Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Home Search Collections Journals About Contact us My IOPscience

Adhesion of nickel–titanium shape memory alloy wires to thermoplastic materials: theory and

experiments

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2012 Smart Mater. Struct. 21 035022

(http://iopscience.iop.org/0964-1726/21/3/035022)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 128.210.126.199
This content was downloaded on 27/04/2016 at 19:46

Please note that terms and conditions apply.


IOP PUBLISHING SMART MATERIALS AND STRUCTURES
Smart Mater. Struct. 21 (2012) 035022 (17pp) doi:10.1088/0964-1726/21/3/035022

Adhesion of nickel–titanium shape


memory alloy wires to thermoplastic
materials: theory and experiments
F C Antico1 , P D Zavattieri1 , L G Hector Jr2 , A Mance2 , W R Rodgers2
and D A Okonski2
1
School of Civil Engineering, Purdue University, West Lafayette, IN 47907-2051, USA
2
GM Research and Development Center, Warren, MI 48090-9055, USA

E-mail: zavattie@purdue.edu

Received 15 September 2011, in final form 16 January 2012


Published 17 February 2012
Online at stacks.iop.org/SMS/21/035022

Abstract
We present a combined experimental/theoretical study aimed at enhancing adhesion between a
NiTi wire and a thermoplastic polyolefin (TPO) matrix in which it is embedded. NiTi wire
surfaces were subjected to the following surface treatments prior to pull-out tests: (i) treatment
with an acid etch or chemical conversion coating and (ii) application of a surface
microgeometry to enhance mechanical interlocking between the wire and the TPO matrix.
Nanometer to micron-scale NiTi wire surface features were examined with atomic force
microscopy. The extent to which each treatment increased the pull-out force was quantified.
Existing theoretical models of wire pull-out based upon strength of materials and linear elastic
fracture mechanics are reviewed. Results from a finite element model (FEM), wherein the
NiTi/TPO matrix interface is modeled with a cohesive zone model, suggest that the interface
behavior strongly depends on the cohesive energy. The FEM model properly accounts for
energy dissipation at the debonding front and inelastic deformation in a NiTi wire during
pull-out. We demonstrate that residual stresses from the molding process significantly
influence mode mixity at the debonding front.
(Some figures may appear in colour only in the online journal)

1. Introduction such as thermoplastic vulcanizate (TPV) and thermoplastic


olefins (TPO), which require manufacturing processes such
The fabrication and application of shape memory alloy (SMA) as injection molding, compression molding or extrusion
wire-reinforced polymer matrix composites for actuators has techniques. Although the high temperatures and pressure
gained significant importance during the last decade [1–4]. levels required for these processes make embedding
The capabilities of these smart composites, especially impact SMA wires particularly challenging, a recent study has
resistance, high-temperature mechanical properties, shape shown that an efficient procedure to fabricate these
control, fatigue resistance, and vibration and buckling composites is by a combination of extrusion and injection
response, have been demonstrated [5–9]. Considerable effort molding [87]. Figure 1(a) shows composite films made
has been focused on applying knowledge acquired in of a nickel–titanium (NiTi) wire-reinforced thermoplastic
laboratory settings directly to industrial applications [8]. vulcanizate using extrusion molding. A subsequent step where
Shape recovery makes these alloys good candidates for the film is overmolded using standard injection molding yields
actuators in transportation industries since they can be a composite material with the NiTi wires located at very
activated via applied stress and temperature change and specific locations, typically eccentric from the neutral plane
do not substantially contribute to vehicle mass. Typical of the composite (figure 1(b)). These types of composites
automotive applications employ thermoplastic polymers, are ideal for bending actuation in automotive applications.

0964-1726/12/035022+17$33.00 1 c 2012 IOP Publishing Ltd Printed in the UK & the USA

Smart Mater. Struct. 21 (2012) 035022 F C Antico et al

10 cm 10 cm

(a) (b)

5 cm
(c) (d)
Figure 1. SMA-reinforced thermoplastic composites for actuation devices. (a) Resulting extruded film with embedded NiTi wires using
two different thermoplastic vulcanizates. (b) Composite material made from overmolding the extruded film using a standard injection
molding process. This technique enables better control of the position of the NiTi wires. Bending of the composites is achieved due to shape
recovery of the prestrained NiTi wires located off of the neutral plane. (c) and (d) Cyclic experiments showing bending actuation of
SMA/thermoplastic composites. Shape recovery can be obtained by temperature-induced martensitic-to-austenitic transformation (e.g. via
electrical current). The white background shows the expected deflection (vertical axis) of the composite as a function of the current
(horizontal axis) that flows through the SMA.

Bending can be achieved by inducing shape recovery to with minimal attention to adhesion in actuators. Poor
prestrained NiTi wires that are embedded in the polymeric interfacial bonding due to low surface energies can be
matrix at different locations through the thickness of a thin addressed through mechanical interlocking [2, 11–16, 22–24],
specimen. If the NiTi wire is bonded to the matrix, it chemical treatments [17–19] or by simply neglecting
transmits forces to the matrix inducing a distributed bending adhesion and facilitating the sliding of the SMA material
moment to the entire composite. Figures 1(c) and (d) show and anchoring it at the ends of the specimen (e.g. by
the type of bending actuation that can be obtained from these inserting it in tubes with internal low adhesion/friction
materials under thermal cyclic loading. Non-trivial challenges surfaces) [20, 21]. Adhesion enhancements via mechanical
associated with actuator fabrication processes and a lack interlocking in SMA actuators have been investigated
of optimal component designs are barriers to wide-scale with only the most rudimentary means such as hand
implementation of SMA materials in transportation industries sanding, sandblasting [11] or wire twisting [12, 13]
in general. A significant barrier is load and displacement to alter otherwise smooth surfaces [11–13]. While these
transfer from NiTi shape memory wires to a polymer matrix; approaches help to alter the smooth surface of the SMA and
this is primarily controlled by chemical and mechanical improve mechanical interlocking between the wire and the
bonding at the interface between these two materials. matrix, these methods are either impractical or prone to leave
Nickel–titanium shape memory alloys typically have low impurities and debris. The existing literature does not address
surface energies which limits their ability to effectively bond repeated actuation [7] or fatigue of devices using SMA.
to other materials. To make matters worse, TPV and TPO However, the degradation of polymers through repetitive
also have low surface energies. Poor interfacial adhesion due heating of an SMA has been identified as a critical obstacle
to the lack of sufficient chemical and/or mechanical bonding to increasing the durability of these materials [14, 25].
between NiTi and polymeric materials results in premature Chemically functionalizing NiTi wire surfaces to enhance
failure and insufficient actuation response. adhesion to polymeric materials is attractive from the
Much of the extant literature on shape memory standpoint that additional mechanical fixtures can be avoided.
materials focuses on quantifying material properties [10] This idea, which is the basis of aluminum sheet conversion

2
Smart Mater. Struct. 21 (2012) 035022 F C Antico et al

coating treatments with chemical coupling agents, was return to its original shape after it is heated to 90 ◦ C,
applied by Smith et al [19] to improve adhesion between the transformation temperature required for shape recovery
NiTi wires and plexiglass. They noted improvements in (i.e. the martensite-to-austenite phase transformation). The
adhesive strength of up to 100% with silane coupling agents. Young’s modulus of the martensitic (cold) phase is 30 GPa,
While this is encouraging, silane coupling agents must be and in the austenitic (hot) phase it is 80 GPa. Additional
hydrolyzed before use, and the hydrolyzed solutions have information on the physical properties and deformation
finite lifetimes. Chemical coupling agents function in several response of NiTi shape memory alloys is in [31, 32].
ways. They can directly bond to a polymer matrix by Currently, Flexinol R
wires are manufactured specifically
reacting with functional groups in the polymeric matrix; they for use in actuators for high cycle applications. All wires
can incorporate themselves into the polymer matrix due to were ‘trained’ through a thermo-mechanical cyclic process
the similarity of their structure to that of the polymer; or by the manufacturer prior to subsequent surface treatment.
they can interact through van der Waals bonding with the See [32–35] for representative applications and for more
polymeric matrix. Oxide surface functionalization using a details on the thermo-mechanical behavior of Flexinol R

coupling agent typically requires hydroxyl groups (−OH) wires. Typically these wires exhibit a dark surface due to
on the NiTi surface. Phosphorus-containing compounds [26] their natural oxide layer which is (approximately) 400 nm
and organosilane-coupling agents react with oxide/hydroxide thick [36]. Whether the NiTi wire surface oxide was grown
groups and bind to metal oxide surfaces via a condensation prior to wire drawing or after drawing is not known for certain
reaction. A heating step may be employed after treatment with without AFM examination.
the coupling agent in order to ensure complete condensation.
In this paper, we explore two experimental methodologies 2.1. NiTi wire preparation
for improving adhesion between a NiTi SMA in wire form
and a polymeric matrix. These are: (i) functionalizing a NiTi Surface contaminants were removed from the wires through
wire surface with a chemical conversion coating or coupling sequential treatment with acetone, 2-propanol and xy-
agent or (ii) application of a surface microgeometry to an lene [17]. The wires were then air dried and briefly soaked
NiTi wire surface for enhancement of mechanical interlocking in deionized water. Each cleaned NiTi wire was immersed in
between the wire and the polymer matrix. Although prior a 3% hydrofluoric acid (HF)/15% nitric acid (HNO3 ) solution
work has focused on enhancing NiTi wire adhesion in carbon until the surface oxide was removed (a 20 min process). The
fiber-reinforced polymers with acid etching [17] and silane HF ensured dissolution of the otherwise insoluble titanium
coupling agents [18, 19], it appears that organophosphorus oxide on the NiTi. Drying was avoided since it allows
compounds have yet to be extensively explored in cases formation of a strong oxide film that is difficult to convert
where an inexpensive polymeric matrix, such as thermoplastic into a surface rich in hydroxyl groups. The wires were
polyolefin (TPO), is required. There is indeed evidence from instead soaked in deionized water after etching and after
both the theoretical [27] and experimental [28, 29] literature each subsequent step, and then immersed in 1 M sodium
that organophosphorus compounds will react with surface hydroxide (NaOH) for 20 min to force the oxide groups to
oxides and hydroxides on NiTi alloys. However, the extent reform as hydroxides. Some (or most) of these hydroxides
to which these compounds improve NiTi wire adhesion with were present as surface-O–Na species (i.e. when exposed to a
a polymeric matrix has not been investigated. Atomic force highly alkaline solution). Hence, a 5 min soak in concentrated
microscopy (AFM) and wire pull-out tests in conjunction with H2 SO4 was used to remove the sodium ions and to leave a
numerical models are used to examine and quantify NiTi surface rich in −OH groups available for reaction with the
wire/TPO matrix interactions. coupling agent. Ultrasonication followed each step.
The remainder of this paper is organized as follows. Once the new oxide/hydroxide layer formed, each
The NiTi wire surface treatments are detailed in section 2 NiTi wire was soaked in an aqueous solution of 1%
followed by AFM measurements in section 3. Mechanical phenylphosphonic acid (PPPA, R–P(=O)(OH)2 , R=C6 H5 )
pull-out tests of wire–polymer adhesion due to the surface at 298 K for 20 min. Another set was subject to a 20 min
treatments, and the theoretical and computational models soak, with ultrasonication in 5% PPPA. Additionally, hand
employed to interpret the results are described in sections 4 sanding (180 grit sand paper) of the wires was used to
and 5, respectively. The paper concludes with a summary apply a directional microgeometry. The NiTi wires that were
of the major observations from the experiments and FEM embedded into the polymeric matrix had diameters ranging
simulations. from 380 to 750 µm and embedded lengths, le , ranging from
70 to 90 mm.
2. Surface treatments
2.2. Phenylphosphonic acid (PPPA) coupling agent
Nickel–titanium (NiTi) shape memory alloy wires, known
by the Flexinol
R
trademark, and manufactured by Dynalloy, The phosphorus atom (P) in PPPA contains a double bond
Inc. [30] were used in this study. These wires exhibit to an oxygen atom (=O) and the P is also bonded to two
a Young’s modulus increase of a factor of 2.6 and a hydroxyl (OH) groups that possess acidic hydrogen atoms.
dimensional change of up to 4–5% (depending on the Although we chose R=C6 H5 , in principle any organic group
amount of prestrain) when heated. If deformed, a wire can that is compatible with −OH groups is possible, and it will

3
Smart Mater. Struct. 21 (2012) 035022 F C Antico et al

those in σ bonds (for example). However, an ample supply


of surface −OH groups must be available (see figure 2(c))
to achieve effective coverage with the coupling agent. The
completed reaction, which shows the coupling agent at
the interface between the polymer and NiTi material, is
schematically detailed in figure 2(d). The solvents used
were histological grade acetone, reagent grade 2-propanol
(a) (b) and reagent grade xylene (mixed isomers). Reagent grade
sodium hydroxide (Baker), nitric acid, sulfuric acid (VWR)
and 98+% PPPA were also used, as was deionized water.

2.3. Wire embedding process

Following surface treatment, the NiTi wires were embedded


into TPO by injection molding. Following this procedure,
the wires were exposed to the same manufacturing process
(c)
as the composites shown in figure 1. The mold was used
to produce dogbone-shaped specimens. Each dogbone was
4 mm thick, 167 mm long (excluding the crimped wire ends)
and 10 mm wide at its center (20 mm wide at its ends).
Two dogbone specimens per molding cycle were produced,
each with a straight gauge section. Wire expulsion from the
polymer matrix had to be prevented to allow measurement
of the interfacial strength of the polymer/NiTi interface [11]
in the mechanical pull-out tests. Additional wire prestrain
was deemed unnecessary. A two-step embedding process was
subsequently developed. In the first step, the end of the wire
(d) that protruded from the specimen was crimped. The crimp was
attached to the mold. A small weight was affixed to the other
Figure 2. (a) Hydrogen bond between the oxide–hydroxides on the end of the wire which was to be embedded in the polymer.
NiTi surface and the phosphonic acid. (b) Condensation (produces
water) which results in bonded P(=O)R groups. (c) Surface −OH The weight was intended to minimize wire bending or kinking
groups on NiTi. (d) Polymer and NiTi substrate interface with during injection molding and mitigate the fluid drag force
coupling agent (P–O bonds to NiTi surface not shown). In the to keep the wire straight. In addition, the weight geometry
present study, R=C6 H5 . helped to keep the wire from moving to the surface of the
solidified polymer, irrespective of the wire surface treatment.
Nevertheless, the effectiveness of the weight was found to be
determine the extent of bonding with a polymer. The hydroxyl largely dependent upon the wire surface treatment. After the
units (i.e. the OH groups) will initially be attracted to polar embedding process, the portion of the specimen containing
oxide and hydroxide groups on oxide surfaces. Interaction the weight was carefully cut without affecting the interface
between the metal oxide surface and the PPPA will first occur between the wire and the polymer.
through hydrogen bonding. Although moderately strong,
these bonds are not as strong as the bonds formed through 3. Surface characterization
a subsequent heat treatment. The P end is bound to the NiTi
oxide–hydroxide via a condensation reaction as illustrated in 3.1. Optical observations
figure 2(a). The resulting P–O bonds are shown schematically
in figure 2(b). The process of applying the coupling agent Figure 3 shows three molded specimens. Figure 3(a) shows
to an as-received NiTi wire surface required removal of the a NiTi wire with no surface treatment. Roughly 50% of the
native oxide/hydroxide layer (via acid etching) followed by wire penetrated the specimen surface (dark region along the
controlled re-growth of the surface oxide. This generated an center of the gauge section moving from left to right across
ample number of surface −OH groups for bonding. Once this figure 3(a)). This shows that the wire moved during molding,
was completed, the re-grown NiTi oxide was treated with the even with the weight, and suggests that wire/TPO polymer
coupling agent. adhesion was insufficient. It should be mentioned that
The TPO polymeric matrix is a saturated hydrocarbon one-half of the specimens with no surface treatment exhibited
material that offers little opportunity for bonding or polar this type of behavior. Only those specimens with uniform
interactions. Polarizable substances, such as aromatic phenyl embedding were considered for the subsequent pull-out tests.
rings, will interact with TPO. This occurs due to the The hand-sanded SMA wire, shown in figure 3(b), shows
high polarizability of the electrons in delocalized p-orbitals. improved embedding. However, some of the wire is still
Electrons in these orbitals are more easily polarized than faintly visible on the right-half of the specimen in figure 3(b).

4
Smart Mater. Struct. 21 (2012) 035022 F C Antico et al

Untreated
3.2.1. As-received NiTi wire surface topography. Figure 4(a)
shows an oblique view of an as-received NiTi wire surface.
(a) Wire through
Areas of bright contrast represent roughness values that
the polymer are high relative to the deepest recessions in the surface
Hand-
(denoted by regions of dark contrast). Height values are
sanded reported in nm in the key, with the total roughness height
(b) Embedded range being 0–451.2 nm. Die marks are indicated with the
NiTi wire
dotted arrows. Note the die marks on the planar portion of
5% PPPA the surface follow the white arrow in the upper right of
figure 4(a): this denotes the drawing direction. This surface
(c)
also contains a series of randomly positioned cracks. The
orientations of crack openings are mostly transverse to the
drawing direction, suggesting that the cracks originated from
one of two possible mechanisms during drawing. In the first
mechanism, crack nucleation resulted from local adhesion of
the wire surface oxide (either natural oxide or re-grown oxide)
to the drawing die surface due to local lubricant starvation
Figure 3. (a) Untreated NiTi wire which is not properly embedded (leading to a stick–crack–slip phenomenon). Alternatively, a
in the TPO matrix—a portion of the wire surface penetrated one condition of mixed-film lubrication may have existed at the
surface of the specimen; (b) a hand-sanded NiTi wire is still not die–wire interface wherein load transfer occurred through
properly embedded in the TPO matrix and (c) 5% PPPA treatment locally entrapped lubricant films. The thickness of these
facilitates proper positioning of the wire within the TPO matrix,
which is indicative of improved wettability with the TPO.
films is impacted by the drawing speed. Those areas of
the wire surface beneath the films were in unconstrained
deformation which creates conditions conducive to surface
On the other hand, the NiTi wire treated with 5% PPPA crack nucleation and growth. This process is very similar
(figure 3(c)) displays the most uniform embedding within to that which generates transverse micro-cracks on rolled
the TPO matrix, which suggests greatly improved adhesion aluminum sheet, for example [37]. However, the surface
between the wire and TPO3 . Additionally, none of these damage shown in figure 4(a) may have resulted from the
specimens showed wire penetration to the surface. These wire training process (as-received wires are typically black;
observations were consistent with additional molding tests in manufacturer specifications are such that the oxide thickness
which the same treatment procedures were repeated. is in excess of 400 nm [34]). What this fails to answer is
whether or not the existing oxide was grown prior to drawing
3.2. AFM analysis or after drawing. This issue is resolved by the longitudinal die
marks in figure 4(a) due to sliding of the wire surface over
The ultrasharp AFM probe, model TESP-HAR from Veeco the die surface. Note that these marks could not have resulted
Instruments, consisted of a silicon cantilever doped with Sb. from any amount of training subsequent to drawing, and this
The tip radius is nominally less than 10 nm and the tip height suggests that the as-received wire surface oxide was grown
is 10–15 µm, with a front angle of 5◦ and a side angle of 5◦ . prior to drawing. The die marks are indicative of the finish on
The cantilever thickness, length and width are 4 µm, 125 µm the die used to manufacture the wires. A cursory examination
and 30 µm, respectively. The cantilever stiffness is 42 N m−1 . of the die marks in figure 4(a) reveals peaks in the 10–30 nm
All wire surfaces were scanned with a Digital Instruments range. This suggests that the die surfaces were reasonably well
D3100 AFM in tapping mode. Typical scanning speeds ranged polished (the wavelength of green light is ∼550 nm), and such
from 0.5 kHz (10 µm s−1 ) to 1.5 kHz (30 µm s−1 ). For a smooth finish may have contributed to the aforementioned
the majority of images gathered from the present work, a lubricant starvation mechanism during wire drawing. Section
0.7 kHz scanning speed was used. All scans were gathered profiles along a line normal to the drawing direction revealed
over a 10 µm × 10 µm area with 512 × 512 pixels2 . Scans a nominal peak-to-valley excursion (representing the distance
over larger areas provided no useful information beyond that from the mean surface plane to the bottom of the measured
reported herein. Great care had to be exercised in establishing cracks) of 282.6 nm. The nominal slope of the crack walls was
AFM scan parameters so as to avoid a myriad of scan artifacts 58.3◦ . The highest regions of the surface are plateau-shaped,
that precluded acquisition of quantitative surface topography indicating that the surface has a good bearing area [38].
data. Several wires with as-received and hand-sanded (180 grit It is unlikely that the majority of the surface cracks fully
sandpaper was used) surfaces, along with some treated with penetrated the oxide layer.
the PPPA coupling agent, were scanned with the AFM.
3.2.2. Hand-sanded NiTi wire surface topography. Fig-
3 ‘Uniformly embedded’ implies that the wire remains straight inside the
ure 4(b) is a representative AFM surface plot of an NiTi
polymer with a fixed distance to the TPO surface along its length. By use of
the term ‘not properly embedded’, it is meant that the wire is curved and its
wire surface topography from hand sanding. The surface
distance to one of the surfaces, along the thickness, of the specimen varies is irregular with jagged peaks corresponding to displaced
such that, in most cases, part of the wire was visibly exposed. material and embedded debris that are good sites for

5
Smart Mater. Struct. 21 (2012) 035022 F C Antico et al

(a) (b)

(c) (d)
Figure 4. (a) AFM height Image of an as-received NiTi wire surface across a 10 µm × 10 µm surface area showing die marks along
plateau regions with cracks. Surface cracks suggest an oxide layer prior to wire drawing. (b) AFM image height of hand-sanded NiTi wire
surface. (c) AFM height images of NiTi surface with 5% PPPA. (d) Phase contrast image corresponding to height image (c).

mechanical interlocking. The bottom part of the image shows these features are due to locally excessive sticking between the
a substantially unidirectional topography and is indicative oscillating AFM probe and the surface. Hence, the chemistry
of the sanding direction. The irregular array of peaks and of the treated surface differs from that of the as-received and
valleys resulting from the grit particles effectively extended hand-sanded surfaces since the probe is momentarily adhered
the surface area which is beneficial for mechanical adhesion to the sample surface while the body of the cantilever deflects,
with a polymeric matrix during the embedding process. thereby jerking the probe away from the sample surface and
The hand-sanded wires are brighter in appearance than the instantaneously altering its oscillation amplitude [39]. This
as-received wire surfaces due to removal of the oxide layer. behavior was noted in all initial scans of the chemically
In addition to data of the type shown in figure 4(b), we also treated wire surfaces for which the initial scan parameters
scanned portions of the hand-sanded surfaces that contained (established for the as-received and hand-sanded wires) were
isolated debris particles. used. Adjustment of the cantilever oscillation amplitude, so
as to apply greater force to the wire surface with each tap,
3.2.3. PPPA treatment. Measurement of NiTi wire surfaces significantly diminished scan artifacts.
with the 5% PPPA coupling treatment initially proceeded with Figure 4(c) is a representative AFM surface plot of an
the AFM scan parameters established for the as-received and NiTi wire surface following treatment with 5% PPPA. Notable
hand-sanded NiTi wires. Topography slope images revealed aspects of this image are its substantially different appearance
high frequency ringing that snakes along valley regions of from the as-received and hand-sanded height images, e.g. its
the treated wire surface. Previous experience with other 2.10 µm maximum roughness height (approx. the extreme
chemically treated metal surfaces led to the conclusion that roughness heights of 451.2 nm and 560.2 nm in figures 4(a)

6
Smart Mater. Struct. 21 (2012) 035022 F C Antico et al

and (b), respectively). This resulted from the acid etch


treatment prior to application of the 5% PPPA coupling agent.
Figure 4(d) is a phase contrast image corresponding to the
height and slope images in figure 4(c). Comparison (not
shown) of height and slope images with the phase image
show that the latter contains contrast features that are not
readily apparent in the former two images. Selected regions
that show contrast unique to the phase image in figure 4(d)
are labeled A–F. Figure 4(d) suggests that phase contrast
measured from the treated NiTi wire was not largely mitigated
by surface topography, but rather through local variations in
probe surface adhesion. Amplitude distribution plots derived
from AFM height data revealed that the 5% PPPA coupling
treatment gave a surface that has negligible skewness (i.e. it is
primarily a Gaussian distribution). The transition from air to
surface is more gradual in the chemically treated wire surfaces
than in the as-received and hand-sanded wire surfaces. This
is advantageous from the standpoint that the surface slopes
are less extreme and offer less resistance to liquid polymer
Figure 5. Arithmetic mean roughness (Ra ) values from 10
wetting than would be the case with that in figure 4(b). specimens of each surface type.
3.3. Roughness comparison
Ten areas of each surface type (i.e. as-received, hand-sanded the mechanics of debonding is well understood [40–43].
and chemically treated) were separately scanned on selected Tensile pull-out testing of NiTi wires embedded in polymeric
NiTi wires. The data from each was used to compute the or composite matrices has been reported in [11, 12, 17, 19,
arithmetic mean roughness (Ra ). 44–46]. In this study, the wires are tested at room temperature.
Figure 5 shows that as-received NiTi wire surfaces are the It is believed that this type of test represents the worst-case
smoothest, with a mean Ra of 36.41 nm (standard deviation of scenario for the NiTi/TPO interface. Here, the wire thinning
5.12 nm). Wire surfaces treated with 5% PPPA have the largest caused by the stress-induced martensitic detwinning of the
Ra values, with a mean Ra value of 222.10 nm and the widest NiTi contributes to an opening mode in the interface, helping
spread of roughness values (standard deviation of 76.9 nm). prevent any additional energy dissipation mechanisms that
The hand-sanded surface roughness values are intermediate may be triggered by mechanical interlocking due to the
to those of the as-received and chemically treated wire roughness of the interface.
surfaces with a mean value of 66.66 nm (standard deviation
of 35.732 nm). We also computed the root-mean-square 4.1. Test configuration
roughness (Rq ) and the extreme peak-to-valley excursions
(1z). The values of Rq follows the same trend as that The inset in figure 6 is a schematic of the pull-out test
for the Ra with the chemically treated NiTi wire surfaces and a typical force–displacement curve is also shown. This
having the greatest spread of values, and the as-received configuration ensures that the specimen does not experience
wire surface having the smallest spread of values. The lateral compressive stress. A monotonically increasing force,
mean Rq for the NiTi wires treated with 5% PPPA was F, is applied to the crimp at the end of the wire until
287.68 nm (standard deviation of 46.72 nm), while the mean debonding is triggered. This test effectively captures the peak
values for the as-received and hand-sanded NiTi wires were (or maximum force, Fmax ) to pull the wire out of the polymer.
51.26 nm (standard deviation of 6.79 nm) and 92.89 nm Each specimen (i.e. a NiTi wire embedded in TPO) was first
(standard deviation of 47.86 nm), respectively. The mean 1z cut through its gauge section exposing one end of the wire.
for the treated wires is 2.21 µm with a standard deviation The specimens were cut to have the same embedded wire
of 493.94 nm. The mean 1z for the hand-sanded wires is length le = 90 mm. However, a few specimens with le =
considerably smaller at 765.01 nm with a standard deviation 70 mm were also measured. The free end of the NiTi wire was
of 336.16 nm. Close to this value is the mean for the loaded with a monotonically increasing force. The following
as-received wires at 638.99 nm with a standard deviation of final set of TPO specimens with embedded NiTi wires were
153.79 nm. investigated in the pull-out tests.

(1) As-received wires (five specimens with le = 90 mm and


4. Pull-out tests
seven specimens with le = 70 mm).
In tensile pull-out, the pull-out force is defined as the applied (2) Hand-sanded wires (four specimens with le = 70 mm and
force to initiate debonding between an NiTi wire and a TPO four specimens with le = 90 mm ).
matrix. This test methodology has been used to measure (3) Pre-treated wires with acid only (oxide/hydroxide layer,
interfacial adhesion in fibrous composites, for example, and see section 2.1) (five specimens with le = 90 mm).

7
Smart Mater. Struct. 21 (2012) 035022 F C Antico et al

Table 1. Test matrix with mean forces recorded for each treatment for pull-out specimens with le = 90 mm. The untreated-wire specimens
require the lowest pull-out force, while those treated with 5% PPPA require the highest pull-out forces. The maximum value achieved with
the untreated NiTi wires is even lower than the maximum values from tests of the treated wires.
Number of specimens Max. force
Treatment tested (le = 90 mm) Mean force ± stand. dev. (N) Min. force recorded (N) recorded (N)
Untreated 5 16.2 ± 5.1 6.9 22.2
Hand-sanded 4 22.7 ± 2.0 19.2 24.3
Acid 5 21.0 ± 2.8 18.9 26.4
1% PPPA 8 22.0 ± 2.6 18.1 25.7
5% PPPA 8 22.4 ± 3.1 18.0 28.0

debonded from the TPO is referred to as the debonding


process zone (denoted by the light gray region labeled
‘Debonding process zone’ in figure 7(a)). Figure 7(b) depicts
the debonding process zone advancing deeper into the TPO
as the applied pull-out force increases during loading stages
I–II. Meanwhile, large axial deformation in the wire is still
observed near the debonding process zone. The debonding
front finally reaches the specimen end opposite that of the
applied force as shown in figures 6 (inset) and 7(c), with
complete debonding shown in figure 7(d).

4.2. Experimental results

Table 1 reports the pull-out forces (newtons) obtained in the


experiments. Due to the large number of experiments, the
table only shows mean force values associated with each
treatment. The mean maximum pull-out forces and equivalent
Figure 6. Typical load–displacement curve obtained from the
tensile stress (with their standard deviations) were plotted in
experiments. Inset figure: schematic of pull-out specimen figure 8. The red dots denote those pull-out specimens with
containing an embedded NiTi wire. le = 90 mm, whereas the green squares are associated with
the shorter pull-out specimens (le = 70 mm). This data shows
that the maximum mean pull-out force for all of the tests was
(4) Wires with fresh oxide/hydroxide layer followed by within the standard deviation regardless of NiTi wire surface
treatment with 1% PPPA (eight specimens with le = treatment. However, as seen in table 1, the wires treated
90 mm and one specimen with le = 70 mm). with 5% PPPA required the largest pull-out force achieved
(5) Wires acid etched and then treated with 5% PPPA (eight in a single test. In contrast, the untreated wires required the
specimens with le = 90 mm). smallest pull-out forces, although a single specimen with an
untreated wire reached a maximum force close to that of the
Three distinct regions can be observed in the treated wires. The forces required with the treated wires were,
force–displacement curve: an initial portion in which the force in general, higher than those associated with the untreated
increases to point I, a plateau in the force between points II wires. The pull-out results suggest that the various treatments
and III, and a final decaying portion down to zero force. The lead to an improvement in adhesion of the NiTi wires to the
translucency of the TPO allowed observation of debonding of TPO matrix.
the NiTi wire from the polymer matrix. Wire deformation and
debonding started at the wire/matrix interface from the force 5. Theoretical models
application end before the force achieved its maximum value
in figure 6. The axial deformation observed in the wire at force Wire pull-out forces typically need to be interpreted through
levels equal to or greater than Fmax indicates that the wire appropriate models to obtain the effective wire/matrix
was undergoing detwinning. Finally, the wire breaks free of properties. Previous studies have used analytical models based
the TPO matrix (point IV) as the debonding process becomes on strength of materials and linear elastic fracture mechanics
unstable. Thinning of the wire in the transverse direction also (LEFM) to convert pull-out forces to interfacial strength or
contributed to wire debonding from the TPO. adhesion energies [11, 14–16, 47–50]. In section 5.1, we
The progression of the debonding process is schemati- briefly review relevant analytical models and discuss the
cally depicted in figure 7. Each inset in figure 7 shows the extent to which they apply to the pull-out tests in this study.
relative position of the image above it with respect to the Both analytical expressions based on LEFM and strength of
entire NiTi/TPO specimen. The region where a NiTi wire materials are presented in section 5.1. Finally an FEM that

8
Smart Mater. Struct. 21 (2012) 035022 F C Antico et al

(a) (b)

(c) (d)

Figure 7. Schematics of the wire pull-out process at different stages of the pull-out test as described in figure 6. Process of debonding
during the pull-out force: if the wire is in the twinned martensitic phase, then it first undergoes a stress-induced detwinning with strains up
to 4%. The white arrow in the NiTi wire denotes the applied force. Debonding starts in (I). A gap is left in the wake of the debonding front
caused by the thinning of the NiTi wire (b). (c) Debonding propagates toward the other end of the specimen until (d) complete debonding is
attained at (IV). Insets beneath each figure show the location where the debonding process zone takes place relative to the entire specimen.

represents the mechanical response of the pull-out specimens where r is the wire radius and τiy is the shear strength
under quasi-static loading is used to analyze the debonding of the wire/matrix interface or polymeric matrix (whichever
process (section 5.2). is smaller). In brittle interfaces, where plastic deformation
around the interface is limited, Fmax is not proportional to le
5.1. Analytical model
because of the nonlinear shear stress distribution around the
According to Penn and Lee [47], wire debonding can interface: this tends to very large values of τiy near the ends
exhibit two distinct behaviors: (i) ductile behavior, where of the wire/matrix interface. Following Piggott [51], Penn and
the interface has a uniform stress distribution, i.e. a large Lee [47] showed that, if a failure criterion based on the energy
fracture process zone, and yielding of the matrix is the main of the interface is assumed, then the following expression
mechanism of deformation, and (ii) brittle behavior, where
provides a valid estimate of the critical strain energy release
the interface has a small debonding process zone compared
with the dimensions of the specimen (which is described by rate, Gc (J m−2 ), or work of separation, in terms of Fmax :
LEFM), and a non-uniform stress distribution along the NiTi
2
wire/matrix interface. [1 + csch2 (ns)]F max
When ductile interface debonding takes place, Fmax Gc = 3
(2)
4π 2 r Ef
is proportional to the embedded wire length (Fmax ∝ le ).
Accounting for the contact area between the wire and a where, for the present case, Ef is the Young’s modulus of the
polymeric (e.g. TPO) matrix Fmax is defined as [51]: NiTi wire, s = (le − a)/r, a is the length of the debonded
Fmax = 2π rle τiy (1) initial region, le is the embedded wire length when a = 0 and

9
Smart Mater. Struct. 21 (2012) 035022 F C Antico et al

manufacturer’s recommended value to guarantee durability


(∼180 MPa for Flexinol R
wires). For optimal actuation,
the design of the composite requires that the actuation
force, and therefore the tensile stress, be maximized without
exceeding the maximum recommended value. Evidently, this
requirement competes with the necessary strength level of
the wire/polymer interface to preclude debonding. Wires
that can be pulled out at tensile values lower than the
manufacturer level cannot exert the maximum force. On
the other hand, wires that need a higher tensile stress for
pull-out are safer for actuation applications. Figure 8 shows
the maximum values of the tensile stress in the NiTi wires
used in the present study compared with the manufacturer’s
recommended level. Only the chemically treated NiTi wires
exceed the recommended values, suggesting that the load
capacity of actuators with embedded wires can be maximized
Figure 8. Maximum pull-out force recorded for NiTi wires with with a suitable wire surface treatment. It should be mentioned
embedded length le = 70 mm (in green squares) and le = 90 mm (in that the values of pull-out force for the treated wire may
red circles). The mean values for each treatment are denoted by
symbols and the standard deviation is indicated by the error bars. be safe mostly for single actuation applications, i.e. where
The tensile stress is obtained from equation (6) for all the the force is applied once monotonically. We surmise that
experimental points. The horizontal dashed line denotes the the required pull-out forces for typical cyclic actuation
manufacturer’s recommended tensile stress of 180 MPa. applications should be higher than those reported here to
compensate for accumulated damage at the interface between
the NiTi and the polymer.
n is defined as [51]
Em
n2 = (3) 5.2. Finite element model
Ef (1 + νm )ln R

r
where Em are νm are the Young’s modulus and Poisson’s ratio, A two-dimensional axisymmetric FEM was used to analyze
respectively, of the matrix and R is the radius of a cylinder the pull-out tests. The geometry of the computational domain
of polymeric material that surrounds the wire. To estimate a and boundary conditions are shown in figure 9(a). The nodes
lower bound for Gc (see equation (2)), R is approximated as located at x1 = 0 are constrained to move in the x1 direction
the minimum dimension of the cross section of the specimen to satisfy the axisymmetric condition. Nodes along part of the
(in the present case, this is 4 mm, which corresponds to the upper boundary of the polymer (at x2 = 70 mm in figure 9(a))
thickness of the specimen). For very short values of le , ns  le were also constrained to move in the x2 direction to properly
and equation (2) reduces to account for the constraints in figure 6. Given the dimensions
2 of the specimens and pull-out tests, the radius of the TPO
Fmax
Gc = . (4) matrix, R (see figure 9(a)), was set to 2 mm to be consistent
4π 2 r3 Ef sinh2 (ns) with the specimen thickness. Preliminary simulations showed
Alternatively, for large embedded lengths, equation (2) that an increase in R from 2 to 4 mm results in only 3.5%
reduces to variation in the values of Fmax . Hence, the axisymmetric
2 approximation has only a very minor effect on our FEM
Fmax
Gc = . (5) results. Additionally, models with embedded lengths of 70 and
4π 2 r3 Ef 90 mm were used to verify that Fmax is indeed independent of
the embedded length of a NiTi wire [47]. In fact, a difference
This is a useful approximation for rectangular specimen of only 1% in the maximum pull-out force was obtained
geometries with large embedded lengths and a small fracture between the two embedded lengths.
process zone (i.e. the LEFM approximation). However, The experiments revealed large deformations (i.e. up
LEFM may underestimate Gc when material nonlinearities are to 5%) in the wires along the axial or pulling directions.
important (as in the present study). This deformation can be attributed to detwinning of the
Another quantity of interest is the tensile stress in the martensitic phase4 (noted in figure 7) [52]. Numerous
NiTi wire, estimated as continuum models for SMA have been developed that make
Fmax use of macroscale plasticity (often with backstress, flow rules
σmax = . (6) and yield functions) [52, 53]. Here, the following nonlinear
π r2
The tensile stress in the wire becomes important for 4 This is because the wires undergo temperature-induced martensitic-to-
actuation purposes, especially when a high number of cycles austenitic phase transformation during the injection molding process and
is required. Stress levels in the wires should remain below the therefore they recover any prestrain from wire drawing.

10
Smart Mater. Struct. 21 (2012) 035022 F C Antico et al

(a)

(b)

(c)

Figure 9. (a) Schematic of the axisymmetric FEM geometry. The light region denotes the polymer matrix, while the dark region denotes a
NiTi wire. Here, le = 70 mm. r is the radius of the wire, R is the maximum radius of the TPO adopted, w is the radial dimension of the TPO
relative to the outer radius of the wire and a is the crack length. (b) Experimental uniaxial stress–strain curve (dotted line) and exponential
hardening law for NiTi (solid line; equation (7)). (c) Schematics of the initially elastic constitutive cohesive law employed to describe the
interface behavior. The red double arrow indicates any irreversible post-peak unloading–loading cycle [65].

isotropic hardening model [54] was used to represent NiTi stress conditions. Additionally, due to the large deformations
deformation in its martensitic phase at 298 K [55]: expected in a NiTi wire during the pull-out tests, our FEM
pl
also accounts for finite deformations. Figure 9(b) shows the
Y(εpl ) = Y∞ + (Y0 − Y∞ )e−βε . (7) experimental true stress–strain curve of NiTi compared to
The plastic strain is ε pl , Y(εpl ) is the radius of the fit response from the model. In this case, the stress is
the yield surface, Y0 is the initial yield stress, β is σ = Y(εpl ) and the total strain is defined as εtot = ε pl + σ/Ef .
a hardening parameter and Y∞ is used here to fit the The experimental stress–strain curve corresponds to our own
constitutive model in equation (7) with the experimental measurements; however, it follows what others have reported
stress–strain curve of the wire. While this model can for Flexinol wires [57, 58]. The linear response of NiTi was
predict the overall effective response of the NiTi wire characterized assuming Ef = 30 GPa based on the information
under monotonically increasing stress conditions, it may not from the manufacturer. The parameters adopted to match
accurately account for the reorientation of the martensitic the nonlinear response of the wire from our experimental
phase and microscale inhomogeneous deformation. As measurements are Y0 = 0.04 GPa, Y∞ = 0.023 GPa and
discussed by Gao and Brinson [56], caution is required β = −100.
here since the SMA deformation/stress response is based on In our FEM study, only quasi-static monotonic loading
underlying mechanisms that differ from those associated with at constant temperature was modeled. It is assumed that the
a reversible plastic response. To apply this model to predict wire is initially stress-free, in the fully martensitic phase
the behavior of the NiTi wire in our particular case, we note and that detwinning occurs only due to the applied stress
that the NiTi is always subjected to a monotonically increasing during loading. The TPO matrix is modeled as linear elastic,
load during pull-out testing and that most of the inelastic an assumption that is valid since stress values are well
deformation of the debonded wire takes place under uniaxial below the yield strength of TPO. Here Em = 0.736 GPa and

11
Smart Mater. Struct. 21 (2012) 035022 F C Antico et al

νm = 0.33 [59]. Each material is represented by a mesh with conducted a systematic parametric analysis of the variation of
four-noded quadrilateral elements. the main cohesive parameters, namely Tmax , α, δn and δt .
The NiTi wire/TPO interface is modeled with a Previous studies neglected residual stresses from in-
cohesive zone model [60–64] to simulate crack initiation jection molding. Residual stresses may delay the onset
and propagation. A potential-based cohesive zone model of interfacial debonding during wire pull-out due to the
implemented as triangular-shaped, four-node, zero-thickness additional compressive normal stresses on the wire–matrix
interface elements [65] is embedded between the wire and interface. The compressive normal stresses increase friction
the TPO matrix [65–71]. The formulation of these elements forces along the wire/TPO interface [75, 76]. The specimens
is based on a non-dimensional effective displacement jump produced in the present study have low NiTi wire/TPO
(λ) that takes into account the current normal and tangential friction compared with adhesion (even for the hand-sanded
displacement jumps at the NiTi wire/TPO interface estimated specimens). However, in the NiTi/TPO system residual
from the FEM. Hence stresses must not be ignored. Figure 7 depicts a case with
s
 2  2 no residual stresses showing that wire thinning behind the
un ut debonding front leads to an important mode I contribution,
λ= + (8)
δn δt leaving a gap in the wake of the debonding front (figure 7(b)).
This mode I contribution helps lower pull-out forces during
where un and ut are the current normal and tangential
debonding. However, residual stress from the injection
displacement jumps, respectively, at the interface estimated by
molding process may cause the matrix to partially close the
the FEM, and δn and δt are the corresponding critical values of
gap, reducing the mode I contribution. To study the impact
the displacement jump at which interface failure takes place.
of residual stresses, we carried out a series of simulations
The displacement jump is related to the normal and tangential
that address shrinkage of the TPO during the injection
traction of the interface using a traction-separation law ([65])
molding process. From previous studies [77, 78], a good
for mode I and mode II, namely
indicator of the ratio of volume change due to thermal
1 − λ∗ un
 
Tmax loads is the thermal expansion coefficient. Assuming 2 ×
Tn = (9) 10−5 mm mm−1 ◦ C−1 [59] and a temperature change of
λ∗ δn (1 − λcr )
100 ◦ C (typical of injection molding), we simulated the TPO
1 − λ∗ ut αTmax
 
Tt = . (10) shrinkage process that preceded the pull-out simulations.
λ ∗ δt (1 − λcr ) Although TPO has very high thermal stability, (i.e. shrinkage
Note that λ∗ is monotonically increased by following the during the injection molding process is very small compared
condition with other thermoplastic polymers), our simulations revealed
that the residual stresses indeed reduced the mode I
λ∗ = max(λmax , λ) (11) contribution, leading to values of Fmax that are 10% higher
where λmax = λcr (initially) and λmax = λ if λ > λmax . than those without residual stresses.
Also, Tmax can be regarded as the cohesive strength of the
interface only when normal opening displacement is applied 5.3. Results and discussion
to the interface and τmax = αTmax as the cohesive strength
when only shear (tangential displacement) is applied to the We analyzed NiTi wire/TPO interface behavior with the
interface. The dimensionless parameter α is the normal to analytical models presented in section 5.1 and the FEM
shear strength ratio and λcr is the critical value of the effective described in section 5.2. The ductile model in equation (1) can
displacement jump where the normal or tangential traction only be applied when the debonding process zone spans the
reach their maximum values. The values of δn and δt were entire bond length (see section 5.1). From our experiments,
derived from the geometrical relation that links the strength this is not the case. An NiTi wire debonds from the matrix
and the cohesive energy (or work of separation) of the NiTi following an interfacial debonding propagation process (see
wire/TPO interface: section 4.2), a process that is initiated from the polymer
surface closest to where the force is applied and propagates all
GIc = 12 δn Tmax , GIIc = 21 δt αT max . (12)
the way to the other end (see the debonding process zone in
Here, GIc is the cohesive energy for mode I (i.e. when figure 7(a)). Application of equation (1) to the forces reported
pure normal opening displacement is applied to the interface) in figure 8 would have resulted in grossly underestimated
and GIIc is the cohesive energy for mode II (only shear). The values of the shear strength, τiy (∼0.1–0.2 MPa), and hence
strength and energy of the cohesive interface are characterized the ductile model is inadequate.
based on information from the pull-out tests. In turn, it is Based on LEFM, Penn and Lee [47] studied the
expected that all these cohesive properties will depend on correlation between Fmax and le . For large le , Fmax is
a specific surface treatment. Figure 9(c) shows the cohesive independent of le . Even though the molded NiTi wire/TPO
traction-separation law for the shearing mode (equation (10)). specimens were expected to develop a small fracture process
While the elastic and inelastic properties of NiTi and TPO are zone, the characterization of the interface behavior with
known, the cohesive parameters in equations (8)–(12) must LEFM is not straightforward since the NiTi mechanical
be quantified. Another characteristic length of interest is the response, under the conditions stated above, is inelastic. This
debonding process zone length, lf ∝ Gc Ef /Tmax2 [72–74]. We model has been used by Jonnalagadda et al [11]. Following

12
Smart Mater. Struct. 21 (2012) 035022 F C Antico et al

for sufficiently long embedded lengths (depicted in figure 6).


Therefore, our simulations are indicative of the complete
macroscale debonding process observed in our pull-out tests
(i.e. the debonding initiation, the detwinning of the wire,
debonding front propagation and the complete failure of a
specimen). Following these observations, we performed a
parametric analysis evaluating Fmax as a function of Tmax and
Gc . The solid red line in figure 10 shows the dependence of
Fmax on Gc including TPO residual stresses. Following [11,
15, 16], we considered values of Tmax in the 10–50 MPa
range. However, it was observed that varying Tmax from 10
to 50 MPa affected Fmax by no more than 1% (and up to
3% when Tmax is above 100 MPa). The overall geometry of
the molded specimen is thus insensitive to Tmax , while it is
highly dependent on Gc . The calculations shown in figure 10
assumed Tmax = 20 MPa.
Besides Tmax and Gc , the shape of the cohesive zone
model may be relevant (see figure 9(c)). Previous studies [79]
Figure 10. Fmax versus Gc . Solid red line: FEM accounting for
inelastic behavior of NiTi considering residual stresses. Gray dashed have shown that the shape of the cohesive zone model can
line: FEM accounting for inelastic behavior of NiTi assuming no have a negligible influence on the overall response of an FEM
residual stresses. Green long-dashed line: FEM calculation solution of pure-mode failure models (i.e. only mode I or only
assuming a linearly elastic behavior for the NiTi. Blue mode II) with an adequate spatial discretization. However,
dashed–dotted line: LEFM analytical prediction. The horizontal it was observed in [80, 81] that the local response close to
dashed lines denote the minimum and maximum values recorded in
table 1 (from our pull-out tests). Circular, triangular and square Fmax may vary, depending on the shape of the cohesive zone
symbols indicate discrete values of Fmax obtained from each set of model only when the cohesive zone length is comparable
simulations, respectively. to other relevant characteristic lengths. Some studies suggest
that the shape of the softening part [79] of the cohesive zone
model may affect the overall response of the numerical model
this model, Fmax is plotted with Gc in figure 10 as the and that this can be neglected, depending on the relationship
blue dashed–dotted line. The horizontal black dashed lines between the stiffness of the TPO, NiTi and the cohesive
denote one standard deviation below the minimum mean force interface [79, 82, 83]. Here, a single cohesive zone model
and one standard deviation above the maximum mean force was adopted to represent the pull-out response of an NiTi
measured in our pull-out experiments detailed in table 1 and wire embedded in a TPO matrix, regardless of the surface
figure 8. This curve clearly shows that Gc ranges from 13 treatment, based upon the observation that the response is
to 85 J m−2 for the peak forces reported in our experiments only dependent on the Tmax and Gc . Following Zavattieri
following LEFM. et al [69, 84, 85], this indicates that Tmax and Gc may both
Finally, a series of FEM calculations using the model be controlled by the adhesion mechanisms resulting from our
described in section 5.2 was carried out for different values various NiTi wire surface treatments. Therefore, the effect of
of the cohesive zone parameters. To properly account for the the shape of the cohesive model on our simulation results can
mixed mode conditions at the debonding front, τmax , Tmax , δn be considered to be negligible.
and δt were individually studied to assess their effect on the Additionally, our simulations consider the residual
overall behavior of the specimen geometry show in figure 9. stresses from the injection molding process (open circles in
Typical force versus displacement curves from the FEM are figure 10 represent discrete values of Fmax obtained from
similar to those shown in figure 6. Variations of the cohesive these simulations). For comparison purposes, figure 10 also
parameters lead to different values of Fmax . Consequently, includes predicted results without residual stresses (gray
we performed a systematic parametric numerical analysis dashed curve; triangular symbols represent discrete values of
wherein calculated Fmax are compared with measured values Fmax obtained from these simulations), showing a difference
from our pull-out experiments (section 4.2, table 1 and in force of about 10%. This difference indicates the effect of
figure 8). Our first observation is that moderate variation of an important mode I contribution when there are no residual
Tmax and δn with respect to τmax and δt , respectively, affected stresses, making it easy for the interface to debond under
Fmax by less than 1%. Hence, the NiTi wire/TPO interface combined mode I/II conditions. Ignoring residual stresses can
behavior strongly depends on the mode II properties rather result in a 20–40% overestimation of Gc .
than on the mode I properties. As such, we show results The FEM calculations allowed us to analyze the stress
for δn = δt and α = 1. This limited our analysis to only state and deformation modes in the NiTi wire in great
two main parameters, namely Tmax = τmax and Gc = GIc = detail. It was observed that the stress state along the wire
GIIc = 21 δn Tmax . Our FEM simulations also suggest that, once during the debonding process differs for wire material behind
the debonding front advances, the pull-out force, at the Fmax the debonding process zone, where the wire is completely
level, remains constant, and therefore it is independent of le debonded, and in locations where the wire is still bonded to

13
Smart Mater. Struct. 21 (2012) 035022 F C Antico et al

the TPO. The former is characterized by a state of uniaxial


stress during the entire debonding process. Before debonding
begins, the initially exposed portion of the wire (i.e. the
portion of the wire that comes out of the matrix) deforms
uniformly until the Fmax is reached. This confirms that the
details of the reorientation of the martensitic phase may not
be necessary in the initial elongation part as long as the
model can properly capture the uniaxial stress–strain response
depicted in figure 9(b). Once the debonding front propagates,
the exposed portion of the wire remains at constant uniaxial
stress (σmax in equation (6)), and only the newly liberated
wire deforms from its elastic regime to σmax . Any portion
of the wire that is still bonded to the matrix (ahead of
the debonding front) remains elastic at very low stress
levels; therefore, detwinning does not occur in those regions.
Rather, detwinning occurs only in a localized plane, oriented
perpendicular to the longitudinal axis of the wire, that follows
the debonding process zone as it propagates. This detwinning Figure 11. Gc calculated for NiTi wires with embedded length
zone divides the two regions of the wire previously described. le = 70 mm (red squares) and le = 90 mm (red circles). The
Since the wire stress state is well characterized (it is either in standard deviation is used for the error bars. These results were
obtained using the FEM simulations considering residual stresses
its elastic regime when it is bonded to the matrix, or under
(figure 10). For comparison purposes, the same values obtained with
mostly uniaxial stress–strain conditions), and the detwinning the LEFM model are included as filled gray circles.
zone is well known at all times, we conclude that our model
described in section 5.2 is well suited for this very important
problem. experimental values of Fmax reported in figure 8 for each
In contrast to the results obtained with LEFM, our FEM surface treatment. We employed the results obtained from our
suggests values of Gc in the range of 48–450 J m−2 for FEM considering residual stresses (solid red line in figure 10).
the same pull-out forces obtained in the experiments. This The Fmax –Gc relationship follows a simple polynomial: Gc =
−36.20+0.84Fmax +0.55Fmax 2 (where force is in newtons and
represents a difference of almost one order of magnitude
−2
the energy is in J m ). The experimental mean and standard
for each wire surface treatment, a discrepancy that can be
explained in terms of Gc . In the LEFM model, the NiTi deviation of Fmax were used to obtain the equivalent mean
wire is assumed to undergo linear elastic deformation with values of Gc (with their equivalent standard deviations) as
Ef = 30 GPa. According to the maximum tensile stresses shown in figure 11. Here, Gc increases (on average) from
obtained in our pull-out tests (i.e. σmax ≈ 200 MPa for the the untreated condition, where it is least, to the 5% PPPA
treated wires), the strains attained in the debonded part of treatment, where it is greatest. For comparison purposes,
the wires are very small (i.e., εtot = 0.006) compared with the values of Gc obtained with the LEFM model were also
those obtained with our FEM and the nonlinear stress–strain included in figure 11 in light gray. Figure 11 also shows
behavior shown in figure 9(b) (εtot = 0.03). These small the cohesive energy for specimens with different embedded
elastic strains predicted in the wire lead to an underestimated lengths. Hence, the greatest adhesion is achieved with the
elastic energy stored in the specimen available to advance the PPPA treatments.
debonding front. According to the fracture mechanics concept
of energy release rate [86], this ultimately represents much 6. Concluding remarks
lower values of Gc than those predicted by the FEM model.
To support these findings, we also carried out additional FEM We have considered methodologies for improving adhesion
calculations (following a similar model to that described in between NiTi shape memory alloys in wire form and a
section 5.2) with the only exception that the NiTi behavior TPO polymer matrix. The methods included functionalizing
was modeled with linear elasticity (the cohesive behavior the NiTi surface with an organophosphorus treatment and
for the interface remains the same). For these new cases, wire surface microgeometry roughening through sanding. We
Fmax versus Gc is plotted with a green line in figure 10. observed that the wires that needed the highest pull-out force
Open square symbols represent computed values of Fmax . Fmax were those treated with 5% phenylphosphonic acid
These results are similar to those obtained with the LEFM (PPPA), followed by those treated with 1% PPPA and the
model, indicating that the inelasticity of the stress-induced hand-sanded specimens. The untreated wires had the lowest
detwinning in the martensitic phase of the NiTi is extremely pull-out forces. Since the TPO by itself allows no significant
important in the debonding. Therefore, we conclude that chemical bonding with an NiTi wire surface, it is concluded
LEFM is not appropriate for modeling pull-out tests wherein that the PPPA greatly improves the bonding and adhesive
large inelastic deformation occurs in a NiTi wire. strength at the interface between the NiTi and the TPO.
Once the FEM was established and the Fmax versus Gc From our experimental results we also conclude that it
curves developed, we proceeded to calculate Gc using the is possible to achieve wire tensile stresses that exceed the

14
Smart Mater. Struct. 21 (2012) 035022 F C Antico et al

manufacturer’s recommended values without debonding from Acknowledgments


the polymer matrix, and hence design actuators could take
advantage of the maximum stresses possible from NiTi wires. The authors are grateful to Nancy Johnson and Alan Browne
However, future effort should be aimed at understanding for many stimulating discussions on shape memory alloys and
the effect of cyclic loads and accumulating damage at the potential applications.
interface to apply these treatments to actuators subject to a
high number of cycles. It should also be mentioned that the References
pull-out tests in this study considered only wires at room
temperature. The temperature effect on the interfacial strength [1] Liang C, Jia J and Rogers C A 1989 Behavior of shape
and energy as well as the effect on the wire deformation memory alloy reinforced composite plates—parts I and II
Proc. 30th Structures, Structural Dynamic and Materials
of the temperature-induced martensitic-to-austenitic phase Conf. (Mobile, AL, April) pp 2011–7 (AIAA-89-1331)
transformation must be explored in future studies. [2] Chaudry Z and Rogers C A 1991 Response of composite
Based upon an existing LEFM theoretical model with beams to an internal actuator force Proc. 32nd Structures,
Fmax as input, all NiTi wire surface treatments investigated Structural Dynamic and Materials Conf. (Washington, DC)
pp 186–93 (AIAA-91-11 66)
lead to a comparable improvement in computed cohesive [3] Lagoudas D C and Tadjbakhsh I G 1992 Active flexible rods
energy, Gc , of the wire–polymer matrix interface relative with embedded SMA fibers Smart Mater. Struct. 1 162–7
to an untreated NiTi wire. However, we demonstrated that [4] Lagoudas D C and Tadjbakhsh I G 1993 Deformations of
this model is grossly inadequate. Through consideration of active flexible rods with embedded line actuators Smart
the nonlinear stress–strain behavior of the NiTi wire and Mater. Struct. 2 71–81
[5] de Blonk B J and Lagoudas D C 1998 Actuation of
residual stresses from molding in a refined FE model, we elastomeric rods with embedded two-way shape memory
find that the estimated values of Gc are not only significantly alloy actuators Smart Mater. Struct. 7 771–83
higher than those predicted with LEFM, but they clearly [6] Moore C L and Bruck H A 2002 A fundamental investigation
show that the PPPA treatment improves the adhesion of the into large strain recovery of one-way shape memory alloy
wires embedded in flexible polyurethanes Smart Mater.
interface. Our models also showed that the residual stresses Struct. 11 130–9
from the manufacturing process have an important effect in [7] Bruck H A, Moore C L and Valentine T L 2002 Repeatable
the debonding process by reducing the mode I contribution bending actuation in polyurethanes using opposing
caused by the thinning of the NiTi wire. Neglecting these embedded one-way shape memory alloy wires exhibiting
residual stresses can result in an overestimation of Gc from large deformation recovery Smart Mater. Struct. 11 509–18
[8] Zheng Y J, Cui L S and Schrooten J 2005 Basic design
20% to 40%. While the values of Gc reported are believed guidelines for SMA/epoxy smart composites Mater. Sci.
to be qualitative in nature, they provide information that Eng. A 390 139–43
qualitatively characterizes the treatments explored in this [9] Han H P, Ang K K, Wang Q and Taheri F 2006 Buckling
work. enhancement of epoxy columns using embedded shape
memory alloy spring actuators Compos. Struct. 72 200–11
The decreasing contribution of mode I at room [10] Lagoudas D C (ed) 2006 Shape memory alloys Mech. Mater.
temperature (the test temperature of interest in this study) 38 389–564
helps prevent additional damage mechanisms due to sliding, [11] Jonnalagadda K, Kline G E and Sottos N R 1997 Local
friction and mechanical interlocking. To provide a more displacements and load transfer in shape memory alloy
composites Exp. Mech. 37 78–86
definitive test of the different chemical and mechanical [12] Umezaki E 2000 Improvement in separation of SMA from
treatments detailed here, future investigations should be matrix in SMA embedded smart structures Mater. Sci. Eng.
focused on wire pull-out experiments wherein the wire A 285 363–9
is actuated (through a temperature-induced martensitic-to- [13] Lau K, Tam W, Meng X and Zhou L 2002 Morphological
austenitic phase transformation), and when the wire is in the study on twisted NiTi wires for smart composite systems
Mater. Lett. 57 364–8
austenitic phase and pull-out through an external force. The [14] Neuking K, Abu-Zarifa A and Eggeler G 2008 Surface
consequences of such a study will reveal any increase in engineering of shape memory alloy/polymer-composites:
Gc due to these additional energy dissipation mechanisms. improvement of the adhesion between polymers and
Along those lines, we suggest that improvements to pseudoelastic shape memory alloys Mater. Sci. Eng. A
481/482 606–11
cohesive models should include characterization of the [15] Rossi S, Deflorian F, Pegoretti A, D’Orazio D and
NiTi surface topography at the micron level (see, for Gialanella S 2008 Chemical and mechanical treatments to
example, [84, 85]), as well as atomistically informed cohesive improve the surface properties of shape memory NiTi wires
laws that properly account for the chemistry and physics Surf. Coat. Technol. 202 2214–22
of the NiTi/organophosphorus/polymer interface. Indeed a [16] Sadrnezhaad S K, Nemati N H and Baheri R 2009 Improved
adhesion of NiTi wire to silicone matrix for smart
numerical model could be developed that accounts for all composite medical applications Mater. Des. 30 3667–72
of the aforementioned features of these composite materials [17] Jang B K and Kishi T 2005 Thermomechanical response of
within the spirit of a multiscale computational modeling and TiNi fiber-impregnated CFRP composites Mater. Lett.
design [88]. The development of thermo-mechanical-based 59 2472–5
[18] Barsoum R G S 1997 Active materials and adaptive structures
cohesive models that account for the effect of microgeometry, Smart Mater. Struct. 6 117–22
cyclic loads and temperature effects will be the focus of future [19] Smith N A, Antoun G G, Ellis A B and Crone W C 2004
work. Improved adhesion between nickel–titanium shape memory

15
Smart Mater. Struct. 21 (2012) 035022 F C Antico et al

alloy and a polymer matrix via silane coupling agents [41] Piggott M R 1991 Failure process in the fiber–polymer
Composites A 35 1307–12 interphase Compos. Sci. Technol. 42 57–76
[20] Wang G and Shahinpoor M 1997 Design, prototyping and [42] Desarmot G and Favre J-P 1991 Advances in pull-out testing
computer simulations of a novel large bending actuator and data analysis Compos. Sci. Technol. 42 151–87
made with a shape memory alloy contractile wire Smart [43] Tsai J H, Patra A and Wetherhold R 2005 Finite element
Mater. Struct. 6 214–21 simulation of shaped ductile fiber pull-out using a mixed
[21] Sun S S, Sun G and Wu J S 2002 Thermo-viscoelastic bending cohesive zone/friction interface model Composites A
analysis of a shape memory alloy hybrid epoxy beam 36 827–38
Smart. Mater. Struct. 11 970–5 [44] Poon C-K, Zhou L-M and Yam L-H 2004 Size effect on the
[22] McBain J W and Hopkins D G 1925 J. Phys. Chem. 29 188 optimum actuation condition for SMA-composites Compos.
[23] Lee G, Cheung K, Chang W and Lee L P 2000 Poster 91 First Struct. 66 503–11
Annual Int. IEEE-EMBS Special Topic Conf. on [45] Poon C-K, Lau K-T and Zhou L-M 2005 Design of pull-out
Microtechnologies in Medicine and Biology (Lyon, Oct.) stresses for prestrained SMA wire/polymer hybrid
[24] Weiss C and Muenstedt H 2002 Surface modification of composites Composites B 36 25–31
polyether ether ketone (peek) films for flexible printed [46] Wang X and Hu G 2005 Stress transfer for SMA fiber pulled
circuit boards J. Adhes. 78 507–19 out from an elastic matrix and related bridging effect
[25] Du X W, Sun G and Sun S S 2004 A study on the deflection of Composites A 36 1142–51
shape memory alloy (SMA) reinforced thermo-viscoelastic [47] Penn L S and Lee S M 1989 Interpretation of experimental
beam Compos. Sci. Technol. 64 1375–81 results in the single pull-out filament test J. Compos.
[26] Hector L G Jr, Opalka S M, Nitowski G, Wieserman L, Technol. Res. 11 23–30
Siegel D J, Yu H and Adams J B 2001 Investigation of vinyl [48] Biro D A, Pleizier G and Deslandes Y 1993 Application of the
phosphonic acid/hydroxylated α-Al2 O3 (0001) reaction microbond technique: effects of hygrothermal exposure on
enthalpies Surf. Sci. 494 1–20 carbon fiber/epoxy interfaces Compos. Sci. Technol.
[27] Nilsing M, Lunell S, Persson P and Ojamäe L 2005 46 293–301
Phosphonic acid adsorption at the TiO2 anatase (101) [49] Gohil P P and Shaikh A A 2009 To study the effect of surface
surface investigated by periodic hybrid HF-DFT roughness and embedded length in steel–polyester
computations Surf. Sci. 582 49–60 composite J. Reinf. Plas. Compos. 28 385–92
[28] Gouzman I, Dubey M, Carolus M D, Schartz J and [50] Baley C, Grohens Y, Busnel F and Davies P 2004 Application
Bernasek S L 2006 Monolayer versus multilayer of interlaminar tests to marine composites. Relation
self-assembled alkylphosphonate films: x-ray photoelectron between glass fibre/polymer interfaces and interlaminar
spectroscopy studies Surf. Sci. 600 773–81 properties of marine composites Appl. Compos. Mater.
[29] Hector L G Jr, Mance A M, Rodgers W R, Zavattieri P D, 11 77–98
Okonski D A, Sherman E and Barvosa-Carter W 2011 [51] Piggott M R 1987 Debonding and friction at fibre–polymer
Method for improving adhesion between a shape memory interfaces. I: criteria for failure and sliding Compos. Sci.
alloy and a polymer US Patent Specification 7,993,537 Technol. 30 295–306
(Aug. 9) [52] Brinson L C 1993 One-dimensional constitutive behavior of
[30] www.dynalloy.com shape memory alloys: thermomechanical derivation with
[31] Funakubo H (ed) 1984 Shape Memory Alloys (New York: non-constant material functions J. Intell. Mater. Syst. Struct.
Gordon and Breach Science) 4 229–42
[32] Brailovski V, Prokoshkin S, Terriarult P and Trochu F (ed) [53] Boyd J G and Lagoudas D C 1994 Thermomechanical
2003 Shape Memory Alloys: Fundamentals, Modeling and response of the shape memory composites J. Intell. Mater.
Applications (Quebec: Ecole de Technologie Superiure) Syst. Struct. 5 333–46
[33] Luntz J E, Brei D, Ypma J, Young J R, Radice J, Johnson N L, [54] Voce E 1955 A practical strain-hardening function Metallurgia
Browne A L and Strom K A 2007 Feasibility study of the 51 219–26
dual chamber SMART (SMA resettable) lift device [55] Taylor R L 2008 User Manual FEAP—a finite element
industrial and commercial applications of smart structures analysis program, version 8.2
technologies Proc. SPIE 6527 65270V [56] Gao X and Brinson L C 2002 A simplified multivariant SMA
[34] Luntz J E, Young J R, Brei D, Radice J and Strom K A 2007 model based on invariant plane nature of martensitic
Modeling and sensitivity study of the dual chamber transformation J. Intell. Mater. Syst. Struct. 13 795–810
SMART (SMA ReseTtable) lift device, modeling, signal [57] Churchill C B and Shaw J A 2008 Shakedown response of
processing, and control for smart structures Proc. SPIE conditioned shape memory alloy wire Behavior and
6523 652305 Mechanics of Multifunctional and Composite Materials
[35] Redmond J A, Brei D, Luntz J, Browne A L and p 6928
Johnson N L 2009 Effect of bending on the performance of [58] Reedlunn B, Daly S, Hector L Jr, Zavattieri P D and
spool-packaged shape memory alloy actuators, industrial Shaw J 2011 Tips & tricks for characterizing shape memory
and commercial applications of smart structures wire part 5: full-field strain measurement by digital image
technologies Proc. SPIE 7290 729007 correlation J. Exp. Tech. doi:10.1111/j.1747-1567.2011.
[36] 2003 Using Nitinol Alloys (San Jose, CA: Johnson Matthey) 00717.x
www.jmmedical.com/ [59] www.matweb.com
[37] Hector L G Jr and Sheu S 1991 Brightness enhancement with [60] Dugdale D S 1960 Yielding of steel sheets containing slits
textured roll US Patent Specification 4,996,113 J. Mech. Phys. Solids 8 100–4
[38] Surface Metrology Guide On-line www.predev.com/smg [61] Barenblatt G I 1962 The mathematical theory of equilibrium
[39] Burnham N A, Behrend O P, Oulevey F, Gremaund G, cracks in brittle fracture Adv. Appl. Mech. 7 55–129
Gallo P-J, Gourdon D, Dupas E, Kulik A J, [62] Tvergaard V and Hutchinson J W 1992 The relation between
Pollock H M and Briggs G A D 1997 How does a tip tap? crack growth resistance and fracture process parameters in
Nanotechnology 8 67–75 elastic–plastic solids J. Mech. Phys. Solids 40 1377–97
[40] Hutchinson J W and Jensen H M 1990 Models of fiber [63] Xu X-P and Needleman A 1994 Numerical simulations of fast
debonding and pull-out in brittle composites with friction crack growth in brittle solids J. Mech. Phys. Solids
Mech. Mater. 9 139–63 42 1397–434

16
Smart Mater. Struct. 21 (2012) 035022 F C Antico et al

[64] Camacho G and Ortiz M 1996 Computational modeling of [75] Tvergaard V 1990 Effect on fibre debonding in a
impact damage in brittle materials Int. J. Solids Struct. whisker-reinforced metal J. Mater. Sci. Eng. A 125 203–13
33 2899–938 [76] Tvergaard V 1990 Analysis of tensile properties for a
[65] Espinosa H D and Zavattieri P D 2003 A grain level model for whisker-reinforced metal–matrix composite Acta Metall.
the study of failure initiation and evolution in Mater 38 185–94
polycrystalline brittle materials. Part I: theory and [77] Lee H, Fasulo D P, Rodgers W R and Paul D R 2006 TPO
numerical implementation Mech. Mater. 35 333–64 based nanocomposites. Part 2. Thermal expansion behavior
[66] Espinosa H D, Zavattieri P D and Dwivedi S K 1998 A finite Polymer 47 3528–39
deformation continuum/discrete model for the description [78] Kim D H, Fasulo D P, Rodgers W R and Paul D R 2008 Effect
of fragmentation and damage in brittle materials J. Mech. of the ratio of maleated polypropylene to organocaly on the
Phys. Solids 46 1909–42 structure and properties of TPO-based nanocomposties.
[67] Espinosa H D, Dwivedi S and Lu H-C 1998 Modeling impact Part II: thermal expansion behavior Polymer 49 2492–506
induced delamination of woven fiber reinforced composites [79] Alfano G 2006 On the influence of the shape of the interface
with contact/cohesive laws Comput. Methods Appl. Mech. law on the application of cohesive-zone models Compos.
Eng. 183 259–90 Sci. Technol. 66 723–30
[68] Zavattieri P D and Espinosa H D 2001 Grain level analysis of [80] Bao G and Suo Z 1992 Remarks on crack-bridging concepts
crack initiation and propagation in brittle materials Acta Appl. Mech. Rev. 45 355–66
Mater. 49 4291–311 [81] Parmigiani J P and Thouless M D 2007 The effects of cohesive
[69] Li S, Thouless M D, Waas A M, Schroeder J and strength and toughness on mixed-mode delamination of
Zavattieri P D 2005 Use of mode-I cohesive-zone models to beam-like geometries Eng. Fract. Mech. 74 2675–99
[82] Song S H, Paulino G H and Buttlar W G 2006 Parameter on
describe the fracture of an adhesively-bonded
asphalt concrete fracture behavior Multiscale and
polymer–matrix composite Compos. Sci. Technol.
Functionally Graded Materials (New York: AIP) pp 730–5
65 281–93
(CP973)
[70] Li S, Thouless M D, Waas A M, Schroeder J and [83] Volokh K Y 2004 Comparison between cohesive zone models
Zavattieri P D 2006 Mixed-mode cohesive-zone models for Commun. Numer. Methods Eng. 20 845–56
fracture of an adhesively-bonded polymer–matrix [84] Zavattieri P D, Hector L G Jr and Bower A F 2007
composite Eng. Fract. Mech. 73 64–78 Determination of the mode-I effective fracture toughness of
[71] Zavattieri P D 2006 Modeling of crack propagation in a sinusoidal interface between two elastic solids Int. J.
thin-walled structures with a cohesive model for shell Fract. 145 167–80
elements Comput. Mech. J. Appl. Mech. 73 948–58 (Special [85] Zavattieri P D, Hector L G Jr and Bower A F 2008 Cohesive
issue) zone simulations of crack growth along a rough interface
[72] Morrissey J W and Rice J R 1998 Crack front waves J. Mech. between two elastic plastic solids Eng. Fract. Mech.
Phys. Solids 46 467–87 75 4309–32
[73] Falk M L, Needleman A and Rice J R 2001 A critical [86] Anderson T L 1995 Fracture Mechanics—Fundamentals and
evaluation of cohesive zone models of dynamic fracture Applications (Boca Raton, FL: CRC Press)
Proc. 5th European Mechanics of Materials Conf. on Scale [87] Okonski D A and Zavattieri P D 2011 Methods of forming
Transitions From Atomistics to Continuum Plasticity vol 11, polymeric articles having continuous support structures US
pp 43–50 Patent Specification 20,110,115,114 (May 19)
[74] Rice J R 1980 The Mechanics of earthquake rupture Physics [88] Krajewski P E, Hector L G Jr, Qi Y, Mishra R K,
of the Earth’s Interior (Proc. Int’l School of Physics Sachdev A K, Bower A F and Curtin W A 2011 Atoms to
‘Enrico Fermi,’ ed A M Dziewonski and E Boschi) autos: a multi-scale approach to modeling aluminum
(Amsterdam: North-Holland) pp 555–649 deformation JOM 63 24–32

17

You might also like