture length is small (l < L(t) < Vs), according to and t* ¼ L/V is the characteristic time of propagation of a crack
eqn (8) viscous dissipation does not occur and G(V) ¼ G0. This of length L at the velocity V. Since in our experiments Rtip is

This journal is ª The Royal Society of Chemistry 2011 Soft Matter, 2011, 7, 9474–9483 | 9479
 View Article Online

comparable with L and the fracture propagation time t* is much

shorter than the relaxation time s, the energy dissipation origi-
nating from this term can be considered as negligible in front of
the surface energy term J ¼ 2gs.
The r
1/2 component s(C) (12)
12 gives the contribution GV (V) to the

viscous dissipated energy:

ð umax  2 du
ð12Þ V ðCÞ t
GV ðV Þ  s12 xJ (12)
2mN s umin u3 s

We recover the prediction of the trumpet model for the linear

case where J ¼ G0.
The r
1 component s(C) (22)
22 gives the contribution GV (V) to the

viscous dissipated energy:

ð
ð22Þ V umax  ðCÞ 2 du mN a4 L
Published on 30 August 2011. Downloaded on 9/24/2018 8:45:22 AM.

GV ðV Þ  s x ln (13)
2mN s umin 22 u3 s l

By using (31) again, we can derive the scaling law from (13):
ð22Þ 
GV ðV Þ J L Rtip L Rtip t L Fig. 8 A cartoon of the viscous relaxation mechanism of a bead at the
 ln  ln  ln (14)
J mN V s l Vs l L s l tip of the fracture. The polymer bridge between beads (1) and (2) just
debonded, forming a loop on bead (2). Bead (1) thus experiences the
Once again this dissipated energy is negligible in our experiments
spring-back force f(t) due to the gel under tension; this will lead to an
since Rtip is comparable with L and the fracture propagation time increasing extra tension on bead (3) and crack propagation at velocity V.
t* is much shorter than the relaxation time s. Note however the
low logarithmic dependence of G22 V (V) with L.

We remark that the necessary conditions for neglecting bulk

viscous dissipations in the fracture energy of a Maxwell fluid are Just before the bead at the tip of the fracture is released (at
more restrictive in the hyperelastic regime than in the simple time t ¼ 0) by the debonding of a polymer bridge, it is submitted
elastic linear case. Eqns (11), (12) and (13) show that the main to the normal force:
condition for neglecting bulk viscous dissipation is that the si d02
characteristic time t* of the opening of crack of length L should f0 ¼ (15)
li
much smaller that the relaxation time of the Maxwell fluid s, as it
is in the simple elastic linear regime. However, a new length scale where li ¼ di/d0 is the elongation of the network at the crack tip,
Rtip  J/mN  a2 – the radius of curvature of the crack tip, or d0 is the mean distance between droplets at rest (d0 x (4pf/3)1/
equivalently the size of the large strain region – becomes relevant.
3
b), hW is the viscosity of the water and si is the local normal
The dissipated energy can sensibly be neglected only when the stress at the tip of the fracture. Note that eqn (15) assumes the
length of the crack is of the same order of magnitude than Rtip as incompressibility of the gel. For t > 0, the bead experiences the
in our experiments. For very short cracks (L < Rtip) eqn (14) following springback force due to the gel under tension (Fig.8):
predicts the term G(22)V to become important, while for longer sðtÞd02
cracks (Rtip < L < Vs) eqn (11) predicts the term G(11) to become f ðtÞ ¼ (16)
V
lðtÞ
important.
From the relationship between s(t) and l(t) ¼ d(t)/d0 given by the
affine network model of the unentangled rubber elasticity
5 Microscopic model for the fracture velocity theory,32 one gets:
 
In the following we will only focus on the first regime, where both Ed 2 1
f ðtÞ ¼ 0 lðtÞ
2 (17)
the fracture energy G and the crack velocity V are constant in 3 l ðtÞ
time (L ( 0.1D0) and also take approximately the same value
V mm s
1 in the two fluids R6 and R12 (Fig. 6). We remark that The size of the bead is submicrometric, and so its motion in water
pffiffiffiffiffiffiffiffiffiffiffiffi obeys the Stokes law:
V  cR, cR z mN =r (of the order of 1 m s
1) being the speed of
sound in the medium. What is the mechanism that explains this dl
6phW bd0 ¼ f ðtÞ (18)
low speed propagation? dt
Basically, we argue that the scaling of the velocity V of the where hW is the viscosity of water. Substituting (17) in (18) and
fracture is given by the characteristic speed of relaxation of integrating the obtained differential equation with the initial
a microemulsion droplet at the opening crack interface, under condition given by (15) gives:
the action of the unbalanced elastic force of the polymer bridge
towards the gel (Fig. 8). This velocity is given by balancing the l(t) ¼ [1 + (l3i
1)exp(
t/s1)]1/3 (19)
elastic force and the viscous drag force acting on the droplet
(Reynolds number Re  1). with a characteristic time

9480 | Soft Matter, 2011, 7, 9474–9483 This journal is ª The Royal Society of Chemistry 2011
 View Article Online

6phW R gpol x (ln2/2)kBTN2/3 P , where NP z 3fr/8pb is the number

3
s1 ¼ (20)
Ed0 density of polymer chains in the sample (cf. Fig. 1). This inter-
facial tension is extremely low and roughly equal to 10 mN m
1
To obtain the characteristic fracture velocity, one can estimate
and is thus negligible compared to the air/gel interfacial tension
that the next bead will debond (causing the growth of the fracture
gs x 45 mN m
1. It has been proved that for many systems G0
of length d0/li) when l ¼ fdli, where l
1 i # fd # 1 is a critical
exhibits a marked dependence on V, and most of the rate
elongation felt by the chain on the next bonded bead, that is
dependence of G(V) then originates from the rate dependence of
difficult to estimate. Indeed, the relaxation of the debonded bead
G0 itself.34 Raphael and de Gennes35 have shown theoretically
induces extra elastic tensions on the next bead at the tip of the
that the surface energy required to debond the connectors
fracture that will grow with time. The time necessary to reach this
between two surfaces is indeed velocity dependent. However, in
critical elongation is then according to eqn (19):
our fluid this argument would apply to the component gpol which
" #
l3i is negligible in front of the dominant term gs, thus resulting in
tp ¼ s1 ln (21) a substantial independence of G0 from V.
fd l3i
1
Although the condition L < Vs is respected throughout the
The time scale is thus given by eqn (21) with logarithmic experiment, the trumpet model can only accurately describe the
Published on 30 August 2011. Downloaded on 9/24/2018 8:45:22 AM.

corrections. Finally the fracture velocity is: first observed regime of crack propagation where L < 0.1D0. This
"  3 #
1 is not surprising, since the modeling is relative to the fracture
d0 1 Ed02 pffiffiffiffi li
1 propagation in a semi-infinite medium. The second regime for
V ¼ pffiffiffiffi ¼ li ln (22)
li tp 6phw b fd l3i
1 L > 0.1D0 must clearly be attributed to the finite size of the soft
pffiffiffiffi  3 
1 filament and to the increasing value of the strain in the
l
1 progressively thinning ligament (cf. Fig. 5) leading to the
The function g: li / li ln 3 i 3 exhibits a maximum
f d li
1 progressive failure of the rough 2D approximation. The increase
whose position and value depend on fd and determines an upper of G for long cracks could partially be caused by the hyperelastic
bound for the fracture velocity. We choose fd ¼ 0.5. Indeed 2 < dissipation term G(11)
V predicted by eqn (11), but a more accurate
li < 8 (li ¼ 2corresponds roughly to the maximal macroscopic mechanical modelling would be needed to explore this
elongation for the network experimentally observed before possibility.
rupture occurs in pendant drop experiments and li y 833 For both R6 and R12 fluids, the value of the crack velocity
corresponds to the maximal elongation usually observed in fluid (V ¼ 4mm s
1, Fig. 6) is in good agreement with inequality
permanent rubbers), one gets max(g) ¼ a ¼ 0.25. We note that (23) that gives V 6 mm s
1 for a reasonable expected value a 
max(g) is a decreasing function of fd. So the choice fd ¼ 0.5 will 0.25. This strongly supports our hypothesis that the kinetics of
give the upper limit for the estimation of the crack velocity: the crack propagation are governed by the elastic relaxation of
the oil droplets after a debonding event, under the action of the
Ed02
V ( 0:25 (23) viscous drag of the solvent. Note that the control of the crack
6phW b
velocity by network/solvent friction has been already proposed
6 Discussion of the experimental results by Baumberger et al.4,36 in the viscoplastic fracture dynamics of
an other class of reversible gels, i.e. gelatin.
For the fluids R6 and R12, the relaxation time s is, respectively,
0.6 and 2 s. The upper bound value of the crack length is the
initial diameter D0 ¼ 0.6 mm of the filament when the rupture 7 Conclusions
process begins, that is L(t) < 0.6 mm. The crack velocity was We performed an original experimental investigation of the
found to be constant at V x 4 mm s
1. For both fluids, the brittle fracture of a viscoelastic Maxwell fluid using a pendant
characteristic time of a fracture event t* < 0.15 s. The condition drop experiment. This configuration allows for an excellent
(t*/s) < 1 is thus fulfilled (note that in fact, as discussed in section reproducibility of the fracture initiation and propagation, as well
3, the experimental analysis we did is valid for L < 0.1D0 so that as a pure elongational stress condition due to the lack of contact
t*/s < 0.025). The radius of curvature of the crack tip is given, with solid interfaces near the fracture region. The fracture of the
respectively, by Rtip ¼ a2/2x174 and 24 mm, which are of the fluid happens in two steps. The fracture initiation step, discussed
same order of magnitude as the observed crack lengths. in a previous work,12 was shown to be governed by the thermally
Therefore the response of the material is completely elastic activated nucleation of a critical crack in the polymer network.
(L < Vs), the size of the sample being too small or, equivalently, The second step, consisting in the rapid propagation of a brittle
the time necessary to fracture the filament is to small (t* < s) to fracture in the fluid, was analysed in detail here and shown to be
see any viscous dissipation on the overall rupture phenomenon. energetically governed by the surface tension of the solvent (oil-
The fracture energy is thus expected to be independent of crack in-water droplet microemulsion).
speed and to equal G0 ¼ 2g ¼ 2(gs + gpol). The first term is the According to the viscoelastic trumpet model of de Gennes, the
classical Dupr
e work needed to form two new air/gel surfaces absence of bulk viscous dissipation was justified by the short
(dry fracture). However, before dry fracture occurs by instability length of the crack in relation to the characteristic length Vs
of the capillary bridge, the wet fracture must have already where the viscous dissipation starts to become effective. More-
occurred and its interfacial cost is 2gpol. As reported in a previous over, an extension of the trumpet model to hyperlasticity has
work,12 this term of purely entropic origin is related to the loss of allowed us to confirm the weakness of the dissipated energy in
conformational entropy of polymer chains at the crack surface terms of a new relevant length scale Rtip  G0/mN, which is found

This journal is ª The Royal Society of Chemistry 2011 Soft Matter, 2011, 7, 9474–9483 | 9481
 View Article Online

to be of the same order of magnitude as the observed crack where p is the crack tip pressure field:
lengths. In agreement with this interpretation, the measured 2mb q 2m
crack opening profiles presented a constant parabolic shape, p¼
cos þ 2 rð3
cosqÞ þ oðrÞ (29)
a2 2 a
which can be related by nonlinear fracture mechanics to a frac-
ture energy equalling the Dupr
e energy for the creation of the For the computation of the J integral, we only need to retain
two solvent surfaces G0 ¼ J ¼2 (gs + gpol) z 2gs. The fact that singular terms. Since the pressure field is bounded, the nominal
the contribution of the polymer network to the surface tension of stress reduces to:
the fluid is negligible, was shown to be the cause of the extreme 2 3
b2 2 q b2 q q
6 a
sin
sin cos
brittleness of the fracture propagation. In fact, the different 2 a 2 27
sðPÞ ¼ mF ¼ m64
7 (30)
sources of local dissipation in the small nonlinear zone were a q a q 5
shown to be of the same order as gpol and are thus energetically
pffiffi sin pffiffi cos
2 r 2 2 r 2
negligible during crack propagation.
On the other hand, the slow velocity of crack propagation ð ð 2p
ma2 1 q pma2
(V z 4 mm s
1 cR) was shown to be governed by the time scales J¼ sab nb ua;1 ds ¼ esin2 dq ¼ (31)
Ge 4 0 e 2 4
of the local crack tip debonding processes, and namely by the
Published on 30 August 2011. Downloaded on 9/24/2018 8:45:22 AM.

elastic relaxation time of the oil droplets after a debonding event The equations for the deformed crack profile (24 and 25) can be
under the action of the viscous drag of the solvent. The combined to express the visual parabolic contour line of the
remarkable properties of this original model system were thus deformed crack lips:
shown to be related to the excellent degree of uncoupling between rffiffiffiffiffiffiffiffi

y1
the energetic and kinetic properties of the fracture propagation, y2 ¼ a (32)
C
due to the presence of a dominant solvent energy term in the
surface tension of the fluid. where we defined the dimensionless coefficient C ¼ b2/a. Since C
is found to be close to 1 in the FE simulations especially in large
strain,20 we can directly relate the opening parameter of the
Appendix 1: Large strain crack tip solutions in plane parabola to J by inverting (31):
strain condition sffiffiffiffiffiffiffi
Let x and y be the coordinates of the physical points in the non 4J
a¼ (33)
deformed and deformed frames according to: pm

y ¼ ^y(x) ¼ x + u(x) If we express the solutions for the true Cauchy stress tensor in the
non deformed frame:21
where u(x) is the displacement field (expressed in the non 2 3
a q
deformed frame). The first order terms of the 2D plane strain
6 C2
p ffiffi sin
2 r 27
non-linear elastic solution for the crack tip fields in a Neo- sðCÞ ¼ m6
4
7
5 (34)
a q a2
Hookean soft solid were provided by Stephenson:21
pffiffi sin
2 r 2 4r
b2 q
y1 ¼
rsin2 (24) we remark the different scaling of the components of the Cauchy
a 2
and Piola stress tensors. In particular the s(C)11 term is bounded,

pffiffi q while the s(C)

22 presents a 1/r singularity. For the sake of estimating
y2 ¼ a r sin (25) physical scaling laws for the energy dissipation, the true Cauchy
2
stress is certainly the most relevant quantity.
where y¼(y1,y2) and x¼(r,q), while a and b2 are two dimensioned
constants. The J integral is computed in the non deformed frame
according to:
Ð
J¼ G3[Wn1
sabnbua,1]ds (26) Appendix 2: Viscous dissipation in the nonlinear zone
of the fracture
where ! n ¼ ðn1 ; n2 Þ ¼ ðcosq; sinqÞ is the normal vector to the
The debonding of a polymer bridge under tension leads to
surface G3 (which is a circle of radius 3 around the crack tip, ds ¼
a viscous dissipation due both to the motion of bead 3 in Fig. 8 in
3dq is the length increment, s is the nominal Piola stress and the
the solvent and to the friction of the polymer bridge linking beads
elastic energy density function is:
1 and 3 with the solvent too. Both forms of dissipated energy will
m
W ðIÞ ¼ ðI
3Þ (27) be of the same order of magnitude, the polymer contribution
2 being even smaller since the hydrodynamic radius of the polymer
I ¼ tr(G)is the invariant of the left Cauchy–Green strain tensor chain is RH  4 nm < b.
Gik ¼ FFT, where Fij ¼ yi,j is the deformation gradient. We focus here to the viscous dissipation due to the motion of
According to Stephenson,21 the nominal Piola stress a bead 3 in font of the fracture tip that will dissipate energy. The
(expressed in the non deformed frame) can be derived from: corresponding energy cost has been already calculated in section 5:
Ð1
s(P) ¼ 2W0 (I)F
pF
T (28) Ed x li f(t)d0dl (35)

9482 | Soft Matter, 2011, 7, 9474–9483 This journal is ª The Royal Society of Chemistry 2011
 View Article Online

where f(t) is given by eqn (17). This leads to a dissipative 12 H. Tabuteau, S. Mora, G. Porte, M. Abkarian and C. Ligoure, Phys.
contribution to the fracture energy DG0: Rev. Lett., 2009, 102, 155501.
13 S. Mora, Soft Matter, 2011, 7, 4908.

2=3 Ed03 1 r2=3 bE 14 T. Tixier, H. Tabuteau, A. Carriere, L. Ramos and C. Ligoure, Soft
DG0 ðV Þz Np =2 ¼ ðp=12Þ1=3 1=3 (36) Matter, 2010, 6, 2699.
2 3 f 15 J.-F. Berret and Y. Serero, Phys. Rev. Lett., 2001, 87, 048303.
16 P. Skrzeszewska, J. Sprakel, F. de Wolf, R. Fokkink, M. A. Cohen
where we have assumed that roughly half of the polymer chains Stuart and J. van der Gucht, Macromolecules, 2010, 43, 3542.
are in bridge configuration37 and that li ¼ 2. For the fluid R12 17 F. Saulnier, T. Ondarçuhu, A. Aradian and R.E., Macromolecules,
(f ¼ 10%, R ¼ 12), DG0(V)  35 mJ m
2 2gs, which is once 2004, 37, 1067–1075.
again of the same order as gpol and is thus negligible in front of 18 E. Michel, M. Filali, R. Aznar, G. Porte and J. Appell, Langmuir,
2000, 16, 8702.
the dominant term 2gs. 19 J. Sprakel, N. A. M. Besseling, F. A. M. Leermakers and M. A. Cohen
Stuart, Phys. Rev. Lett., 2007, 99, 104504.
20 V. R. Krishnan, C. Y. Hui and R. Long, Langmuir, 2008, 24, 14245.
Acknowledgements 21 R. A. Stephenson, J. Elasticity, 1992, 12, 65–99.
22 L. B. Freund, Dynamic Fracture Mechanics, Cambridge University
We thank G. Porte and C. Fretigny for stimulating discussions. Press, Cambridge, 1990.
This work has been supported by the ANR (contract No. ANR- 23 M. Filali, R. Aznard, M. Svensson, G. Porte and J. Appell, J. Phys.
Published on 30 August 2011. Downloaded on 9/24/2018 8:45:22 AM.

06-BLAN-0097(TSANET)). Chem. B, 1999, 103, 7293.

24 Y. M. Stokes, E. O. Tuck and L. W. Schwartz, Q. J. Mech. Appl.
Math., 2000, 53, 565.
References 25 L. B. Smolka, A. Belmonte, D. Henderson and T. P. Witelski, Eur. J.
Appl. Math., 2004, 15, 679.
1 A. A. Griffith, Phil. Trans. Roy. Soc. London, 1921, A221, 163–198. 26 G. R. Irwin, J. Appl. Mech., 1957, 24, 361–364.
2 A. N. Gent and J. Shultz, J. Adhes., 1972, 3, 281. 27 C. Y. Hui, A. Jagota, S. J. Bennison and J. D. Londono, Proc. R. Soc.
3 E. H. Andrews and A. Kinloch, J. Proc. R. Soc. (London), 1973, 401, London, Ser. A, 2003, 459, 1489–1516.
A332–385. 28 J. Rice, J. Appl. Mech., 1968, 35, 379–386.
4 T. Baumberger, C. Caroli and D. Martina, Nat. Mater., 2006, 5, 552. 29 P.-G. de Gennes, CR Acad. Sci. Paris Srie II, 1988, 307, 1949.
5 M. E. Seitz, D. Martina, T. Baumberger, V. R. Krishnan, C. Hui and 30 P.-G. de Gennes, Langmuir, 1996, 12, 4497–4500.
K. Shull, Soft Matter, 2009, 5, 447. 31 M. L. Williams, R. F. Landel and J. D. Ferry, J. Am. Chem. Soc.,
6 P. Cordier, F. Tournhilac, C. Souli
e-Zakovich and L. Leibler, Nature, 1995, 77, 3701.
2008, 451, 0669. 32 M. Rubinstein and R. H. Colby, Polymer Physics, Oxford University
7 C. L. Mowery, A. J. Crosby, D. Ahn and K. R. Shull, Langmuir, 1997, Press, New York, 2003.
13, 6101. 33 L. R. G. Treloar, The physics of rubber elasticity, 3rd edition,
8 A. Tripathi, K. C. Tam and G. H. McInley, Macromolecules, 2006, 39, Clarendon Press, Oxford, 1975.
1981. 34 A. N. Gent, Langmuir, 1996, 12, 4492.
9 A. Bhardwaj, E. Miller and J. P. Rothstein, Rheol. Acta, 2007, 51, 693. 35 E. Rapha€el and P. de Gennes, J. Phys. Chem., 1992, 96, 4002.
10 A. Bhardwaj, D. Richter, M. Chellamuthu and J. P. Rothstein, J. 36 T. Baumberger, C. Caroli and D. Martina, Eur. Phys. J. E, 2006, 21,
Rheology, 2007, 46, 86. 81.
11 H. Tabuteau, S. Mora, L. Ramos, G. Porte and C. Ligoure, Prog. 37 V. Testard, J. Oberdisse and C. Ligoure, Macromolecules, 2008, 41,
Theor. Phys., Suppl., 2008, 175, 47. 7219.

This journal is ª The Royal Society of Chemistry 2011 Soft Matter, 2011, 7, 9474–9483 | 9483