Polymer Testing 71 (2018) 65–71

Contents lists available at ScienceDirect

Polymer Testing
journal homepage: www.elsevier.com/locate/polytest

Test Method

Thermal expansion/shrinkage measurement of battery separators using a T

dynamic mechanical analyzer
Shutian Yana, Jie Dengb, Chulheung Baeb, Xinran Xiaoa,∗
a
Mechanical Engineering, Michigan State University, USA
b
Department of Energy Storage Research, Ford Motor Company, USA

A R T I C LE I N FO A B S T R A C T

Keywords: Polymer separators for lithium ion batteries are thin, porous membranes of 20–30 μm thickness. At elevated
CTE measurement temperatures, some separators can shrink considerably. To predict the deformation and stresses in the separator
Polymer thin films in battery cells, it is necessary to measure the expansion/shrinkage property of a separator as expressed by the
Thermal expansion/shrinkage coefficient of thermal expansion (CTE). The reported CTE measurement techniques for thin films have their
Dynamic mechanical analyzer (DMA)
limitations. This paper examines the use of a dynamic mechanical analyzer (DMA) for the measurement of the
thermal expansion/shrinkage property of polymer separators. The effect of the dimensional change of the testing
fixture was considered. The influences of the preloading levels and heating rates were investigated. The mea-
surements were carried out for three common types of polymer based separator. The CTE as a function of
temperature was determined from the DMA data. DMA offers continuous measurements in an automatic fashion,
which is an efficient and convenient method to characterize the thermal expansion/shrinkage behavior of thin
polymer films.

1. Introduction surface in an oven with an optical window and using non-contact op-
tical methods such as interferometry [4] or digital image correlation [5]
In lithium ion batteries (LIBs), the separator provides electrical in- to measure the dimensional change. For metal foils, the CTE can be
sulation between opposite electrodes while allowing ionic conductivity. measured using a bi-layer configuration with a substrate of known CTE
The commonly used LIB separators are polymer based, thin, porous [6]. The optical measurement requires special equipment. The bi-layer
membranes of 20–30 μm thickness [1]. configuration is not suitable for polymer thin films because the stiffness
It is known that, with increasing temperatures, polymer separators of these materials are too low to generate needed deformation with
first expand and then start to shrink before final fracture. The amount of commonly used substrates.
shrinkage can be significant for some separators [1]. In a constrained The thin film CTE measurement has also been attempted for a
condition, a larger shrinkage will induce higher stresses in the separator clamped sample. El-Tonsy [7] presented a system that measures the
which may cause earlier failure. As the mechanical integrity of the length change of a clamped polymer film sample under a small hanging
separator is critical to the safety of LIBs, it is desired to predict the weight. This system has some similarity to the thermal mechanical
stresses and strains in the separator under extreme conditions such as analyzer (TMA) and dynamic mechanical analyzer (DMA) commonly
thermal runaway. In such predictions, the coefficient of thermal ex- used in polymer characterization. TMA [1,8] and DMA [9–12] have
pansion (CTE) is one of the required inputs. Because of that, it is critical been used widely in the characterization of battery separators. Baldwin
to have a technique that can measure the thermal expansion/shrinkage et al. [9] characterized the shrinkage of battery separators under iso-
of these films continuously and accurately. thermal conditions using a DMA under tensile mode [9]. These works
The standard methods determine the CTE of a material through indicate the potential of TMA and DMA in thin film CTE measurement.
measuring the volume or lengthwise change of a free standing sample The current study explores this possibility.
[2,3]. These techniques are not suitable for materials in the form of thin This paper presents a systematical investigation on the use of a DMA
films. under tensile mode for the thermal expansion/shrinkage measurement
For thin films, several CTE measurement techniques have been re- of battery separators. Factors such as the calibration procedure, the
ported. A common method is laying an unconstrained sample on a flat influence of the tensile load and the temperature ramp rate were

∗
Corresponding author.
E-mail address: [email protected] (X. Xiao).

https://doi.org/10.1016/j.polymertesting.2018.08.028
Received 17 July 2018; Accepted 26 August 2018
Available online 27 August 2018
0142-9418/ © 2018 Elsevier Ltd. All rights reserved.
S. Yan et al. Polymer Testing 71 (2018) 65–71

Fig. 1. Surface microstructure of Celgard® 2400 [1]. The two in-plane material
directions are referred to as the machine direction (MD) and the transverse
direction (TD).

examined. Using the established method, the measurements were car-

ried out for three common types of LIB separator from ambient to the Fig. 3. The MD sample and TD sample.
maximum temperature that the measurement can be performed.
Celgard® 2400 was tested first to investigate the effect of the test
2. Experimental parameters such as the level of the tensile load and the temperature
ramp rate. The experiments were performed at multiple heating rates
2.1. Materials and force levels. A preload of 0.01 N and 0.001 N was selected for the
MD and TD samples, respectively, and a heating rate of 3 °C/min was
Three representative separators were investigated in this work. selected for both samples. The results of these investigations will be
They were: (1) Celgard® 2400, a monolayer PP; (2) Celgard® 2325, a presented in Section 3.1.
trilayer PP/PE/PP; and (3) Celgard® Q20S1HX, a ceramic coated tri-
layer PP/PE/PP. The film thickness was 25 μm for 2400 and 2325, and
20 μm for Q20S1HX. 2.3. Calibration
Fig. 1 shows the surface microstructure of a typical PP separator [1].
As seen, the pores are a fraction of micrometer long and tens to hundred Prior to the CTE measurement, the DMA needs to be calibrated
nanometer wide, with splits. The fibrous like structures that separate using a material with known CTE. In this work, the calibration was
the splits are amorphous. The thick regions are semi-crystalline. This performed with Al thin foil of 25 μm thickness with a purtiy of 99%.
unique microstructure is formed through a multi-step process including The sample dimensions and their mounting in the DMA were the same
melt-extrusion, annealing, stretching and heat fixation [1]. The re- as that described in 2.2. The measurement was performed under a small
sulting material is highly anisotropic. The separator is usually treated as tensile loading of 0.01 N over the temperature range from ambient to
an orthotropic material with the two in-plane material directions re- 220 °C with a heating rate of 3 °C/min.
ferred to as the machine direction (MD) and the transverse direction Fig. 4a presents the measured grip displacement with increasing
(TD). temperature obtained with the Al samples. The slopes of the curves
appear to be negative, implying a negative CTE, i.e. shrinkage instead
2.2. Measurement of expansion. For a sample with a known CTE, the change in sample
gage length with temperature can be calculated. For pure Al, the CTE is
The experiment was performed with a TA Q800 DMA under tensile reported as 24.5 × 10−6/⁰C over the range of 20–200 °C [13]. The
mode. DMA is a common mechanical testing equipment to characterize displacement predicted this way is presented by the dashed line in
the viscoelastic behavior of materials. Fig. 2 shows the experimental Fig. 4a. The discrepency between the predicted curve and the mea-
set-up. The sample was gripped at both ends with tension clamps. Then, suremed curves indicates that the measured grip displacement included
the cell was closed and the temperature ramped. During the test, a the contribution from other components in addition to that of the
constant force was maintained and the lengthwise variation of the sample.
sample gage length was measured. In TA DMA systems, the grip displacement is measured by a sensor
The samples, in the form of long strips with a length of 45 mm and located on the drive shaft [14]. The measured displacement includes
width of 6 mm, were cut using a razor blade. The sample width was the dimensional change of the sample, the grip, and the drive shaft
then measured, on a flat surface, using a caliper. Samples were prepared relative to a stationary component [14], in addition to that of the
both along MD and TD, as shown in Fig. 3. In DMA measurement, the sample. The standard calibration procedure for the DMA uses a steel
testing length of the samples was about 15 mm. sample and the calibration is performed at ambient temperature. Based
on the calibration, the contribution of the testing fixture compliance is
automatically excluded from the DMA output. This is sufficientlly ac-
curate for mechanical testing, as verified by comparing the DMA data
with the digital image correlation measurement [15]. In measurements
over a temperature range, all components, including the stationary
component, will experience thermal expansion/shrinkage. In mechan-
ical testing, the strain due to the dimensional change of the testing
fixture is usally negligible compared to the deformation experienced by
the sample. For CTE measuremnts, additional calibration is needed. The
calibration should be performed with the same experimental set-up and
under the same testing condition.
To determine the true dimensional change of the sample, the con-
tribution of the testing fixture must be known in advance. This amount
Fig. 2. (A) A sample is mounted in the tensile clamp in DMA Q800. (b) The is designated as Δlcalibration, determined by the difference between the
furnace will then be closed for temperature ramp. predicted dimensional change Δlpredicted over the gage length l of the

66
S. Yan et al. Polymer Testing 71 (2018) 65–71

Fig. 4. (A) The measured Al sample length change and the predicted sample length expansion. (b) The calibration curve. (c) The measured Cu sample length change
before and after correction, and the predicted length change.

sample used in calibration, and the measured dimensional change 3. Results and discussions
Δlmeasrued by the DMA [16].
3.1. Expansion/shrinkage measurement
Δlcalibration = Δlpredicted − Δlmeasured (1)
3.1.1. Effect of tensile load
The predicted gage length change of the calibration sample is To keep the sample straight during the measurement, a small tensile
computed as load is needed. The load should be as small as possible to avoid the
influence of creep strain. On the other hand, it is difficult to maintain a
Δlpredicted = CTE × l 0 × (T − T0) (2) very small load accurately. The effect of tensile load on the measure-
ment was investigated using Celgard® 2400 at a ramp rate of 3 °C/min.
where l0 is the initial sample length. These measurements were repeated two times at each load level. The
Fig. 4b presents the Δlcalibration for the current CTE measurement set- repeatability of the measurement was very good. The averaged curves
up obtained with the Al samples. The true dimensional change of the are presented in Fig. 5.
sample Δlcorrected can be determined by the following correction Fig. 5 shows that the load level had a significant effect on the
measurement. For the MD samples (Fig. 5a), the measurements were
Δlcorrected = Δlcalibration + Δlmeasured (3)
performed under three load levels, 0.01 N, 0.02 N and 0.45 N, corre-
The established calibration was verified by testing a Cu foil of 50 μm sponding to stress levels of 0.067 MPa, 0.13 MPa and 3 MPa. At 3 MPa,
thickness with a purtiy of 99.9%. Fig. 4c presents the Δlmeasured for two the sample did not shrink as the creep strain had exceeded the thermal
Cu samples and Δlcorrected calculated using Eq. (3). Then, the engineering strain. Under 0.067 MPa and 0.13 MPa, the measured expansion/
strain was calculated as ε = Δlcorrected/l0. The CTE is defined as a change shrinkage was almost identical, indicating that a stress level of
in strain in response to a change in temperature 0.067 MPa is sufficiently low for the MD samples. For the TD (Fig. 5b),
the measurement was made under 0.007 MPa, 0.014 MPa, 0.067 MPa
dεT and 0.13 MPa. Unlike the MD, the expansion/shrinkage curve and the
CTE = peak/valley values of the TD samples continue to change with the
dT (4)
stress. Finally, a stress level of 0.007 MPa was selected as this corre-
The CTE for the Cu foil determined in this way was 17.1 × 10−6/⁰C, sponds to the minimum force of 0.001 N which the DMA can maintain.
which is within 1.2% of the reported CTE value of 17.3 × 10−6/⁰C for
pure Cu over the range of 20–200 °C [17].
3.1.2. Effect of heating rate
The effect of the temperature ramp rate on the expansion/shrinkage
behavior was investigated using Celgard® 2400 TD samples. The results

67
S. Yan et al. Polymer Testing 71 (2018) 65–71

Fig. 5. The measured expansion/shrinkage strain under different stress levels for Celgard® 2400 (a) MD and (b) TD samples.

Fig. 6. The effect of the temperature ramp rate on Celgard® 2400 TD samples. (a) The measured thermal expansion/shrinkage, and (b) TShrink, the temperature when
the sample starts to shrink.

Table 1
Test conditions for MD and TD samples.
Testing conditions MD samples TD samples

Tensile stress (MPa) 0.067 0.007

Temperature ramp rate (⁰C/min) 3

are shown in Fig. 6. For 3 °C/min and 5 °C/min, the amount of the
expansion below 120 °C and shrinkage between 140 °C and 160 °C were
about the same. At the two higher rates, however, these amounts de-
creased. This is probably due to the thermal lag between the tem-
perature measured by the thermocouple of the DMA and the sample
[18,19]. A faster temperature ramp rate causes a larger thermal lag
[18]. TShrink, the temperature when the sample starts to shrink, as
marked on Fig. 6a, may be another indication of the thermal lag effect.
As shown in Fig. 6b, TShrink increased sharply when the rate increased
from 3 °C/min to 5 °C/min. Therefore, a 3 °C/min ramp rate was se- Fig. 7. Celgard® 2400 MD samples after isothermal conditioning for 30 min in
lected. Table 1 lists the testing conditions for further investigations. an oven. From left to right: 25 °C, 130 °C, 155 °C, 160 °C, 170 °C and 180 °C.

3.1.3. Comparison with oven isothermal test temperature. The length of the sample before and after was measured
The shrinkage of separators is often measured with an oven iso- with a caliper. The oven tested samples are shown in Fig. 7. The change
thermal test [9,20–23]. In this method, the sample is heated in an oven in the sample length and appearance is noticeable. The sample tested at
to the desired isothermal temperature for a period of 30 min to 1 h. The 170 °C became transparent and curled. At 180 °C, the sample had
dimensions of the sample are measured immediately after it has been melted as it became attached to the glass slide.
removed from the oven. Table 2 compares the shrinkage measured using the isothermal
In this study, oven isothermal tests were performed with Celgard® method and by the DMA. The agreement between the two methods is
2400 MD samples of 55 × 6 mm at 130 °C, 155 °C, 160 °C, 170 °C and rather good except at 180 °C. The glass slide apparently had restricted
180 °C. The sample was placed on a clean and smooth microscope slide the shrinkage of the sample during the isothermal conditioning.
and heated for 30 min in a conventional oven at the desired Comparing the two methods, the isothermal test is relatively simple

68
S. Yan et al. Polymer Testing 71 (2018) 65–71

Table 2 Table 3
Celgard® 2400 MD shrinkage strains measured using oven isothermal test and Shrinkage from manufacturer's data sheet measured after being kept at the
DMA temperature ramp test. temperature for 1 h and from DMA measurement at 3 °C/min.
Testing temperature (oC) 130 155 160 170 180 Celgard® 2400 Celgard® 2325 Celgard® Q20S1HX

Shrinkage measured by oven −9.1 −37.3 −40.0 −43.6 −50.9 Thickness 25 μm 25 μm 20 μm

isothermal test (%) Shrinkage Datasheet DMA Datasheet DMA Datasheet DMA
Shrinkage measured using DMA (%) −6.6 −31.8 −38.7 −44.4 −73.6 MD, 90 °C < 5.0% 0.3% < 5.0% 0.4%
TD, 90 °C 0.0% 0.0% 0.0% 0.0%
MD, 105 °C 2.21% 1.2%
TD, 105 °C 0.0% 0.0%
and requires only an oven. However, the measurements can be per-
MD, 130 °C < 2% 0.5%
formed only at selected temperatures. The DMA provides a continuous TD, 130 °C < 1% 0.0%
measurement and the measurement is automatically recorded. Overall, MD, 150 °C < 4% 0.8%
the DMA is less time consuming and more accurate than the isothermal TD, 150 °C < 2% 1.7%
method.
0% indicates there was no shrinkage.

3.2. Expansion/shrinkage of three separators

the shrinkage reported in the datasheet.
Fig. 8a compares the typical expansion/shrinkage behaviors of three Fig. 8d compares the shrinkage behavior in the TD. The tendency of
types of separators in the MD. In general, Celgard® 2400 and 2325 show shrinkage in the TD for these three separators is opposite to that in the
similar behavior, whereas Q20S1HX is distinctively different. For a MD. The TD shrinkage value is Q20S1HX > 2400 > 2325, although
better comparison, Fig. 8b and c provide zoom-in views for 20–80 °C their differences were smaller than that in the MD. Q20S1HX started to
and small strain regions, respectively. As shown in Fig. 8b, 2400 and shrink at about 130 °C. The shrinkages was about 0% at 130 °C and
2325 expanded slightly up to 60 °C, and then started to shrink. At 1.7% at 150 °C. These values compared well with the values in data-
180 °C, the shrinkage was about 75%. On the other hand, Q20S1HX sheet. The datasheet reports no shrinkage for 2400 up to 90 °C and for
shrank steadily with increasing temperature. Its maximum shrinkage 2325 up to 105 °C. These are confirmed by the DMA data in this work.
was less than 1%. The sample ruptured at 220 °C. Both 2400 and 2325 displayed expansion behavior up to about 130 °C.
Table 3 compares the shrinkage data provided in manufacturer's The significant shrinkage of Celgard® 2400 and 2325 in the MD is
datasheet and the results from the current DMA measurement. The due to their microstructures showed in Fig. 1. The pores and fibers were
shrinkage in the datasheet was indicated as 1 h at the temperature. introduced by uniaxially stretching the film along the MD [1]. As a
For 2400 and 2325 MD, the shrinkage reported in the datasheet result, the polymer chains are disentangled and elongated. When the
is < 5% at 90 °C. The value measured using the DMA was about 0.3%. temperature was increased, the oriented amorphous phase shrank be-
At 105 °C, the shrinkage for 2325 is 2.21% in the datasheet. The DMA fore the fusion of crystals. Then, the crystals melted, and the chains
value was about 1.2%. For Q20S1HX MD, the DMA data were within became disoriented. This lead to the large thermal shrinkage along the

Fig. 8. Comparison of expansion/shrinkage behaviors of three types of separators, (a) the MD samples, (b) the MD samples from 20 to 80 °C, (c) the MD samples for
small strain range, and (d) the TD samples.

69
S. Yan et al. Polymer Testing 71 (2018) 65–71

Table 5
CTE for Celgard® 2400, 2325 and Q20S1HX.
Material Temp (°C) CTE (x10−6/°C)

2400 MD 30–65 44
65–175 − 3.45 × 10−13T7.6262
2400 TD 30–130 1.4 T + 26
130–140 0
140–160 −88 T + 12044
2325 MD 30–60 30
60–175 − 5.1 × 10−13T7.5641
2325 TD 30–120 1.6 T + 25
120–130 0
130–150 −124 T + 16087
Q20S1HX MD 30–160 −1.2 T + 39
160–200 29
Q20S1HX TD 30–110 T-1
110–135 0
135–175 28T-5479
175–185 540T-101850

provides a summary for the temperature when necking/rupture occurs.

For the TD samples, these values were 15 °C–35 °C lower than the MD
samples for all three separators. Table 4 compares the temperatures
when the samples started to shrink and neck/fracture for three se-
parators in the MD and TD.
In summary, the shrinkage data obtained by the DMA measurement
Fig. 9. Samples after the experiment: (a) Celgard® 2400 MD (2325 MD is si- yielded a similar trend as that in the datasheet. At 90 °C, Celgard® 2400
milar), (b) Celgard® Q20S1HX MD, (c) Celgard® 2400 TD (2325 TD is similar); and 2325 have similar shrinkage behavior and the shrinkage is much
and (d) Celgard® Q20S1HX TD sample. higher in the MD. Q20S1HX shows more isotropic shrinkage behavior.

Table 4 3.3. Coefficient of thermal expansion (CTE)

Temperatures when the samples started to shrink and neck/fracture.
Celgard® 2400 Celgard® 2325 Celgard® Q20S1HX As described by Eq. (4), the CTE is the derivative of the strain-
temperature curve. Since the curves in Fig. 8 were all nonlinear, each
MD TD MD TD MD TD curve was divided into several segments and the CTEs were determined
T At Shrink (oC) ∼65 ∼133 ∼60 ∼125 ∼30 ∼115
in a piecewise fashion. This procedure is illustrated in Fig. 10 using the
At Neck/fracture (oC) > 180 ∼165 > 180 ∼150 ∼220 ∼185 Celgard® 2400 MD sample as an example.
The strain-temperature curve of 2400 MD sample (Fig. 8a) can be
divided into two segments. The first segment is 25 < T < 65 °C,
MD [8]. The shrinkage of Celgard® 2400 and 2325 TD samples was less where the sample expanded, and the strain increased linearly. This
than 2% as the TD was not stretched during manufacturing. This ani- segment is replotted in Fig. 10a. The curve was fitted using a linear
sotropic shrinkage would be restricted with a ceramic coated layer, as function as shown by the blue dashed curve. The slope of the curve was
in the case of Q20S1HX. constant and is given by the black line with a scale shown by the sec-
Fig. 9 compares the samples after the experiment. For 2400 and ondary vertical axis.
2325 MD samples, the measurement terminated when the gage length The second segment is 65 < T < 180 °C. It was replotted in
reached the minimum value allowed by the tension clamp. The length Fig. 10b. In this segment, the sample shrank and the strain decreased
of the tested samples shrank from 15 mm to about 5 mm without no- nonlinearly. The curve was fitted using a power function. The deriva-
ticeable change in the width (Fig. 9a). Their rupture temperature would tive of the curve provides the CTE function. Fig. 10c presents the
be higher than 180 °C. On the other hand, the measurement for combined strain and CTE curves. As shown, the piecewise method
Q20S1HX lasted until 220 °C when the separator actually ruptured yields a smooth relation for the CTE over the entire temperature range.
(Fig. 9b). This corresponds to a sharp increase in strain in Fig. 8a. At the This relation can be used conveniently in finite element analysis.
end of the tests, the Celgard® 2400 and 2325 TD samples necked Following the same procedure, the CTE was determined for Celgard®
(Fig. 9c) while Q20S1HX TD samples ruptured (Fig. 9d). Table 4 2400 TD, 2325 and Q20S1HX MD and TD. These results are

Fig. 10. Determine the CTE of Celgard® 2400 MD. The shrinkage curve was divided into two segments, curve fitting was performed for each segment, and the slope of
the curve for the segment gave the CTE. (a) The first segment, (b) the second segment, and (c) the combined curve.

70
S. Yan et al. Polymer Testing 71 (2018) 65–71

summarized in Table 5. interferometry, Strain 42 (2006) 197–205.

[5] P. Bing, X. Hui-min, H. Tao, A. Asundi, Measurement of coefficient of thermal ex-
pansion of films using digital image correlation method, Polym. Test. 28 (2009)
4. Conclusions 75–83.
[6] W. Fang, C. Lo, On the thermal expansion coefficients of thin films, Sensor.
The use of a dynamic mechanical analyzer (DMA) in the measure- Actuator. 84 (2000) 310–314.
[7] M.M. El-Tonsy, Automatic measurement of the absolute CTE of thin polymer
ment of thermal expansion/shrinkage of battery separators was ex- samples: I—effect of multiple processing on thermal expansion of polypropylene
plored. For this measurement, it is important to perform a calibration films, Polym. Test. 23 (2004) 355–360.
test to determine the dimensional change of the testing fixture and to [8] B. Xiong, R. Chen, F. Zeng, J. Kang, Y. Men, Thermal shrinkage and microscopic
shutdown mechanism of polypropylene separator for lithium-ion battery: in-situ
compensate for this contribution in the DMA measurement. The pre- ultra-small angle X-ray scattering study, J. Membr. Sci. 545 (2018) 213–220.
loading level and heating rate were found to affect the measurement. [9] R.S. Baldwin, W.R. Bennet, E.K. Wong, M.R. Lewton, M.K. Harris, Battery Separator
The shrinkage data obtained from the DMA agreed well with the values Characterization and Evaluation Procedures for NASA's Advanced Lithium-ion
Batteries, NASA/TM-2010-216099.
obtained with an unconstrained sample under iso-thermal conditions.
[10] C.T. Love, Thermomechanical analysis and durability of commercial micro-porous
From the DMA results, the CTE relationships for three typical separators polymer Li-ion battery separators, J. Power Sources 196 (2011) 2905–2912.
have been established. The DMA offers a continuous measurement in an [11] A. Sheidaei, X. Xiao, X. Huang, J. Hitt, Mechanical behavior of a battery separator
automatic fashion. It is an efficient and convenient method to char- in electrolyte solutions, J. Power Sources 196 (2011) 8728–8734.
[12] J. Xu, L. Wang, J. Guan, S. Yin, Coupled effect of strain rate and solvent on dynamic
acterize the thermal expansion/shrinkage behavior of thin films. mechanical behaviors of separators in lithium ion batteries, Mater. Des. 95 (2016)
Celgard® 2400 and 2325 have similar shrinkage behavior. Their 319–328.
shrinkage in the MD is much higher than that in the TD. Q20S1HX [13] P. Hidnert, H.S. Krider, Thermal expansion of aluminum and some aluminum al-
loys, J. Res. Natl. Bur. Stand. 48 (1952) 209–220.
shows more isotropic shrinkage behavior. For all three separators, the [14] TA Instruments, Dynamic Mechanical Analyzers: How Do They Work? http://www.
necking/fracture temperatures of the TD samples were 15 °C–35 °C tainstruments.com/pdf/literature/TA229.pdf (accessed June 18, 2018).
lower than the MD samples. [15] S. Yan, J. Deng, C. Bae, Y. He, A. Asta, X. Xiao, In-plane Orthotropic Property
Characterization of a Polymeric Battery Separator, submitted.
[16] PerkinElmer, Inc., Expansion Coefficient of Polymeric Materials Using DMA 8000,
Acknowledgement (2007) Waltham.
[17] S.J. Bennett, The thermal expansion of copper between 300 and 700 K, J. Phys.
Appl. Phys. 11 (1978) 777–780.
This work is supported by the Ford-MSU Alliance Program. [18] TA Instruments, Dynamic Mechanical Analyzer, Operator's Manual, (2002) New
Castle.
Appendix A. Supplementary data [19] W. Sun, A.P. Vassilopoulos, T. Keller, Effect of thermal lag on glass transition
temperature of polymers measured by DMA, Int. J. Adhesion Adhes. 52 (2014)
31–39.
Supplementary data related to this article can be found at https:// [20] C. Shi, J. Dai, X. Shen, L. Peng, C. Li, X. Wang, P. Zhang, J. Zhao, A high-tem-
doi.org/10.1016/j.polymertesting.2018.08.028. perature stable ceramic-coated separator prepared with polyimide binder/Al2O3
particles for lithium-ion batteries, J. Membr. Sci. 517 (2016) 91–99.
[21] J. Song, M.H. Ryou, B. Son, J.N. Lee, D.J. Lee, Y.M. Lee, J.W. Choi, J.K. Park, Co-
References polyimide-coated polyethylene separators for enhanced thermal stability of lithium
ion batteries, Electrochim. Acta 85 (2012) 524–530.
[1] P. Arora, Z.M. Zhang, Battery separators, Chem. Rev. 104 (2004) 4419–4462. [22] J. Li, Q. Zhong, Y. Yao, S. Bi, T. Zhou, X. Guo, M. Wu, T. Feng, R. Xiang,
[2] ASTM D696 Test Method for Coefficient of Linear Thermal Expansion of Plastics Electrochemical performance and thermal stability of the electrospun PTFE nano-
between -30°C and 30°C with a Vitreous Silica Dilatometer, ASTM International. fiber separator for lithium-ion batteries, J. Appl. Polym. Sci. 135 (2018) 46508.
[3] ASTM E831-06 Standard Test Method for Linear Thermal Expansion of Solid [23] Y. Li, H. Pu, Y. Wei, Polypropylene/polyethylene multilayer separators with en-
Materials by Thermo-mechanical Analysis, ASTM International. hanced thermal stability for lithium-ion battery via multilayer coextrusion,
[4] C. Dudescu, J. Naumann, M. Stockmann, S. Nebel, Characterization of thermal Electrochim. Acta 264 (2018) 140–149.
expansion coefficient of anisotropic materials by electronic speckle pattern

71