Composites Part B 120 (2017) 27e34

Contents lists available at ScienceDirect

Composites Part B
journal homepage: www.elsevier.com/locate/compositesb

Prediction of modulus at various strain rates from dynamic

mechanical analysis data for polymer matrix composites
Steven Eric Zeltmann a, *, Keerthana A. Prakash a, Mrityunjay Doddamani b, Nikhil Gupta a
a
Composite Materials and Mechanics Laboratory, Mechanical and Aerospace Engineering Department, Tandon School of Engineering, New York University,
Brooklyn, NY 11201, USA
b
Advanced Manufacturing Laboratory, Department of Mechanical Engineering, National Institute of Technology Karnataka, Surathkal, India

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

Article history: Understanding and modeling the behavior of polymers and composites at a wide range of quasi-static
Received 20 February 2017 and high strain rates is of great interest to applications that are subjected to dynamic loading condi-
Received in revised form tions. The Standard Linear Solid model or Prony series frameworks for modeling of strain rate dependent
27 March 2017
behavior are limited due to simplicity of the models to accurately represent a viscoelastic material with
Accepted 28 March 2017
Available online 29 March 2017
multiple relaxations. This work is aimed at developing a technique for manipulating the data derived
from dynamic mechanical analysis to obtain an accurate estimate of the relaxation modulus of a material
over a large range of strain rate. The technique relies on using the time-temperature superposition
Keywords:
Polymer-matrix composites (PMCs)
principle to obtain a frequency-domain master curve, and integral transform of this material response to
Thermoplastic resin the time domain using the theory of viscoelasticity. The relaxation function obtained from this technique
Analytical modeling is validated for two polymer matrix composites by comparing its predictions of the response to uniaxial
Mechanical testing strain at a prescribed strain rate to measurements taken from a separate set of tension experiments and
excellent matching is observed.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction dynamic loading of structures. There are numerous difficulties in

determining strain rate sensitivity using traditional testing. First, a
Dynamic mechanical analysis (DMA) is a widely used technique large number of specimens are required to be tested at a wide range
in the polymer science fields for characterizing thermal transitions of strain rates. Second, universal testing machines do not have the
in materials. It has traditionally been used primarily for the response rates to measure load and deformations accurately at
detection of thermal transitions, and associated properties such as higher rates. This is particularly challenging in composite materials
glass transition temperature (Tg), polymer blend miscibility and because the linear elastic region is often small and composites may
composite interfacial bonding [1e10]. Recent works have focused undergo brittle failure before only a handful of data points can be
on expanding the utility of the DMA technique to other applications captured at moderate strain rates. High strain rate properties of
such as to determine the properties of heterogeneous materials and materials are studied experimentally [14e20] using techniques
for determination of crack location in composites [11e13]. In these such as the tensile Hopkinson bar for use in analytical studies for
existing cases, storage (E0 ) and loss (E00 ) moduli measured through structures and foams [21]. The use of multiple techniques for
DMA are interpreted with respect to the material microstructure. different strain rate regimes and unavailability of cross-correlation
The focus of this work is to determine the strain rate sensitivity between properties obtained from quasi-static and dynamic test
of materials by developing a transform to convert the frequency methods has emerged as a significant limitation. A central chal-
domain DMA results to time domain. Such a transform can signif- lenge to experimental characterization by traditional methods is
icantly improve the usefulness of DMA results in materials and that over small ranges of strain rate (a few orders of magnitude)
mechanics fields by providing the modulus at a wide range of strain many properties may appear to vary linearly with the logarithm of
rates, which is required in modeling and simulation studies on the strain rate (i.e. follow a power law) [22,23], while over a wider
range this is usually not the case [24,25]. Extrapolation of the trends
observed in the low-moderate strain rate regime to high strain
* Corresponding author. rates can lead to significant error. This work explores the use of an
E-mail address: steven.zeltmann@nyu.edu (S.E. Zeltmann).

http://dx.doi.org/10.1016/j.compositesb.2017.03.062
1359-8368/© 2017 Elsevier Ltd. All rights reserved.
28 S.E. Zeltmann et al. / Composites Part B 120 (2017) 27e34

alternative method for predicting strain rate sensitivity in polymers 2.2. Sample fabrication
and composites: using time-temperature superposition (TTS) of
dynamic mechanical analysis (DMA) data to determine the fre- Cenospheres are used in 20 and 40 wt% in HDPE to fabricate two
quency domain response over a sufficiently wide range that it can types of syntactic foams named HDPE20 and HDPE40, respectively,
be transformed to the time domain relaxation function. on a polymer injection molding (PIM) machine (Windsor, 80 ton
Using the DMA transform technique of Zeltmann et al. [26], capacity). Operating and processing parameters of the PIM machine
predictions of the strain rate sensitivity of the modulus of high were optimized in a set of earlier studies [23,46e48] and are set at
density polyethylene (HDPE) matrix fly ash cenosphere reinforced 160  C temperature and 30 kg/cm2 (2.9 MPa) pressure. Samples of
composites called syntactic foams are computed and verified dimensions 60  12.7  3.3 mm3 (length  width  height) are
against results from tensile testing. Various properties of syntactic molded. The length of the specimens is later reduced by cutting
foams have been widely studied in the literature [27e29], and with a diamond saw to 35 mm for DMA testing. Samples for tensile
experimental studies are available on the high strain rate behavior testing are also injection molded, with their geometry following
of syntactic foams [30]. This method does not rely on an a priori ASTM D638.
form for the constitutive model, as often used in modeling strain
rate sensitivity [31]. This work also explores further the application 2.3. Dynamic mechanical analysis
of the method, including using the Kramers-Kronig relations to
compute an associated loss modulus function that provides a better Dynamic mechanical analysis is conducted using a TA In-
indication of whether sufficient information has been captured to struments (New Castle, DE) Q800 DMA. Specimens are tested in the
accurately represent the response. From the loss modulus master single cantilever configuration with a span length of 17.5 mm.
curve one can find the center of the transition, which can then be Testing is conducted in the strain control mode with a maximum
used to determine the critical strain rate at which the transition displacement of 25 mm. This displacement is achievable within the
causes a significant curve in the strain rate sensitivity of viscoelastic load limits of the machine at all temperatures for all the tested
properties. specimens. This ensures that the full specified strain is achieved
This work aims at providing a resource for simulation and throughout the entire test.
modeling efforts by providing a more sophisticated means of DMA testing is conducted in two phases: (a) using the temper-
characterizing the dynamic response of the material and by ature sweep mode at constant frequency of 1 Hz and (b) using the
correlating the elastic and viscoelastic properties. The results of the frequency sweep mode at a series of constant temperatures. In the
DMA transform technique can be utilized directly in many existing temperature sweep test, the temperature is ramped from 100  C
mechanical models [32e36] through the elastic-viscoelastic cor- to 130  C at a rate of 1  C/min. At least three specimens of each type
respondence principle and for many other polymer matrix com- are tested in this phase. In the frequency sweep testing, the tem-
posites systems of interest [37e41]. perature is stepped from 10  C to 130  C in increments of 5  C. At
This work uses fly ash hollow particles dispersed in HDPE to each temperature step the specimen is soaked for 8 min to ensure
create a type of composite known as syntactic foam [42,43]. This thermal equilibrium. The dynamic properties are measured at 20
class of composite materials has been widely studied for its me- discrete frequencies logarithmically spaced between 1 and
chanical properties and strain rate sensitivity [43,44]. Particulate 100 Hz at each temperature step. Testing is halted if E0 drops below
composites such as the syntactic foams studied in this work are an 10 MPa, in order to prevent melting of the specimen in the clamp.
ideal model system for exploring this transformation method At least three specimens of each type are tested in this phase. The
because their mechanical properties are isotropic, so the properties results of temperature and frequency sweeping are combined using
measured in bending in the DMA are assumed to be the same as the time-temperature superposition (TTS) principle to generate
those measured in tension in the strain rate sensitivity experi- master curves describing the behavior of syntactic foams over a
ments. Since the measurements are all taken on bulk properties, wider range of frequencies.
more complex systems such as those with viscoelastic fillers in a
viscoelastic matrix are also able to be treated by this method, so 2.4. Tensile testing
long as they are within the linear elastic regime. Nevertheless, the
transformation method discussed here is not limited to isotropic Tensile testing is conducted using an Instron 4467 universal test
materials; in the case of anisotropy the material functions in system. The specimen geometry conforms to ASTM D638. Strain
different directions would be determined from separate measure- data is collected using an Instron 100 gage length extensometer
ments and combined using just as in the familiar theory of linear attached to the test specimen. Tests are conducted with the
elasticity. crosshead speed set to give initial strain rates of 105-102 s1,
which are taken as the nominal strain rates for the tests. At least
five specimens of each material type are tested at each strain rate.
2. Experimental
2.5. Scanning electron microscopy
2.1. Materials
A Hitachi S3400 N scanning electron microscope is used for
HDPE of grade HD50MA180 supplied by Reliance Polymers, imaging of the syntactic foam microstructure. Specimens are
Mumbai, India, is used as the matrix material. The HDPE had a melt sputter coated with gold using sputter coater to improve conduc-
flow index of 20 g/10 min (190 C/2.16 kg) and a mean molecular tivity and prevent charging.
weight of 97,500 g mol1. Cenospheres of CIL-150 grade, supplied
by Cenosphere India Pvt. Ltd., Kolkata, India, are used as hollow 3. Results
fillers. Cenospheres are used in the as-received condition, without
any surface treatment. Chemical, physical and sieve analysis results Scanning electron micrographs of freeze-fractured surfaces of
on the same type of cenospheres have been reported previously the syntactic foams are shown in Fig. 1. Fly ash cenospheres are
[45]. The cenospheres primarily comprise of alumina, silica, cal- observed to be distributed uniformly throughout the syntactic
cium oxide and iron oxides. foam. Various factors such as matrix porosity entrapment, particle
 S.E. Zeltmann et al. / Composites Part B 120 (2017) 27e34 29

Fig. 1. Scanning electron micrographs of (a) HDPE20 and (b) HDPE40 freeze-fractured surfaces, showing uniform distribution of hollow particles. Particle fracture during processing
does not appear to be significant.

crushing and quality of particle-matrix interface depend on the tensile strength and failure strain can be found in the literature
specimen fabrication method. These factors affect the mechanical [46,47]. In syntactic foams, the matrix is usually assumed to be the
properties, including elastic and viscoelastic properties, of syntactic only strain rate sensitive component, as the ceramic particles
foams. Extensive literature is available on all these factors for HDPE/ commonly used have insignificant damping compared to the
cenosphere syntactic foams as well as numerous other types of matrix.
syntactic foams. However, that information is not replicated here
because the focus of this work is to examine if a transform can be
developed to convert the frequency domain DMA results to time
3.2. DMA temperature sweep
domain data to estimate the elastic modulus at various strain rates
for these composites. Effect of various factors would be similar on
Results from the temperature sweep at 1 Hz are shown in Fig. 4
tensile and DMA results and the developed transform should not be
and Fig. 5 for HDPE20 and HDPE40, respectively. Both syntactic
dependent on the material quality.
foams have a peak in the loss modulus at about 37  C, which is due
to the a-transition in HDPE [49]. Based on the loss modulus peak or
3.1. Tensile testing the increase in tan d, the onset of this transition is at about 0  C. The
transition continues until the HDPE is melted. At 100  C there is
A representative set of stress-strain curves for the HPDE resin evidence of another peak in the loss modulus, though the data do
are shown in Fig. 2 at 105 to 102 s1 strain rates. These curves do not include low enough temperatures to fully capture this peak. The
not show distinct elastic and plastic regions. In addition, the effect peak in HDPE at 110  C is generally associated with the glass
of strain rate is clearly visible due to the viscoelastic nature of the transition. Since the HDPE is above its Tg and within the a-transition
material. Representative stress-strain curves for the HDPE/fly ash range at room temperature, strong strain rate sensitivity in its
syntactic foams at strain rates from 105 to 102 s1 are shown in mechanical properties are expected.
Fig. 3. The syntactic foams show an initial nearly linear region up to
about 0.5% strain, after which yielding occurs. The modulus in the
initial elastic region and the yield stress increase with the strain 3.3. DMA frequency sweep and time-temperature superposition
rate, while the strain at yielding decreases. A more detailed analysis
of the effect of processing parameters and composition on the Following the temperature sweep, frequency sweeps are con-
ducted from 1 to 100 Hz at stepped temperatures from 10 to
130  C. A set of representative frequency sweeps for HDPE20 is
presented in Fig. 6a. These frequency sweeps are combined using
the time-temperature superposition (TTS) principle by shifting the
various frequency sweeps along the frequency axis so that they
form a single master curve [50]. No vertical shifting of the curves is
employed in the TTS procedure. Representative master curves ob-
tained for the two types of syntactic foams are shown in Fig. 6b. One
step change in the storage modulus is observed in the master curve.
The glass transition whose onset was observed at the lowest tem-
peratures in the temperature sweep does not appear to have been
captured here. Because the experiments are conducted using liquid
nitrogen cooling, the total run time is limited. Since frequency
sweeps require dwelling at constant temperature and substantial
soaking times, the minimum temperature that can be reached
before the cooling system is depleted is higher than for a constant
Fig. 2. Representative stress-strain curves for neat HDPE at various strain rates. frequency experiment.
30 S.E. Zeltmann et al. / Composites Part B 120 (2017) 27e34

Fig. 3. Representative stress-strain curves for (a) HDPE20 and (b) HDPE40 at various strain rates.

Fig. 4. Temperature sweep at 1 Hz for HDPE20 syntactic foam: (a) E0 and E00 and (b) tand.

Fig. 5. Temperature sweep at 1 Hz for HDPE40 syntactic foam: (a) E0 and E00 and (b) tand.

4. Discussion and design purposes. From E0 , and using the linear theory of
viscoelasticity, the time domain relaxation modulus E(t) can be
4.1. Transformation to time-domain found using [51]

Using the TTS principle, E0 is found over a sufficiently wide range Z∞

of frequencies to adequately characterize a viscoelastic function of 2 E0 ðuÞ
EðtÞ ¼ sinðutÞdu (1)
the material. This frequency-domain viscoelastic function can, us- p u
ing an appropriate transformation, be converted to any of the other 0

viscoelastic functions which may be more useful for engineering

where u and t represent angular frequency and time, respectively.
 S.E. Zeltmann et al. / Composites Part B 120 (2017) 27e34 31

of the data yields d z a. The frequency at which E0 (u) ¼ 0, and

below which E0 is negative, is

bc  atanhðd=aÞ
ujE’ ¼0 ¼ exp½  (3)
b

which is of the order of 1016 Hz for both syntactic foams. While it is

simple to enforce that E0 (u) ¼ 0 in the curve fitting, this yields a
poorer fit in the moderate frequency ranges which are most
important to the response. In the unconstrained fit, the frequency
at which the function is negative is sufficiently small that the
negative E0 will not substantially affect the resulting relaxation
function, so the parameters from the unconstrained fit are used in
the subsequent analysis.
The Kramers-Kronig (K-K) relations can be used to obtain the
loss modulus function E00 (u) that corresponds to a given function
for E0 (u) via the integral transform [52]

Z∞
00 2u E0 ðlÞ  E0 ðuÞ
E ðuÞ ¼ dl (4)
p l2  u2
0

which can be approximated by

00 p dE0 ðuÞ
E ðuÞz (5)
2 d ln u
Though the transform could be integrated numerically, and
more accurate but complex approximations exist [53], this simple
form has been found to be widely applicable [51,54] and allows the
use of an analytical form using the fitting function selected here.
Based on the approximation of the Kramers-Kronig relations, the
sigmoidal storage modulus function yields the loss modulus
function

Fig. 6. (a) Representative set of frequency sweeps for HDPE20 and (b) storage modulus 00 pab
master curves constructed using 25  C as the reference temperature for HDPE syntactic E ðuÞ ¼ sechðbðlogðuÞ þ cÞ2 (6)
2
foams.

which will be referred to as the “K-K prediction.” The frequency at

which this is maximal, uT, and which is the typical definition of the
Since this is an improper integral, the experimental data must be
location of the transition, is found to be
extrapolated to zero and infinite frequencies. To do so the storage
modulus master curve at a chosen temperature is fitted to a  
1
sigmoidal function of log(u) having the form uT ¼ exp (7)
c
E0 ðuÞ ¼ a tanhðbðlogðuÞ þ cÞÞ þ d (2) Thus one can obtain a frequency-temperature pair that corre-
sponds to a transition. The transition temperature corresponding to
where a, b, c, and d are the fit coefficients and log(u) is the natural this transition at another frequency can be found using the WLF
logarithm. The fit coefficients and the goodness of fit are given in equation and its experimentally determined coefficients [55]. Since
Table 1. A fit of this form imposes that there is one smooth step transition temperatures are usually defined at 1 Hz, this method
transition in the E0 curve, corresponding to one peak in E00 , and that could be used to find transitions that are below the temperature
the behavior is asymptotic as frequency goes to zero or to positive range of the test equipment by using the higher frequency data and
infinity. For the case where the experimental data captures multi- TTS to extend the range. The magnitude of the extension of the
ple transitions, a mixture of functions of this form can be applied, temperature range by this technique would depend on the TTS shift
and the basis of the analysis in the following paragraphs would still factors, which quantify the relationship between changes in fre-
apply. Such choice satisfies the physically required positive and quency and changes in temperature. The shift factors outside the
bounded behavior of the relaxation function at zero and infinite experimental temperature range can be estimated by the WLF [56]
frequencies [51] if d > a. However as the HDPE is above its Tg its or Arrhenius equations, as applicable. The fit parameter b is related
behavior is like that of a viscoelastic liquid (i.e. E0 (u) ¼ 0) and fitting to the breadth, or “sharpness” of the transition.
The K-K prediction of the loss modulus curve is compared to the
experimentally determined loss modulus master curves in Fig. 7.
Table 1
Fit coefficients from the single-sigmoidal fit of the storage modulus master curve
The loss modulus master curves are constructed by shifting the loss
(Equation (2)), with E0 in MPa and u in radians/sec. modulus curves from the frequency sweeps using the shift factors
determined from shifting the storage modulus curves. At fre-
Material a b c d R2
quencies below the peak, excellent agreement is observed between
HDPE20 1224.9 0.06155 3.6566 1167.3 0.9993 the prediction and the master curve. However, above the peak the
HDPE40 1276.1 0.05935 4.1236 1193.9 0.9992
K-K prediction tends towards zero while the master curve remains
32 S.E. Zeltmann et al. / Composites Part B 120 (2017) 27e34

Fig. 7. Comparison of the experimental loss master curves at 25  C with the predictions of the approximate K-K relations (Equation (6)), using the coefficients of the fit function for
(a) HDPE20 and (b) HDPE40.

higher. This is likely due to the presence of another peak at higher well beyond practical time scales.
frequencies (lower temperatures) than what was captured in the The time-domain relaxation function determines the stress
frequency sweep experiments. The use of the K-K prediction in history generated by a specified strain history (for the one-
comparison to the experimental loss modulus master curve is a dimensional case) according to [51]
more sensitive means to assess the goodness of fit, since transitions
near the minimum and maximum frequency are difficult to discern Zt
dεðtÞ
from the storage modulus master curve. Such deviations appear sðtÞ ¼ E  dε ¼ Eðt  tÞ dt (9)
dt
clearly on the loss modulus master curve since the loss modulus ∞
shows peaks at these transitions.
The transform in Equation (1) is applied to the storage modulus where s, ε and t represent stress, strain and time variable used for
fitting function by integrating numerically, to yield the time integration, respectively. In anisotropic materials, this becomes a
domain relaxation function for the material. The integral is the tensor expression (see Ref. [51]). However, we assume that the
product of a positive and decreasing function with a sinusoid, so the syntactic foams used in this work are isotropic as the reinforcing
improper integral can be converted to the summation of an alter- particles are spherical and uniformly distributed. In fiber-
nating series whose terms are the integral evaluated between the reinforced composites, DMA measurements would need to be
zeroes of the sinusoid. An accelerated convergence approach can performed in different orientations to build the relaxation function
then be used to approximate the infinite summation. tensor, which could be used for arbitrary states of strain.
The relaxation function generated using the room temperature For constant strain rate deformation with a strain rate of ε_
master curve is shown in Fig. 8. The relaxation function can be beginning at t ¼ 0, which is the idealized deformation state in a
observed to satisfy the requirements of fading memory and standard tension test, the convolution integral simplifies to
nonnegative stored and dissipated energy as expressed by Ref. [57]
Zt
2
E(t)0, dE(t)/dt  0, d E(t)/dt  0 2
(8) sðtÞ ¼ ε_ EðtÞdt (10)
0
within the range of times displayed. Since d < a in the fitted
functions, at some time the relaxation function will violate the first and the corresponding strain is obviously
condition and yield a negative value for E(t). However, this cross-
over is observed to happen at around 1014 s (about 3 million years), εðtÞ ¼ ε_ t (11)
From this procedure, the linear viscoelastic response of the
material at a constant strain rate can be determined. From the
stress-strain relationship that results, relevant material properties
can be derived.
Since the elastic response of HDPE is not perfectly linear, the
predictions are compared with the results from tension experi-
ments for HDPE matrix syntactic foams using the secant modulus at
a prescribed strain. The 0.5% secant modulus derived from the DMA
transform technique are compared with the results from the ten-
sion experiments in Fig. 9. The DMA transform line is shown only
for one specimen but multiple specimens were tested and they
showed consistent behavior. Excellent agreement is observed for
both syntactic foams between the secant modulus and DMA
transformation results. This method is not expected to be affected
by factors such as imperfect interface in the composite, particles
crushed during composite fabrication or air entrapment in the
Fig. 8. Relaxation functions for HDPE syntactic foams at 25  C. matrix because both the DMA measurements and the tensile test
 S.E. Zeltmann et al. / Composites Part B 120 (2017) 27e34 33

Fig. 9. Strain rate sensitivity of (a) HDPE20 and (b) HDPE40 compared to results from tensile experiments.

results will be affected similarly by such factors. Therefore, the 011016e011016-7.

[2] Saba N, Jawaid M, Alothman OY, Paridah MT. A review on dynamic mechanical
DMA transformed results should still match with the tensile test
properties of natural fibre reinforced polymer composites. Constr Build Mater
data. In composites containing low volume fraction of particles, 2016;106:149e59.
segregation of particles may be a concern that can provide a large [3] Jose S, Thomas S, Parameswaranpillai J, Aprem AS, Karger-Kocsis J. Dynamic
variability in the test results. However, segregation or microstruc- mechanical properties of immiscible polymer systems with and without
compatibilizer. Polym Test 2015;44:168e76.
tural inhomogeneity would only require testing a larger number of [4] Jones DS. Dynamic mechanical analysis of polymeric systems of pharmaceu-
specimens for both tests as per the routine statistical procedures tical and biomedical significance. Int J Pharm 1999;179(2):167e78.
due to large standard deviation values for a batch of specimens. [5] Chandra R, Singh SP, Gupta K. Damping studies in fiber-reinforced composites
e a review. Compos Struct 1999;46(1):41e51.
Successful application of the transform developed in this work [6] Poveda RL, Achar S, Gupta N. Viscoelastic properties of carbon nanofiber
eliminates the need for conducting a large number of tensile test reinforced multiscale syntactic foam. Compos Part B Eng 2014;58:208e16.
specimens at a wide range of strain rates. The curve of modulus [7] Yang S, Taha-Tijerina J, Serrato-Diaz V, Hernandez K, Lozano K. Dynamic
mechanical and thermal analysis of aligned vapor grown carbon nanofiber
with respect to strain rate captured through this transform can be reinforced polyethylene. Compos Part B Eng 2007;38(2):228e35.
used in modeling and simulation studies. [8] Yang F, Ou Y, Yu Z. Polyamide 6/silica nanocomposites prepared by in situ
polymerization. J Appl Polym Sci 1998;69(2):355e61.
[9] Wang X, Hu Y, Song L, Yang H, Xing W, Lu H. In situ polymerization of gra-
5. Conclusions phene nanosheets and polyurethane with enhanced mechanical and thermal
properties. J Mater Chem 2011;21(12):4222e7.
In this work, integral transform of dynamic mechanical data was [10] Lin N, Chen G, Huang J, Dufresne A, Chang PR. Effects of polymer-grafted
natural nanocrystals on the structure and mechanical properties of poly(-
used to convert the frequency spectrum obtained from time tem-
lactic acid): a case of cellulose whisker-graft-polycaprolactone. J Appl Polym
perature superposition into a time domain relaxation function. Sci 2009;113(5):3417e25.
Using the relaxation function, the linear viscoelastic response to a [11] Kostka P, Holeczek K, Ho €hne R, Filippatos A, Modler N. Extension and appli-
cation of dynamic mechanical analysis for the estimation of spatial distribu-
given strain history can be obtained. Predictions of the response in
tion of material properties. Polym Test 2016;52:184e91.
constant strain rate uniaxial deformation were compared to tension [12] Jia Z, Amirkhizi AV, Nantasetphong W, Nemat-Nasser S. Experimentally-based
experiments and excellent agreement was observed. In addition, relaxation modulus of polyurea and its composites. Mech Time-Depend.
approximations to the Kramers-Kronig relations were used to Mater 2016;20(2):155e74.
[13] Nielsen C, Nemat-Nasser S. Crack healing in cross-ply composites observed by
obtain a well-matched estimate of the loss modulus master curve at dynamic mechanical analysis. J Mech Phys Solids 2015;76:193e207.
frequencies up to the transition. Comparison of the loss master [14] Song B, Chen WW, Lu WY. Mechanical characterization at intermediate strain
curve with the K-K prediction provides a better indication of the rates for rate effects on an epoxy syntactic foam. Int J Mech Sci 2007;49(12):
1336e43.
appropriateness of the fit function, as transitions near the edge of [15] Yoon, S.-h. and Siviour, C.R., Application of the virtual fields method to a
the experimental frequency range cause a small kink in the storage relaxation behaviour of rubbers. J Mech Phys Solids.
modulus curve versus peaks in the loss modulus. [16] Zhou Y, Akanda SR, Jeelani S, Lacy TE. Nonlinear constitutive equation for
vapor-grown carbon nanofiber-reinforced SC-15 epoxy at different strain rate.
Mater Sci Eng A 2007;465(1e2):238e46.
Acknowledgements [17] Jia Z, Hui D, Yuan G, Lair J, Lau K-t, Xu F. Mechanical properties of an epoxy-
based adhesive under high strain rate loadings at low temperature environ-
ment. Compos Part B Eng 2016;105:132e7.
This work is supported the Office of Naval Research Global [18] Hao X, Kaschta J, Schubert DW. Viscous and elastic properties of polylactide
N00014-10-1-0988. Authors acknowledge Dr. Keshav Prabhu, Mr. melts filled with silica particles: effect of particle size and concentration.
Puneeth and Mr. Praveen of Konkan Speciality Polyproducts Pvt. Compos Part B Eng 2016;89:44e53.
[19] Zhang B, Lin Y, Li S, Zhai D, Wu G. Quasi-static and high strain rates
Ltd., Mangalore, Karnataka, India for providing the injection
compressive behavior of aluminum matrix syntactic foams. Compos Part B
molding facility for sample fabrication. William Peng is thanked for Eng 2016;98:288e96.
useful mathematical discussions. Partial support from NYU Abu [20] Peroni L, Scapin M, Fichera C, Lehmhus D, Weise J, Baumeister J, et al.
Investigation of the mechanical behaviour of AISI 316L stainless steel syntactic
Dhabi Office of Research for Keerthana and NYUAD Dean of Engi-
foams at different strain-rates. Compos Part B Eng 2014;66:430e42.
neering for Mrityunjay is acknowledged. The views expressed in [21] Romero PA, Zheng SF, Cuitin ~ o AM. Modeling the dynamic response of visco-
this article are those of the authors, not of the funding agencies. elastic open-cell foams. J Mech Phys Solids 2008;56(5):1916e43.
[22] Luong DD, Pinisetty D, Gupta N. Compressive properties of closed-cell poly-
vinyl chloride foams at low and high strain rates: experimental investigation
References and critical review of state of the art. Compos Part B Eng 2013;44(1):403e16.
[23] Bharath Kumar BR, Singh AK, Doddamani M, Luong DD, Gupta N. Quasi-static
[1] Qiao J, Amirkhizi AV, Schaaf K, Nemat-Nasser S. Dynamic mechanical analysis and high strain rate compressive response of injection-molded cenosphere/
of fly ash filled polyurea elastomer. J Eng Mater Technol 2010;133(1):
34 S.E. Zeltmann et al. / Composites Part B 120 (2017) 27e34

HDPE syntactic foam. JOM 2016;68(7):1861e71. CAAI Trans Intell Technol 2016;1(1):1e3.
[24] Sarva SS, Deschanel S, Boyce MC, Chen W. Stressestrain behavior of a poly- [42] Labella M, Zeltmann SE, Shunmugasamy VC, Gupta N, Rohatgi PK. Mechanical
urea and a polyurethane from low to high strain rates. Polymer 2007;48(8): and thermal properties of fly ash/vinyl ester syntactic foams. Fuel 2014;121:
2208e13. 240e9.
[25] Luong DD, Shunmugasamy VC, Strbik III OM, Gupta N. High strain rate [43] Rohatgi P, Weiss D, Gupta N. Applications of fly ash in synthesizing low-cost
compressive behavior of polyurethane resin and polyurethane/Al2O3 hollow MMCs for automotive and other applications. JOM Journal of the Minerals,
sphere syntactic foams. J Compos 2014;2014(795984):1e10. Metals Mater Soc 2006;58(11):71e6.
[26] Zeltmann SE, Bharath Kumar BR, Doddamani M, Gupta N. Prediction of strain [44] Shunmugasamy VC, Gupta N, Nguyen NQ, Coelho PG. Strain rate dependence
rate sensitivity of high density polyethylene using integral transform of dy- of damage evolution in syntactic foams. Mater Sci Eng A 2010;527(23):
namic mechanical analysis data. Polymer 2016;101:1e6. 6166e77.
[27] Nian G, Shan Y, Xu Q, Qu S, Yang Q. Failure analysis of syntactic foams: a [45] Bharath Kumar BR, Doddamani M, Zeltmann SE, Gupta N, Ramesh MR,
computational model with cohesive law and XFEM. Compos Part B Eng Ramakrishna S. Data characterizing tensile behavior of censophere/HDPE
2016;89:18e26. syntactic foam. Data Brief 2016;6:933e41.
[28] Huang R, Li P. Elastic behaviour and failure mechanism in epoxy syntactic [46] Bharath Kumar BR, Doddamani M, Zeltmann SE, Gupta N, Ramesh MR,
foams: the effect of glass microballoon volume fractions. Compos Part B Eng Ramakrishna S. Processing of cenosphere/HDPE syntactic foams using an in-
2015;78:401e8. dustrial scale polymer injection molding machine. Mater Des 2016;92:
[29] Tagliavia G, Porfiri M, Gupta N. Influence of moisture absorption on flexural 414e23.
properties of syntactic foams. Compos Part B Eng 2012;43(2):115e23. [47] Bharath Kumar BR, Zeltmann SE, Doddamani M, Gupta N, Uzma, Gurupadu S,
[30] Huang R, Li P, Wang Z, Liu T. X-ray microtomographic characterization and et al. Effect of cenosphere surface treatment and blending method on the
quantification of the strain rate dependent failure mechanism in cenosphere tensile properties of thermoplastic matrix syntactic foams. J Appl Polym Sci
epoxy syntactic foams. Adv Eng Mater 2016;18(9):1550e5. 2016;133(35):1e11.
[31] Buckley CP, Dooling PJ, Harding J, Ruiz C. Deformation of thermosetting resins [48] Bharath Kumar BR, Doddamani M, Zeltmann SE, Gupta N, Uzma, Gurupadu S,
at impact rates of strain. Part 2: constitutive model with rejuvenation. J Mech et al. Effect of particle surface treatment and blending method on flexural
Phys Solids 2004;52(10):2355e77. properties of injection-molded cenosphere/HDPE syntactic foams. J Mater Sci
[32] Porfiri M, Gupta N. Effect of volume fraction and wall thickness on the elastic 2016;51(8):3793e805.
properties of hollow particle filled composites. Compos Part B Eng [49] Khanna YP, Turi EA, Taylor TJ, Vickroy VV, Abbott RF. Dynamic mechanical
2009;40(2):166e73. relaxations in polyethylene. Macromolecules 1985;18(6):1302e9.
[33] Bardella L, Genna F. On the elastic behavior of syntactic foams. Int J Solids [50] Williams ML, Landel RF, Ferry JD. The temperature dependence of relaxation
Struct 2001;38(40e41):7235e60. mechanisms in amorphous polymers and other glass-forming liquids. J Am
[34] Bardella L, Sfreddo A, Ventura C, Porfiri M, Gupta N. A critical evaluation of Chem Soc 1955;77(14):3701e7.
micromechanical models for syntactic foams. Mech Mater 2012;50:53e69. [51] Christensen RM. Theory of viscoelasticity: an intoduction. New York: Aca-
[35] Nguyen VT, Hwu C. Holes, cracks, or inclusions in two-dimensional linear demic Press; 1982.
anisotropic viscoelastic solids. Compos Part B Eng 2017;117:111e23. [52] Booij HC, Thoone GPJM. Generalization of Kramers-Kronig transforms and
[36] Fedorov VA, Barkanov EN. Homogenisation of viscoelastic damping in unidi- some approximations of relations between viscoelastic quantities. Rheol Acta
rectional composites under longitudinal shear. Compos Part B Eng 2017;113: 1982;21(1):15e24.
72e9. [53] Brather A. Numerisch einfache Beziehungen zwischen Verlust- und Spei-
[37] Liu H, Dong M, Huang W, Gao J, Dai K, Guo J, et al. Lightweight conductive cherkomponente des dynamischen Schermoduls und der dynamischen
graphene/thermoplastic polyurethane foams with ultrahigh compressibility Nachgiebigkeit. Rheol Acta 1978;17(4):325e41.
for piezoresistive sensing. J Mater Chem C 2017;5(1):73e83. [54] Schapery RA, Park SW. Methods of interconversion between linear visco-
[38] Liu H, Gao J, Huang W, Dai K, Zheng G, Liu C, et al. Electrically conductive elastic material functions. Part IIdan approximate analytical method. Int J
strain sensing polyurethane nanocomposites with synergistic carbon nano- Solids Struct 1999;36(11):1677e99.
tubes and graphene bifillers. Nanoscale 2016;8(26):12977e89. [55] Shunmugasamy VC, Pinisetty D, Gupta N. Viscoelastic properties of hollow
[39] Petrie KG, Osazuwa O, Docoslis A, Kontopoulou M. Controlling MWCNT par- glass particle filled vinyl ester matrix syntactic foams: effect of temperature
titioning and electrical conductivity in melt compounded polypropylene/ and loading frequency. J Mater Sci 2013;48(4):1685e701.
poly(ethylene-co-octene) blends. Polymer 2017;114:231e41. [56] Ferry JD. Viscoelastic properties of polymers. New York: John Wiley & Sons,
[40] Zhang X, Gao H, Guo M, Li G, Liu Y, Li D. A study on key technologies of un- Inc; 1961.
manned driving. CAAI Trans Intell Technol 2016;1(1):4e13. [57] Christensen RM. Restrictions upon viscoelastic relaxation functions and
[41] Alippi C. A unique timely moment for embedding intelligence in applications. complex moduli. Trans Soc Rheol. 1972;16(4):603e14.