- Email: [email protected]
Enhanced thermal conductivity of photopolymerizable composites using surface modified hexagonal boron nitride fillers
Accepted Manuscript Enhanced thermal conductivity of photopolymerizable composites using surface modified hexagonal boron nitride fillers Nir Goldin, Hanna Dodiuk, Dan Lewitus PII:
S0266-3538(17)30528-6
DOI:
10.1016/j.compscitech.2017.09.001
Reference:
CSTE 6890
To appear in:
Composites Science and Technology
Received Date: 11 March 2017 Revised Date:
28 August 2017
Accepted Date: 2 September 2017
Please cite this article as: Goldin N, Dodiuk H, Lewitus D, Enhanced thermal conductivity of photopolymerizable composites using surface modified hexagonal boron nitride fillers, Composites Science and Technology (2017), doi: 10.1016/j.compscitech.2017.09.001. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Enhanced thermal conductivity of photopolymerizable composites using surface modified hexagonal boron nitride fillers. Nir Goldina, Hanna Dodiuka, and Dan Lewitusa*
a
RI PT
Author and affiliations: Department of Polymers and Plastics Engineering, Shenkar - Engineering. Art. Design,
Ramat-Gan, Israel
SC
*Corresponding author
M AN U
Abstract:
The interest in photocurable polymers has risen greatly in the past few years, in part due to the additive manufacturing revolution. Still, their widespread use is hindered by various inherent physical properties, such as thermal insulation. This work is aimed
TE D
towards the development of photopolymerizable polymer composites that are thermally conductive, while maintaining their photocurable characteristics. We developed photocurable acrylic-based photopolymer composites with hexagonal boron nitride
EP
(hBN) using the following method: pristine hBN underwent two chemical surface modifications, was added to the monomers, and the mixture then underwent radiation
AC C
curing. The success of the synthesis was verified in two ways: FTIR and XPS analyses in which the formation of carbonyl groups at the surface of the treated hBN was tracked, as well as tracking the increase in the homogeneity of the pre-polymerized solution. The addition of a reaction accelerator (o-benzoic sulfimide) to the photoinitiator system allowed for an increase of conversion percentage from ~60% to ~95%, even with high hBN loadings. Thermal conductivity (measured via modulated differential scanning calorimetry (MDSC)) increased with respect to hBN content by more than 300% when using 35wt% hBN. Young's modulus and viscosity increased with hBN content, while
ACCEPTED MANUSCRIPT coefficient of thermal expansion (CTE) decreased. We have thus developed a photocurable monomer system that is thermally conductive and applicable in various radiation curing processes.
RI PT
Key words: A. Particle-reinforced composites. B. Thermal properties. D. Differential scanning calorimetry (DSC). C. Modeling. Radiation curing
SC
Main Text:
M AN U
1. Introduction
Efficient thermal dissipation has become imperative to the advancement of many fields, including the electronics [1–3] and mold making [4] industries. Traditional thermally conductive materials, such as metals and ceramics, have several disadvantages that make them unfavorable for many applications. They are both heavy,
TE D
and their molding/shaping processes are time-consuming and costly. Moreover, these materials' inability to achieve extremely complex shapes often makes them impractical
EP
[5].
Radiation-cured polymeric materials have many applications, some of which would
AC C
greatly benefit from enhanced thermal conductivity. In the coating industry, for example, a decrease in the difference of thermal expansion between the substrates and coatings may allow for better adhesion [6–8]. Additive manufacturing (commonly known as "3D printing") may allow for inexpensive production of finely detailed items that had been traditionally made of metals or ceramics, with limited detail [9]. Similarly, the field of electronic circuitry [10–13] uses photolithography quite heavily, and could certainly benefit from better heat dissipation. Plastics are inherent thermal insulators, as most have a thermal conductivity value of 0.19 – 0.35 W/mK [14]. Thermal 2
ACCEPTED MANUSCRIPT conductivity enhancement is not a new concept [1,15–17], and thermally conductive plastics have been offered commercially for quite some time. These plastics are mostly achieved through the incorporation of high loadings of thermally conductive fillers into a polymer [18–21]. While this addition allows for the resulting composite to be
suffers from decreased mechanical properties [22].
RI PT
thermally conductive, it also yields a much heavier, more viscous product that often
Successful fabrication with methods using irradiation curing of photosensitive
SC
monomers (e.g. stereolithography 3D printing) places much significance on the
M AN U
viscosity of the polymer, its ability to photo-cure effectively, and the stability of the additive-monomer solution [23,24]. When incorporating thermally conductive fillers, however, the filler "shields" the polymer as the thickness increases, thus reducing the depth of cure [25,26]. The phenomenon is even more pronounced at high loadings. This factor, together with the oxygen sensitive radiation curing mechanism [27–30], may be
TE D
the reason that most thermal conductivity studies of thermoset composites have been conducted using thermally cured thermosets, like epoxy, polybenzoxazine, polyimide,
EP
etc. [1,15,16,18,31–35].
There has therefore been much focus on reducing the loading of the thermally
AC C
conductive fillers as much as possible. One method with which this is achieved uses compatibilized fillers [17,18,36]. Chemical compatibilization is prevalent for both carbon-based and ceramic fillers. In hexagonal boron nitride (hBN), surface modification is a rather popular route, as there are hydroxyl and amine groups on the surface of the particle [37,38]. Yu et al. [38], for example, compatibilized hBN using a hyperbranched polymer to slightly increase the thermal conductivity and mechanical properties of epoxy-based composites. Likewise, Hou et al.[39] modified the surface of hBN with a silane treatment, achieving the high thermal conductivity of 1.178 W/mK 3
ACCEPTED MANUSCRIPT while using only 30% by weight modified hBN in epoxy resin. Quite recently, Wu et al. [35] showed a 263% increase in thermal conductivity for styrene-butadiene rubber while using boron nitride nanosheets compatibilized using a silane coupling agent. In this study, we produced thermally conductive photopolymerizable polymers.
RI PT
Hexagonal boron nitride was used as the conductive filler, and its capacity to undergo chemical surface modifications was demonstrated. By using a hybrid initiator system,
we were able to increase the percentage of hBN added to the composite material, while
SC
still maintaining a high conversion percentage and photo-curing ability. Later, we
M AN U
evaluated the effect of different hBN compositions on the thermal and mechanical properties (thermal conductivity, coefficient of thermal expansion, and modulus of elasticity) of radiation-cured acrylic composites with different filler loadings. The results were then compared to relevant mathematical models.
TE D
2. Experimental
2.1.hBN surface modifications
hBN (5µm, M.K. Impex Corp.) was treated in the following ways: γ-Methacryloxypropyltrimethoxysilane, (MPTS): A methacryloxy-
EP
(i)
AC C
terminated silane coupling agent (Dynasylan® MEMO, Dynasylan) was added to 250 mL of water (1% v/v). Acetic acid (100%, Merck) was added dropwise to the solutions until a pH of 4 was reached. Approximately 7.5 gr of hBN was then added to the mixture while stirring. The reaction was stirred at room temperature for 24 hours. The mixture was then vacuumfiltered using a 2.5 µm mesh filter paper and washed with cold water 5 times. The treated particles were then dried in a vacuum oven overnight at 80°C.
4
ACCEPTED MANUSCRIPT (ii)
Carbodiimide treatment (CDI): 5 gr of hBN was added to 250 mL of THF (technical, Bio-Lab) in a round-bottomed flask, and stirred continuously. 195 mg of DMAP (4-dimethylaminopyridine, Sigma-Aldrich), 250 µL of DIC (N,N'-diisopropylcarbodiimide, Sigma-Aldrich) and 1 mL of acrylic acid
RI PT
(99%, Acros Organics) were then added, the flask was covered, and the
reaction was stirred for 24 hours at room temperature. The mixture was then vacuum-filtered using 2.5 µm mesh filter paper and washed with cold water
M AN U
2.2.Composite sample preparation
SC
5 times. The treated particles were then dried in an oven overnight at 110°C.
A mixture of thermosetting acrylic monomers was obtained from Stratasys Ltd, and used as the matrix. Treated and untreated hBN were added separately at loadings of 5%, 15%, 25% and 35% by weight into solutions with the acrylic monomer mixture and
TE D
with 2,2-Dimethoxy-2-phenylacetophenone (Sigma-Aldrich), phenylbis(2,4,6trimethylbenzoyl)phosphine oxide (Sigma-Aldrich), and o-benzoic sulfimide (Sigma-
EP
Aldrich). All mixtures were sonicated in a bath sonicator (Transsonic TS540, Elma) for 30 minutes to break up aggregates and to better disperse the filler within the mixture
AC C
(where viscosity permitted). The 25wt% mixtures were too viscous to sonicate, and were therefore mixed with a vortex mixer (Velp Scientifica) for 1 minute. With the exception of the 35wt% mixtures, all samples were poured into silicone molds after sonication. The 35wt% mixtures were highly viscous and had a paste-like consistency, and were therefore vortexed and then transferred manually to the mold using a spatula. Each mixture was then covered and rapidly radiation-cured using a high-intensity UV curing lamp (SB-100PC, Mercury 100W bulb, Spectroline). The UV lamp was situated 10 cm above the molds, and was pre-warmed for two minutes before curing. The 5
ACCEPTED MANUSCRIPT samples were cured for 2 minutes from one side, for 1 minute from the other side, and then removed from under the lamp and from their silicone molds. They were then left to complete their cure cycle at room temperature while exposed to air. Finally, the samples were cut to size for the required tests.
2.3.1
RI PT
2.3.Characterization methods
Modulated Differential Scanning Calorimetry (MDSC).
SC
Thermal conductivity measurements were taken using MDSC (DSC Q200, TA Instruments) in a modified procedure based on ASTM E 1952. TA Instruments has
M AN U
shown that this procedure is effective and accurate up to 4 W/mK [40]. Samples were cut into a right cylindrical shape at two thicknesses (~0.4 mm and over 3 mm), and placed directly onto the DSC sensor, after the sensor was moistened with a small amount of silicone oil. The reversing heat capacity was recorded after 15 minutes of a
TE D
quasi-isothermal run at 25°C with an amplitude of 0.5°C and an oscillation time of 20 seconds in a nitrogen atmosphere. The reversing heat capacities of the thin and thick disks (specific heat capacity,
, and apparent heat capacity, , respectively) were
EP
recorded. These heat capacities were then inserted into a simplified form of Fourier's
AC C
law of heat conduction using the following equation: =
8∙ ∙ ∙ ∙
∙
(1)
where L is the length of the thick sample, M is the mass of the thick sample, d is the diameter of both samples, and P is the oscillation time of the MDSC. 2.3.2
Modeling Thermal Conductivity Thermal conductivity was modeled using three existing models:
6
ACCEPTED MANUSCRIPT The Maxwell model [41] assumes a binary mixture with non-interacting spherical filler particles in small concentrations. The effective thermal conductivity is expressed as
where
+ 2( − −( −
) )
(2)
RI PT
+2 +2
=
is the thermal conductivity of the polymer (base fluid),
is the volumetric fraction of filler.
SC
conductivity of the filler (particle), and
is the thermal
The Nielsen Model [42], also known as the Lewis-Nielsen Model, is a modified
M AN U
version of a model predicting the modulus of composites, taking into account both the shape and packing of the filler in the polymer. The model is written as follows: =
1+ 1−
TE D
where
∙
−1
EP
=
AC C
(3)
+1
1− =1+!
"
"
, (4)
# , (5)
and A is a function of the geometry of the particles, and
"
is the maximum packing
fraction (the maximum amount of filler that still allows for a continuous matrix). The Agari-Uno [43] model is based on a generalization of the parallel and series models, and assumes that the dispersion is random within the material. It introduces two empirical factors,
%
and
.
% describes
the effect that the particles have on the
7
ACCEPTED MANUSCRIPT secondary structure of the polymeric matrix, while
describes the ease with which the
particles form conductive paths in the material [12]. Their model is given by the following equation:
2.3.3
=
∙
∙ &'(
+ (1 − ) ∙ log,
%
∙
- (6)
RI PT
&'(
Dynamic Mechanical Analysis (DMA).
Modulus of elasticity measurements were conducted using a DMA (Q800, TA
SC
Instruments) in controlled-force mode. The force was ramped from an initial force of 0.01N to 18N at a rate of 1N/min, at 298K and in an atmosphere of air. Coeffcient of
M AN U
thermal expansion (CTE) measurements [44] were measured using a DMA (Q800, TA Instruments) in multi-frequency (stress) mode. 0.4 mm thick rectangular samples were ramped from a temperature of 273K to 423K at a rate of 3K/min, with a force of 0.05N, and a frequency of 10Hz. The initial force was 0.01N, the initial amplitude was 0.1µm,
TE D
and the measurements were conducted in an atmosphere of air. Data for CTE measurements was taken from 283K to 303K. X-Ray Photoelectron Spectroscopy (XPS).
EP
2.3.4
XPS spectra were recorded with an X-Ray photoelectron spectrometer (Axis
AC C
Ultra, Kratos Analytical LTD) using an Al Kα monochromatic radiation X-ray source (1486.7eV). Data was collected and analyzed using a Casa XPS and Vision Data processing program. High-resolution XPS spectra were collected with a takeoff angle of 90°. Vacuum condition in the chamber was held at 1.9 × 1034 Torr for the C 1s, O 1s, N 1s, B 1s and Si 2p levels with pass energy of 20 and 0.1 eV step size. The binding energies were calibrated using C 1s peak energy as 285.0 eV.
8
ACCEPTED MANUSCRIPT 2.3.5
Fourier Transform Infrared Spectroscopy - Attenuated Total Reflectance (FTIRATR). The infrared spectra were obtained with an FTIR spectrometer (Alpha-P,
Bruker), using single-bounce ATR with a diamond crystal, and by collecting and
RI PT
averaging 24 scans at the range of 400 to 4000 cm-1 with a resolution of 4 cm-1. A heavy glass pane was used to press the samples against the ATR crystal, and liquid silane was dripped directly onto the crystal. To calculate degree of conversion for the acrylic
SC
polymer, ATR curves were converted to FTIR curves, and the degree of conversion was
equation: 5' = 100 × 61 −
7'89:&;<= >'&?9=8 = >=: @ (7) 7'89:&;<= 9'7'9=8 = >=:
Viscosity measurements.
TE D
2.3.6
M AN U
then calculated in reference to the normalized peak of the monomer using the following
Qualitative viscosity measurements were modeled after flow cup measurements such as Ford cup viscosity [45]. Liquid materials were poured into a 1mL syringe
EP
without a plunger, and then allowed to flow freely. This process was video-recorded and then analyzed in slow-motion in order to determine the amount of time taken for the
AC C
materials to flow. 2.3.7
Thermogravimetric Analysis (TGA). Thermogravimetric analysis was performed using a high-resolution TGA (Q500,
TA instruments). The samples were heated at a rate of 30°C/min to 600°C at a resolution of 4°C in a nitrogen atmosphere. 2.3.8
High-Resolution Scanning Electron Microscopy (HR-SEM).
9
ACCEPTED MANUSCRIPT Field emission HR-SEM (Gemini Ultra Plus, Zeiss) with a low voltage of 1eV and a working distance of 4.9 mm was used to observe the surface of the synthesized composites. The specimens were not coated, and the SEM images were magnified up to x10000.
RI PT
3. Results and Discussion. 3.1.Surface chemistry characterization.
SC
The hBN surface chemistry was evaluated by both FTIR-ATR and XPS. FTIRATR analysis revealed that hexagonal boron nitride has two extremely prominent peaks
M AN U
at approximately 800 cm-1 and 1360 cm-1 (Figure 1), which can be ascribed to its inplane and out-of-plane ring stretching vibrations, respectively [46,47]. In addition to these peaks, a shoulder can be seen at approximately 925 cm-1.
When the MPTS treatment was applied, the two main peaks shifted to yield
TE D
peaks at 787 cm-1 and 1358 cm-1. In addition, a new peak appeared at 1727 cm-1, implying a carbonyl group on the surface, which matches the carbonyl peak of the MEMO silane (shown in the supplementary information). This finding suggests that the
EP
silane was covalently bonded to the hBN surface, with the peak's small size corresponding to the small amount of hBN edge defects available for bonding [47].
AC C
Slight shoulders can be seen adjacent to the main 1358 cm-1 hBN peak at about 1000 – 1150 cm-1, a region typically associated with Si-O [48]. This further supports the assumption that a covalent bond was formed between the silanol and the hBN surface. FTIR-ATR analysis of the carbodiimide-treated (CDI) hBN showed a shift of the two main peaks to 774 cm-1 and 1347 cm-1. Furthermore, three new peaks formed at 1651 cm-1, 1564 cm-1, and 1193 cm-1 (as a shoulder). The peak at 1193 cm-1 can be ascribed to the C-N stretch of the amide [49], while the peaks at 1651 cm-1 and 1564
10
ACCEPTED MANUSCRIPT cm-1 are slightly more ambiguous. These peaks can be attributed either to the amide carbonyl group (1651 cm-1: C=O stretch, 1564 cm-1: N-H bend) or to the C=C double bond, however they more likely indicate the amide carbonyl, as the carbonyl peak is usually stronger than that of the double bond. The fact that there is only one carbonyl
RI PT
peak indicates that only the amine groups on hBN's surface reacted, since they are both more prevalent and more basic than the hydroxyl groups [37]. In similar experiments, in which pyridine was used as an activating basic agent or when the reaction was run for
SC
96 hours (instead of 24), the same results were found, indicating that the free hydroxyl
1 0.9 0.8 0.7 0.6 0.5
1740
1700
1660
1620
0.3 0.2 0.1 0 4000
1600
1575
3600
TE D
0.4
M AN U
group on the hBN is not reactive under the conditions explored [50,51].
3200
1550
2800
EP
CDI
2400
MPTS
2000
1600
1200
800
400
Untreated
AC C
Figure 1. FTIR-ATR plots of untreated hBN (dotted), MPTS hBN (light gray) and CDI hBN (black). The insets display enlargements of the regions between 1550 - 1600 cm-1, and 1620 - 1740 cm-1
The results of all XPS tests were similar to one another (Figure 2), with the exception of the MPTS-treated sample which showed two additional peaks resulting from the added Si (2p). The untreated hBN exhibited a carbon peak at its surface,
11
ACCEPTED MANUSCRIPT possibly as a result of either a small amount of carbon-containing contaminant, or an
RI PT
artifact caused by the sample holder
CPS
N (1s)
B (1s)
O (1s)
C (1s)
1000
800 CDI
600
400
200
M AN U
1200
SC
Si (2p)
MPTS
0
Untreated
Figure 2. XPS survey plots of untreated hBN (dotted), MPTS hBN (light gray) and CDI
TE D
hBN (black)
The deconvoluted C (1s) peaks provided further insight into the material's surface. Both the MPTS- and CDI-treated hBN showed a C=O peak within their
EP
deconvoluted C (1s) peaks, suggesting the addition of a carbonyl group to the surface of the hBN. This information is best viewed in percentages, as shown in Table 1.
AC C
Table 1. Deconvoluted C(1s) peaks of all hBN samples (in precentages) Adventitious
C-O bond
C=O bond
Untreated
80.79%
19.21%
-
MPTS
71.21%
17.80%
10.99%
CDI
72.43%
25.44%
2.13%
12
ACCEPTED MANUSCRIPT From these results, it can be concluded that the MPTS was more successful than CDI as a surface treatment, especially when considering that the carbonyl group is a much larger section of the CDI treatment than of the MPTS treatment. Figure 3 shows the results of a thermogravimetric analysis performed for the
RI PT
different hBN powders. The pristine hBN showed no more than a ~0.5% weight loss
when heated to 600°C. The MPTS-treated hBN, however, showed a 1.6% decrease in weight, suggesting that an organic layer had been previously present. The main
SC
temperature decrease was recorded at approximately 441.2°C, and was confirmed by the
M AN U
derivative weight change curve. Since this temperature is above the boiling point of the original MEMO silane, we concluded that the organic layer was the result of a covalent bond between the silane and the hBN surface, rather than simple adsorption. These findings concur with other thermogravimetric findings of silane-treated hBN [52,53]. The CDI-treated hBN showed a 4.8% weight decrease, with a main peak on the
TE D
derivative weight change curve at approximately 177°C, again hinting at a new covalent
AC C
EP
bond between the acrylic acid and the hBN surface.
13
ACCEPTED MANUSCRIPT 101
99 98 97
RI PT
Weight percent (%)
100
96 95 94 150
225
300
375
Temperature (°C) MPTS
525
600
Untreated
M AN U
CDI
450
SC
75
Figure 3. Weight percent loss vs. temperature plots of untreated hBN (dotted), MPTS hBN (light gray) and CDI hBN (black)
The combined results of the chemical analyses allow for an estimate of the
TE D
chemical modifications made to the pristine hBN, using the two methods discussed. A
AC C
EP
schematic representation can be seen in Scheme 1.
Pristine hBN
CDI hBN
MPTS hBN
Scheme 1. Schematic representation of the chemical modifications made to pristine hBN
3.2.Uncured monomer mixture characterization.
14
ACCEPTED MANUSCRIPT Viscosity measurements using a modified "Ford Cup" method (measured in seconds) were performed on the neat monomer as well as on 5% and 10% (w/w) hBNloaded samples. Higher concentrations of hBN proved too viscous to be poured into the syringe used. Table 2 details the viscosities found.
Untreated hBN
MPTS hBN
CDI hBN
0.69s
5wt%
0.86s
0.82s
0.83s
10wt% 1.26s
1.10s
1.30s
M AN U
SC
0wt%
RI PT
Table 2. Viscosity measurement of different formulations of treated and untreated hBN
From Table 2 it is clear that the addition of hBN to the neat acrylic monomer increased the viscosity of the mixture. The smaller increases in viscosity seen with
TE D
MPTS-modified samples possibly indicate superior compatibility between the filler and the matrix when compared to CDI. The monomer's viscosity is an important factor in applications that rely on the use of thin liquids, and as such, future research in the field
EP
is required to minimalize this increase in viscosity with the addition of hBN [54,55].
AC C
Dispersion longevity was evaluated visually for 1wt% hBN dispersions in the uncured monomer. The dispersions were sonicated for 45 minutes, set aside for various amounts of time, and photographed periodically. Figure 4 depicts the changes in the dispersions over a 12-day period.
15
SC
RI PT
ACCEPTED MANUSCRIPT
M AN U
Figure 4. Treated and untreated hBN dispersions in an uncured monomer over the course of 12 days
Based on these visual results, the MPTS treatment seemed to allow for the least amount of precipitation. The CDI treatment showed slightly better dispersion than
TE D
untreated hBN. A similar approach to stabilizing mixtures has been used extensively with CNTs, wherein functional groups are attached to surface defects on the CNTs. This method has been shown to achieve enhanced stability of dispersion in polymeric
AC C
EP
matrices [56–58].
3.3.Degree of Conversion
16
ACCEPTED MANUSCRIPT 1 0.9 0.8 0.7 0.6 0.5
0.3 0.2
0 4000
1800
1700
3600
1600 3200
Monomer
1500
1400
2800
1300
2400
2000
1600
1200
800
400
SC
0.1
RI PT
0.4
No accelerator
Accelerator, surface
Accelerator, bulk
M AN U
Figure 5. FTIR transmittance of the neat monomer and the polymer when cured under different conditions. In the inset, the region between 1300 and 1800 cm-1 is enlarged. The acrylic monomer was radiation-cured two separate ways: using only 2,2Dimethoxy-2-phenylacetophenone, and using the two photoinitiators and the curing
TE D
accelerator. In order to neutralize the two dominant hBN peaks in the FTIR graph, the polymer was cured without hBN. When the curing accelerator was used, both the surface and the bulk of the cured polymer were examined for degree of conversion
EP
calculations. Figure 5 depicts the FTIR graphs of the polymers cured in different
AC C
conditions. In addition, Table 3 shows the examination of the double bond peak at ~1406 cm-1 using the carbonyl peak at ~1721 cm-1 as an internal standard [59,60]. Table 3. Absolute and normalized intensties of carbonyl and double bond peak areas of different curing procedures Carbonyl peak ~1721
C=C peak ~1406 cm-1
cm-1 Absolute
Degree of conversion
Normalized
Absolute
Normalized
(%)
17
ACCEPTED MANUSCRIPT (%)
Int.
(%)
Monomer
3.0187
100
1.2003
39.8
-
No acc.
3.4247
100
0.5354
15.6
60.7
Acc.,
1.5864
100
0.0799
5
87.3
0.2997
100
0.0054
1.8
Surface Acc., Bulk
RI PT
Int.
95.5
SC
The cured polymers showed a distinctly higher degree of conversion when the curing
M AN U
accelerator was used, both at the surface and within the material, going from approximately 60% of conversion to more than 95%. 3.4.Cured composites characterization.
0.8
TE D
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
5
10
AC C
0
EP
Thermal Conductivity (W/mK)
0.9
15
20
25
30
35
40
%hBN by weight CDI
MPTS
Untreated
Figure 6. Thermal conductivity of Acrylic/hBN composites. For clarity, values are shown without error bars Figure 6 depicts the thermal conductivities of untreated and treated hBNs as a function of hBN weight fraction. Within the compositions studied, the influence of the surface
18
ACCEPTED MANUSCRIPT treatments seems negligible. However, as the hBN content increased in the synthesized composites, the MPTS-treated hBN showed a slight increase in thermal conductivity as compared to the other treated hBNs. This may be as a result of the enhanced compatibility between the MPTS and the polymeric matrix [17], however the difference
RI PT
in values is not pronounced enough to show a clear trend. The fact that all plots in
Figure 6 can be fitted rather well with linear trend lines in this region suggests that the hBN content in these compositions is too low to fully achieve thermally conductive
SC
paths in the composite material, thus limiting their thermal conductivity [14].
M AN U
Predictions regarding thermal conductivity were made with three known models (Maxwell, Nielsen, Agari & Uno), while using calculated volume fractions and the data gathered on pristine hBN. Figure 7 shows the thermal conductivity of unmodified hBN/acrylic composites from 1% by weight up to 35%. As can be seen in the graph, the Maxwell model, which does not take particle shape into account, is less accurate for the
TE D
non-spherical hexagonal boron nitride particles. The Nielsen model, on the other hand, achieved better results using
= 10 and
"
= 0.637. The shape parameter, A,
EP
signifies that the hBN particles are non-spherical, and have a higher aspect ratio. The values found in these calculations are close to those given by other authors [61]. The
AC C
maximum packing fraction,
",
is in accordance with three-dimensional random close
orientation of irregularly shaped fillers [62]. In order to plot the Agari-Uno model, the factors
%
= 1.1;
= 0.9193 were calculated using measurements made for the
untreated hBN. The very high
%
value appears to hint that the polymer has a closer
pack after the addition of hBN, while the high
value signifies that hBN is able to
efficiently form conductive paths when sufficient loading is used in the formulation. These findings indicate that the thermal conductivity of photopolymerizable composites does not behave differently from that of other thermally polymerized materials. 19
ACCEPTED MANUSCRIPT 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0
0.2
0.4
0.6
0.8
1
SC
Weight fraction of hBN
RI PT
Thermal Conductivty (W/mK)
1
Maxwell Model
Agari-Uno Model
Measured Values
M AN U
Nielsen Model
Figure 7. Thermal conductivity of composites containing different hBN concentrations according to three different models (Maxwell, Nielsen, and Agari-Uno) and the experimental values measured for untreated hBN. The graphs are cut-off at an arbitrary
TE D
1 W/mK value to better show models in contrast with the experimental results Because most thermal conductivity measurements are based on calculations
EP
rather than direct measurement, the results are prone to large errors. Here, two independent models fit well with the experimental results, providing much more
AC C
credence to the thermal conductivity measurement method used. In addition, the applied models have been proven to be rather effective for a variety of composite materials. Nielsen's model, for example, has been used for calculations involving boron nitride nanosheets in SBR rubber [35], hBN and secondary carbon-based fillers in poly(phenylene sulfide) [63], hBN-filled thermoplastics [64], hBN in several typed of epoxy [61], and aluminum nitride filled epoxy [20], to name a few. Agari and Uno's model has also been proven to be effective, and examples can be found for PVC and polyethylene filled with either carbon black or graphite [65], silicon-reinforced HDPE 20
ACCEPTED MANUSCRIPT [66], polystyrene filled with aluminum nitride [67], and epoxy filled with several ceramic fillers [12]. Although HRSEM images could not conclusively show improved adhesion between the matrix and the treated fillers, the CDI treatment did seem to allow for a
RI PT
more homogenic composite, while the MPTS treatment and the untreated hBN seem to create similar composite materials. HRSEM images are provided in the supplementary
SC
information (Figure 12).
The coefficients of thermal expansion of the different composites were
M AN U
investigated using DMA. The experiments described the change in sample dimension as a function of the original length, and the CTE can be then be calculated from this information. All DMA curves had two wide-ranged regions, corresponding to the glassy state and to the rubbery state, and a transition between the two [68]. However, even
TE D
within the same region, the slope was found to change as the temperature rose, which is indicative of the expected CTE change.
Because the CTE curves change continuously, the values reported in Figure 8
EP
are extracted from the range of ~ 283K – 303K, which is in the glassy region for all composites. As the filler concentration increased, the CTE generally decreased, which is
AC C
to be expected with a ceramic filler (ceramics tend to have low CTEs, while polymers usually have higher CTEs [12,69]). One of the most common models for calculating CTEs is the Schapery model
[66,70,71], which assumes that the composites are isotropic and applies extremum values based on energy principles. The model establishes upper and lower bounds for CTE composites. It is clear from Figure 8 that the measured values are well within Schapery's bounds, with the single exception of 35% by weight hBN composites, which
21
ACCEPTED MANUSCRIPT all seem to be just under the lower bound. These deviations may be the result of the hBN platelets orienting themselves perpendicular to the testing direction, as their CTE is about one order of magnitude smaller for transverse testing. 80
RI PT
60 50 40 30 20 10 0 0%
5%
10%
15%
20%
25%
CDI
MPTS
30%
M AN U
hBN content in % by weight
SC
CTE (μm/mK)
70
Untreated
Upper bound
35%
40%
Lower bound
Figure 8. CTE of the different composites fabricated at 298K compared to the
TE D
theoretical Schapery's model
The mechanical properties of the fabricated composites were investigated using their stress-strain curves, generated from DMA measurements. Overall, the samples
EP
showed trends similar to one another when their Young's Moduli were examined using a
AC C
linear trend line at the beginning of their stress-strain curves (Figure 9).
22
ACCEPTED MANUSCRIPT 7000 6000
E (MPa)
5000 4000 3000
1000 0 5%
10%
15%
20%
25%
hBN content in % by weight CDI
MPTS
Untreated
30%
35%
40%
SC
0%
RI PT
2000
M AN U
Figure 9. Modulus of elasticity of the different composites at varying concentrations Small amounts of boron nitride within polymeric matrices have been shown to improve mechanical strength [69]. As a result, the slight increase in Young's modulus for the 5% by weight composite seen in Figure 9 is to be expected. The decreased
TE D
modulus at 15% by weight is most likely the result of hBN aggregates slipping over one another, causing the well-documented "lubricating" effect [69]. The second increase,
EP
starting with approximately 25% by weight composites, is most likely due to the increased particulate matter, leading to an overall stronger, more brittle material. When
AC C
examining the stress-strain curves of the highest-filled composites for each series, all materials showed a higher Young's modulus, as expected. The untreated hBN composites exhibited embrittlement, showing small amounts of strain before failure. Both treated hBN composites, however, showed much higher elongation, i.e. improved ductility of both composites. Stress-strain curves for all 35wt% compositions and for the neat acrylic are presented in the supplementary information (Figure 13). 4. Conclusions.
23
ACCEPTED MANUSCRIPT To the best of our knowledge, this is the first investigation of thermal conductivity of a photopolymerizable thermosetting material. We synthesized photocurable composites through the addition of a small amount of curing accelerator and radiation polymerization with high loadings of thermally conductive hBN. The physical
RI PT
properties of the polymerized composites were not significantly affected by hBN
covalent functionalization using silane (MPTS) and carbodiimide reaction (CDI).
However, the effects of this functionalization on the prepolymer stability and viscosity
SC
were prominent. These finding suggest that functionalization may be critical when
considering the shelf life of the examined compositions. Overall, the addition of hBN
M AN U
resulted in a significant enhancement of thermal conductivity (from 0.19 W/mK up to 0.84 W/mK), while still allowing the composite to cure using only irradiation. In addition, the mathematical calculations fit well with the relevant models (Nielsen, Agari & Uno).
TE D
5. Supplementary information.
Further information regarding the FTIR-ATR results and peak assignments for
EP
MPTS, the thermal conductivity results for each composition of acrylic/hBN (including omitted error bars), the HRSEM images of the 35wt% compositions, and the stress-
AC C
strain curves for all 35wt% compositions as well as for the unfilled material are included in the supplementary information section. 6. References.
[1]
H. Ishida, S. Rimdusit, Very high thermal conductivity obtained by boron nitridefilled polybenzoxazine, Thermochim. Acta. 320 (1998). doi:10.1016/S00406031(98)00463-8.
[2]
R. Mahajan, G. Chrysler, Cooling a Microprocessor Chip, Proc. IEEE. 94 (2006) 24
ACCEPTED MANUSCRIPT 1476–1486. doi:10.1109/JPROC.2006.879800. [3]
G.-W. Lee, M. Park, J. Kim, J.I. Lee, H.G. Yoon, Enhanced thermal conductivity of polymer composites filled with hybrid filler, Compos. Part A Appl. Sci.
[4]
RI PT
Manuf. 37 (2006) 727–734. doi:10.1016/j.compositesa.2005.07.006. H.H. Chiang, C.A. Hieber, K.K. Wang, A unified simulation of the filling and postfilling stages in injection molding. Part I: Formulation, Polym. Eng. Sci. 31
[5]
SC
(1991) 116–124. doi:10.1002/pen.760310210.
K.P. Karunakaran, S. Suryakumar, V. Pushpa, S. Akula, Low cost integration of
M AN U
additive and subtractive processes for hybrid layered manufacturing, Robot. Comput. Integr. Manuf. 26 (2010) 490–499. doi:10.1016/j.rcim.2010.03.008. [6]
A. Bjorneklett, L. Halbo, H. Kristiansen, Thermal conductivity of epoxy adhesives filled with silver particles, Int. J. Adhes. Adhes. 12 (1992) 99–104.
[7]
TE D
doi:10.1016/0143-7496(92)90030-Y.
R.G. Humfeld, D.A. Dillard, Residual stress development in adhesive joints to
thermal
cycling,
J.
Adhes.
65
(1998)
277–306.
EP
subjected
doi:10.1080/00218469808012250. G. Viana, M. Costa, M. Banea, L. da Silva, A review on the temperature and
AC C
[8]
moisture degradation of adhesive joints, Proc. Inst. Mech. Eng. Part L J. Mater. Des. Appl. 0 (2016) 1–14. doi:10.1177/1464420716671503.
[9]
H. Lipson, M. Kurman, Fabricated: The new world of 3D printing, 2013.
[10] C. Zweben, Ultrahigh-thermal-conductivity packaging materials, Semicond. Therm. Meas. Manag. Symp. 2005 IEEE Twenty First Annu. IEEE. (2005) 168– 174. doi:10.1109/stherm.2005.1412174. 25
ACCEPTED MANUSCRIPT [11] A. a Balandin, S. Ghosh, W. Bao, I. Calizo, D. Teweldebrhan, F. Miao, C.N. Lau, Superior Thermal Conductivity of Single-Layer Graphene 2008, Nano Lett. 8 (2008) 902–907. doi:10.1021/nl0731872. [12] C.P. Wong, R.S. Bollampally, Thermal Conductivity , Elastic Modulus , and
RI PT
Coefficient of Thermal Expansion of Polymer Composites Filled with, 74 (1999) 3396–3403.
Future
Directions,
Proc.
IEEE.
94
M AN U
doi:10.1109/JPROC.2006.879796.
SC
[13] R. Prasher, Thermal Interface Materials: Historical Perspective, Status, and (2006)
1571–1586.
[14] Z. Han, A. Fina, Thermal conductivity of carbon nanotubes and their polymer nanocomposites:
A
review,
Prog.
Polym.
Sci.
36
(2011)
914–944.
doi:10.1016/j.progpolymsci.2010.11.004.
TE D
[15] T.-L. Li, S.L.-C. Hsu, Enhanced thermal conductivity of polyimide films via a hybrid of micro- and nano-sized boron nitride., J. Phys. Chem. B. 114 (2010)
EP
6825–9. doi:10.1021/jp101857w.
[16] A. Yu, P. Ramesh, M.E. Itkis, E. Bekyarova, R.C. Haddon, Graphite
AC C
Nanoplatelet-Epoxy Composite Thermal Interface Materials, J. Phys. Chem. C. 111 (2007) 7565–7569. doi:10.1021/jp071761s.
[17] I. a. Tsekmes, R. Kochetov, P.H.F. Morshuis, J.J. Smit, Thermal conductivity of polymeric composites: A review, 2013 IEEE Int. Conf. Solid Dielectr. (2013) 678–681. doi:10.1109/ICSD.2013.6619698. [18] Y. Xu, D.D.L. Chung, Increasing the thermal conductivity of boron nitride and aluminum nitride particle epoxy-matrix composites by particle surface
26
ACCEPTED MANUSCRIPT treatments,
Compos.
Interfaces.
7
(2000)
243–256.
doi:10.1163/156855400750244969. [19] P. Bujard, Thermal Conductivity of Boron Nitride Filled Epoxy Resin: Temperature Dependence and Influence of Sample Preparation, Intersoc. Conf.
RI PT
Therm. Phenom. Fabr. Oper. Electron. Components I-THERM ’88. (1988) 41– 49.
SC
[20] P. Bujard, J.P. Ansermet, Thermally conductive aluminium nitride-filled epoxy resin for electronic packaging, Fifth Annu. IEEE Semicond. Therm. Temp. Meas.
M AN U
Symp. (1989) 126–130.
[21] S. Ino, T. Shiobara, Thermal Conductivity Of Molding Compounds For Plastic Packaging, IEEE Trans. Components Packag. Manuf. Technol. Part A. 17 (1994) 527–532. doi:10.1109/95.335037.
TE D
[22] Z. Pu, J.E. Mark, J.M. Jethmalani, W.T. Ford, Effects of Dispersion and Aggregation of Silica in the Reinforcement of Poly(methyl acrylate) Elastomers,
EP
Chem. Mater. 9 (1997) 2442–2447. doi:10.1021/cm970210j. [23] K. V. Wong, A. Hernandez, A Review of Additive Manufacturing, ISRN Mech.
AC C
Eng. 2012 (2012) 1–10. doi:10.5402/2012/208760. [24] R. Singh, Process capability study of polyjet printing for plastic components, Evol. Ecol. 25 (2011) 1011–1015. doi:10.1007/s12206-011-0203-8.
[25] J.G. Leprince, P. Leveque, B. Nysten, B. Gallez, J. Devaux, G. Leloup, New insight into the “depth of cure” of dimethacrylate-based dental composites, Dent. Mater. 28 (2012) 512–520. doi:10.1016/j.dental.2011.12.004. [26] A.C. Shortall, How light source and product shade influence cure depth for a 27
ACCEPTED MANUSCRIPT contemporary
composite,
J.
Oral
Rehabil.
32
(2005)
906–911.
doi:10.1111/j.1365-2842.2005.01523.x. [27] J.V. Koleske, Radiation Curing of Coatings, ASTM International, 2002.
RI PT
[28] C. Decker, UV-radiation curing chemistry, Pigment Resin Technol. 30 (2001) 278–286. doi:10.1108/03699420110404593.
[29] P. Gloeckner, Radiation Curing (European Coatings Tech Files), 2008.
SC
[30] S. Adanur, Y. Arumugham, Characteristics of Ultraviolet Cured Glass-Epoxy Textile Composites: Part 1: Experimental Procedures and Testing, J. Ind. Text.
M AN U
32 (2002) 93–106. doi:10.1106/152808302031090.
[31] J.J. Park, D.M. Park, J.H. Youk, W.R. Yu, J. Lee, Functionalization of multiwalled carbon nanotubes by free radical graft polymerization initiated from photoinduced
surface
groups,
Carbon
N.
Y.
48
(2010)
2899–2905.
TE D
doi:10.1016/j.carbon.2010.04.024.
[32] F.H. Gojny, J. Nastalczyk, Z. Roslaniec, K. Schulte, Surface modified multi-
EP
walled carbon nanotubes in CNT/epoxy-composites, Chem. Phys. Lett. 370
AC C
(2003) 820–824. doi:10.1016/S0009-2614(03)00187-8. [33] R. Kochetov, Thermal and Electrical Properties of Nanocomposites, Including Material Processing, 2012.
[34] P.C. Irwin, Y. Cao, A. Bansal, L.S. Schadler, Thermal and mechanical properties of polyimide nanocomposites, 2003 Annu. Rep. Conf. Electr. Insul. Dielectr. Phenom. (2003) 1–4. [35] X. Wu, H. Liu, Z. Tang, B. Guo, Scalable fabrication of thermally conductive elastomer / boron nitride nanosheets composites by slurry compounding, 28
ACCEPTED MANUSCRIPT Compos.
Sci.
Technol.
123
(2016)
179–186.
doi:10.1016/j.compscitech.2015.12.010. [36] C.W. Nan, R. Birringer, D.R. Clarke, H. Gleiter, Effective thermal conductivity of particulate composites with interfacial thermal resistance, J. Appl. Phys. 81
RI PT
(1997) 6692–6699. doi:10.1063/1.365209.
[37] M.T. Huang, H. Ishida, Investigation of the boron nitride / polybenzoxazine
doi:10.1002/(SICI)1099-0488(19990901)37.
SC
interphase, J. Polym. Sci. Part B Polym. Phys. 37 (1999) 2360–2372.
M AN U
[38] J. Yu, X. Huang, C. Wu, X. Wu, G. Wang, P. Jiang, Interfacial modification of boron nitride nanoplatelets for epoxy composites with improved thermal properties,
Polymer
(Guildf).
53
(2012)
471–480.
doi:10.1016/j.polymer.2011.12.040.
TE D
[39] J. Hou, G. Li, N. Yang, L. Qin, M.E. Grami, Q. Zhang, N. Wang, X. Qu, Preparation and Characterization of Surface Modified Boron Nitride Epoxy Composites with Enhanced Thermal Conductivity, RSC Adv. 4 (2014) 44282–
EP
44290. doi:10.1039/C4RA07394K.
AC C
[40] E. Verdonck, G. Dreezen, Thermal Conductivity Measurements of Conductive Epoxy Adhesives by MDSC, Therm. Libr. Appl. Br. TA312. (n.d.). http://www.tainstruments.co.jp/application/pdf/Thermal_Library/Applications_B riefs/TA312.PDF.
[41] C. Maxwell, James, A treatise on electricity and magnetism, (1873). [42] L.E. Nielsen, Thermal conductivity of particulate-filled polymers, J. Appl. Polym. Sci. 17 (1973) 3819–3820. doi:10.1002/app.1973.070171224.
29
ACCEPTED MANUSCRIPT [43] Y. Agari, T. Uno, Estimation on thermal conductivities of filled polymers, J. Appl. Polym. Sci. 32 (1986) 5705–5712. doi:10.1002/app.1986.070320702. [44] P.F. Rios, S. Kenig, R. Cohen, A. Shechter, The effect of carbon nanotubes on the thermal expansion isotropy of injection molded carbon fiber reinforced
RI PT
thermoplastics, Polym. Compos. 34 (2013) 1367–1374. doi:10.1002/pc.22551.
[45] ASTM International, Standard Test Method for Viscosity by Ford Viscosity Cup,
SC
(2010).
[46] R. Geick, C.H. Perry, G. Rupprecht, Normal Modes in Hexagonal Boron Nitride,
M AN U
Phys. Rev. 146 (1966) 543–547. doi:10.1103/PhysRev.146.543.
[47] K. Sato, H. Horibe, T. Shirai, Y. Hotta, H. Nakano, H. Nagai, K. Mitsuishi, K. Watari, Thermally conductive composite films of hexagonal boron nitride and polyimide with affinity-enhanced interfaces, J. Mater. Chem. 20 (2010) 2749.
TE D
doi:10.1039/b924997d.
[48] M.A. Villegas, J.M.F. Navarro, Characterization of BaOa-SiOa glasses prepared
EP
via sol-gel, J. Mater. Sci. 23 (1988) 2464–2478. doi:10.1007/BF01111904. [49] M. Habibi, R. Amrollahi, M.H.S. Alavi, Polymerization of Acrylic Acid by a 4kJ
AC C
plasma focus device, (2012) 1–4.
[50] B. Neises, W. Steglich, Simple Method for the Esterification of Carboxylic Acids,
Angew.
Chemie
Int.
Ed.
English.
17
(1978)
522–524.
doi:10.1002/anie.197805221. [51] C.A.G.N. Montalbetti, V. Falque, Amide bond formation and peptide coupling, Tetrahedron. 61 (2005) 10827–10852. doi:10.1016/j.tet.2005.08.031. [52] K. Kim, M. Kim, J. Kim, Fabrication of UV-curable polyurethane acrylate 30
ACCEPTED MANUSCRIPT composites containing surface-modified boron nitride for underwater sonar encapsulant
application,
Ceram.
Int.
40
(2014)
10933–10943.
doi:10.1016/j.ceramint.2014.03.092. [53] K. Kim, M. Kim, Y. Hwang, J. Kim, Chemically modified boron nitride-epoxy
RI PT
terminated dimethylsiloxane composite for improving the thermal conductivity, Ceram. Int. 40 (2014) 2047–2056. doi:10.1016/j.ceramint.2013.07.117.
SC
[54] B. Derby, Inkjet Printing of Functional and Structural Materials: Fluid Property Requirements, Feature Stability, and Resolution, Annu. Rev. Mater. Res. 40
M AN U
(2010) 395–414. doi:10.1146/annurev-matsci-070909-104502.
[55] S.D. Hoath, O.G. Harlen, I.M. Hutchings, Jetting behavior of polymer solutions in drop-on-demand inkjet printing, J. Rheol. (N. Y. N. Y). 56 (2012) 1109. doi:10.1122/1.4724331.
coupling
TE D
[56] P.C. Ma, J. Kim, B.Z. Tang, Functionalization of carbon nanotubes using a silane agent,
Carbon
N.
Y.
44
(2006)
3232–3238.
EP
doi:10.1016/j.carbon.2006.06.032.
[57] P.C. Ma, J.K. Kim, B.Z. Tang, Effects of silane functionalization on the
AC C
properties of carbon nanotube/epoxy nanocomposites, Compos. Sci. Technol. 67 (2007) 2965–2972. doi:10.1016/j.compscitech.2007.05.006.
[58] P.-C. Ma, N.A. Siddiqui, G. Marom, J.-K. Kim, Dispersion and functionalization of carbon nanotubes for polymer-based nanocomposites: A review, Compos. Part A
Appl.
Sci.
Manuf.
41
(2010)
1345–1367.
doi:10.1016/j.compositesa.2010.07.003. [59] Application Note AN # 103 Determination of the degree of cure of a varnish, n.d.
31
ACCEPTED MANUSCRIPT [60] S. Imazato, J.. McCabe, H. Tarumi, A. Ehara, S. Ebisu, Degree of conversion of composites measured by DTA and FTIR, Dent. Mater. 17 (2001) 178–183. doi:10.1016/S0109-5641(00)00066-X. [61] R.F. Hill, P.H. Supancic, Thermal Conductivity of Platelet-Filled Polymer
RI PT
Composites, J. Am. Ceram. Soc. 85 (2002) 851–857. doi:10.1111/j.11512916.2002.tb00183.x.
(1986) 125–140. doi:10.1002/pc.750070302.
SC
[62] D.M. Bigg, Thermally conductive polymer compositions, Polym. Compos. 7
M AN U
[63] M.O. Khan, Thermally Conductive Polymer Composites for Electronic Packaging Applications, University of Toronto, 2012.
[64] C. Raman, A. Murugaiah, B. Xiang, R. Roden, Thermally conductive but electrically insulating plastics for thermal management applications, Proc. 12th
TE D
Automot. Compos. Conf. (2012) 23.
[65] Y. Agari, T. Uno, Thermal Conductivity of Polymer Filled With Carbon
EP
Materials: Effect of Conductive Particle Chains on Thermal Conductivity., J. Appl. Polym. Sci. 30 (1985) 2225–2235. doi:10.1002/app.1985.070300534.
AC C
[66] T.K. Dey, M. Tripathi, Thermal properties of silicon powder filled high-density polyethylene
composites,
Thermochim.
Acta.
502
(2010)
35–42.
doi:10.1016/j.tca.2010.02.002.
[67] S. Yu, P. Hing, X. Hu, Thermal conductivity of polystyrene–aluminum nitride composite, Compos. Part A Appl. Sci. Manuf. 33 (2002) 289–292. doi:10.1016/S1359-835X(01)00107-5. [68] E.S.A. Rashid, K. Ariffin, C.C. Kooi, H.M. Akil, Preparation and properties of 32
ACCEPTED MANUSCRIPT POSS/epoxy composites for electronic packaging applications, Mater. Des. 30 (2009) 1–8. doi:10.1016/j.matdes.2008.04.065. [69] G. Wypych, Handbook of Fillers (3rd Edition), 2010.
on
Energy
Principles,
J.
Compos.
doi:10.1177/002199836800200308.
Mater.
RI PT
[70] R.A. Schapery, Thermal Expansion Coefficients of Composite Materials Based 2
(1968)
380–404.
SC
[71] A.M. Drews, Control of thermal expansion coefficient of a metal powder
AC C
EP
TE D
M AN U
composite via ceramic nanofiber reinforcement, University of Akron, 2009.
33
S0266-3538(17)30528-6
DOI:
10.1016/j.compscitech.2017.09.001
Reference:
CSTE 6890
To appear in:
Composites Science and Technology
Received Date: 11 March 2017 Revised Date:
28 August 2017
Accepted Date: 2 September 2017
Please cite this article as: Goldin N, Dodiuk H, Lewitus D, Enhanced thermal conductivity of photopolymerizable composites using surface modified hexagonal boron nitride fillers, Composites Science and Technology (2017), doi: 10.1016/j.compscitech.2017.09.001. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Enhanced thermal conductivity of photopolymerizable composites using surface modified hexagonal boron nitride fillers. Nir Goldina, Hanna Dodiuka, and Dan Lewitusa*
a
RI PT
Author and affiliations: Department of Polymers and Plastics Engineering, Shenkar - Engineering. Art. Design,
Ramat-Gan, Israel
SC
*Corresponding author
M AN U
Abstract:
The interest in photocurable polymers has risen greatly in the past few years, in part due to the additive manufacturing revolution. Still, their widespread use is hindered by various inherent physical properties, such as thermal insulation. This work is aimed
TE D
towards the development of photopolymerizable polymer composites that are thermally conductive, while maintaining their photocurable characteristics. We developed photocurable acrylic-based photopolymer composites with hexagonal boron nitride
EP
(hBN) using the following method: pristine hBN underwent two chemical surface modifications, was added to the monomers, and the mixture then underwent radiation
AC C
curing. The success of the synthesis was verified in two ways: FTIR and XPS analyses in which the formation of carbonyl groups at the surface of the treated hBN was tracked, as well as tracking the increase in the homogeneity of the pre-polymerized solution. The addition of a reaction accelerator (o-benzoic sulfimide) to the photoinitiator system allowed for an increase of conversion percentage from ~60% to ~95%, even with high hBN loadings. Thermal conductivity (measured via modulated differential scanning calorimetry (MDSC)) increased with respect to hBN content by more than 300% when using 35wt% hBN. Young's modulus and viscosity increased with hBN content, while
ACCEPTED MANUSCRIPT coefficient of thermal expansion (CTE) decreased. We have thus developed a photocurable monomer system that is thermally conductive and applicable in various radiation curing processes.
RI PT
Key words: A. Particle-reinforced composites. B. Thermal properties. D. Differential scanning calorimetry (DSC). C. Modeling. Radiation curing
SC
Main Text:
M AN U
1. Introduction
Efficient thermal dissipation has become imperative to the advancement of many fields, including the electronics [1–3] and mold making [4] industries. Traditional thermally conductive materials, such as metals and ceramics, have several disadvantages that make them unfavorable for many applications. They are both heavy,
TE D
and their molding/shaping processes are time-consuming and costly. Moreover, these materials' inability to achieve extremely complex shapes often makes them impractical
EP
[5].
Radiation-cured polymeric materials have many applications, some of which would
AC C
greatly benefit from enhanced thermal conductivity. In the coating industry, for example, a decrease in the difference of thermal expansion between the substrates and coatings may allow for better adhesion [6–8]. Additive manufacturing (commonly known as "3D printing") may allow for inexpensive production of finely detailed items that had been traditionally made of metals or ceramics, with limited detail [9]. Similarly, the field of electronic circuitry [10–13] uses photolithography quite heavily, and could certainly benefit from better heat dissipation. Plastics are inherent thermal insulators, as most have a thermal conductivity value of 0.19 – 0.35 W/mK [14]. Thermal 2
ACCEPTED MANUSCRIPT conductivity enhancement is not a new concept [1,15–17], and thermally conductive plastics have been offered commercially for quite some time. These plastics are mostly achieved through the incorporation of high loadings of thermally conductive fillers into a polymer [18–21]. While this addition allows for the resulting composite to be
suffers from decreased mechanical properties [22].
RI PT
thermally conductive, it also yields a much heavier, more viscous product that often
Successful fabrication with methods using irradiation curing of photosensitive
SC
monomers (e.g. stereolithography 3D printing) places much significance on the
M AN U
viscosity of the polymer, its ability to photo-cure effectively, and the stability of the additive-monomer solution [23,24]. When incorporating thermally conductive fillers, however, the filler "shields" the polymer as the thickness increases, thus reducing the depth of cure [25,26]. The phenomenon is even more pronounced at high loadings. This factor, together with the oxygen sensitive radiation curing mechanism [27–30], may be
TE D
the reason that most thermal conductivity studies of thermoset composites have been conducted using thermally cured thermosets, like epoxy, polybenzoxazine, polyimide,
EP
etc. [1,15,16,18,31–35].
There has therefore been much focus on reducing the loading of the thermally
AC C
conductive fillers as much as possible. One method with which this is achieved uses compatibilized fillers [17,18,36]. Chemical compatibilization is prevalent for both carbon-based and ceramic fillers. In hexagonal boron nitride (hBN), surface modification is a rather popular route, as there are hydroxyl and amine groups on the surface of the particle [37,38]. Yu et al. [38], for example, compatibilized hBN using a hyperbranched polymer to slightly increase the thermal conductivity and mechanical properties of epoxy-based composites. Likewise, Hou et al.[39] modified the surface of hBN with a silane treatment, achieving the high thermal conductivity of 1.178 W/mK 3
ACCEPTED MANUSCRIPT while using only 30% by weight modified hBN in epoxy resin. Quite recently, Wu et al. [35] showed a 263% increase in thermal conductivity for styrene-butadiene rubber while using boron nitride nanosheets compatibilized using a silane coupling agent. In this study, we produced thermally conductive photopolymerizable polymers.
RI PT
Hexagonal boron nitride was used as the conductive filler, and its capacity to undergo chemical surface modifications was demonstrated. By using a hybrid initiator system,
we were able to increase the percentage of hBN added to the composite material, while
SC
still maintaining a high conversion percentage and photo-curing ability. Later, we
M AN U
evaluated the effect of different hBN compositions on the thermal and mechanical properties (thermal conductivity, coefficient of thermal expansion, and modulus of elasticity) of radiation-cured acrylic composites with different filler loadings. The results were then compared to relevant mathematical models.
TE D
2. Experimental
2.1.hBN surface modifications
hBN (5µm, M.K. Impex Corp.) was treated in the following ways: γ-Methacryloxypropyltrimethoxysilane, (MPTS): A methacryloxy-
EP
(i)
AC C
terminated silane coupling agent (Dynasylan® MEMO, Dynasylan) was added to 250 mL of water (1% v/v). Acetic acid (100%, Merck) was added dropwise to the solutions until a pH of 4 was reached. Approximately 7.5 gr of hBN was then added to the mixture while stirring. The reaction was stirred at room temperature for 24 hours. The mixture was then vacuumfiltered using a 2.5 µm mesh filter paper and washed with cold water 5 times. The treated particles were then dried in a vacuum oven overnight at 80°C.
4
ACCEPTED MANUSCRIPT (ii)
Carbodiimide treatment (CDI): 5 gr of hBN was added to 250 mL of THF (technical, Bio-Lab) in a round-bottomed flask, and stirred continuously. 195 mg of DMAP (4-dimethylaminopyridine, Sigma-Aldrich), 250 µL of DIC (N,N'-diisopropylcarbodiimide, Sigma-Aldrich) and 1 mL of acrylic acid
RI PT
(99%, Acros Organics) were then added, the flask was covered, and the
reaction was stirred for 24 hours at room temperature. The mixture was then vacuum-filtered using 2.5 µm mesh filter paper and washed with cold water
M AN U
2.2.Composite sample preparation
SC
5 times. The treated particles were then dried in an oven overnight at 110°C.
A mixture of thermosetting acrylic monomers was obtained from Stratasys Ltd, and used as the matrix. Treated and untreated hBN were added separately at loadings of 5%, 15%, 25% and 35% by weight into solutions with the acrylic monomer mixture and
TE D
with 2,2-Dimethoxy-2-phenylacetophenone (Sigma-Aldrich), phenylbis(2,4,6trimethylbenzoyl)phosphine oxide (Sigma-Aldrich), and o-benzoic sulfimide (Sigma-
EP
Aldrich). All mixtures were sonicated in a bath sonicator (Transsonic TS540, Elma) for 30 minutes to break up aggregates and to better disperse the filler within the mixture
AC C
(where viscosity permitted). The 25wt% mixtures were too viscous to sonicate, and were therefore mixed with a vortex mixer (Velp Scientifica) for 1 minute. With the exception of the 35wt% mixtures, all samples were poured into silicone molds after sonication. The 35wt% mixtures were highly viscous and had a paste-like consistency, and were therefore vortexed and then transferred manually to the mold using a spatula. Each mixture was then covered and rapidly radiation-cured using a high-intensity UV curing lamp (SB-100PC, Mercury 100W bulb, Spectroline). The UV lamp was situated 10 cm above the molds, and was pre-warmed for two minutes before curing. The 5
ACCEPTED MANUSCRIPT samples were cured for 2 minutes from one side, for 1 minute from the other side, and then removed from under the lamp and from their silicone molds. They were then left to complete their cure cycle at room temperature while exposed to air. Finally, the samples were cut to size for the required tests.
2.3.1
RI PT
2.3.Characterization methods
Modulated Differential Scanning Calorimetry (MDSC).
SC
Thermal conductivity measurements were taken using MDSC (DSC Q200, TA Instruments) in a modified procedure based on ASTM E 1952. TA Instruments has
M AN U
shown that this procedure is effective and accurate up to 4 W/mK [40]. Samples were cut into a right cylindrical shape at two thicknesses (~0.4 mm and over 3 mm), and placed directly onto the DSC sensor, after the sensor was moistened with a small amount of silicone oil. The reversing heat capacity was recorded after 15 minutes of a
TE D
quasi-isothermal run at 25°C with an amplitude of 0.5°C and an oscillation time of 20 seconds in a nitrogen atmosphere. The reversing heat capacities of the thin and thick disks (specific heat capacity,
, and apparent heat capacity, , respectively) were
EP
recorded. These heat capacities were then inserted into a simplified form of Fourier's
AC C
law of heat conduction using the following equation: =
8∙ ∙ ∙ ∙
∙
(1)
where L is the length of the thick sample, M is the mass of the thick sample, d is the diameter of both samples, and P is the oscillation time of the MDSC. 2.3.2
Modeling Thermal Conductivity Thermal conductivity was modeled using three existing models:
6
ACCEPTED MANUSCRIPT The Maxwell model [41] assumes a binary mixture with non-interacting spherical filler particles in small concentrations. The effective thermal conductivity is expressed as
where
+ 2( − −( −
) )
(2)
RI PT
+2 +2
=
is the thermal conductivity of the polymer (base fluid),
is the volumetric fraction of filler.
SC
conductivity of the filler (particle), and
is the thermal
The Nielsen Model [42], also known as the Lewis-Nielsen Model, is a modified
M AN U
version of a model predicting the modulus of composites, taking into account both the shape and packing of the filler in the polymer. The model is written as follows: =
1+ 1−
TE D
where
∙
−1
EP
=
AC C
(3)
+1
1− =1+!
"
"
, (4)
# , (5)
and A is a function of the geometry of the particles, and
"
is the maximum packing
fraction (the maximum amount of filler that still allows for a continuous matrix). The Agari-Uno [43] model is based on a generalization of the parallel and series models, and assumes that the dispersion is random within the material. It introduces two empirical factors,
%
and
.
% describes
the effect that the particles have on the
7
ACCEPTED MANUSCRIPT secondary structure of the polymeric matrix, while
describes the ease with which the
particles form conductive paths in the material [12]. Their model is given by the following equation:
2.3.3
=
∙
∙ &'(
+ (1 − ) ∙ log,
%
∙
- (6)
RI PT
&'(
Dynamic Mechanical Analysis (DMA).
Modulus of elasticity measurements were conducted using a DMA (Q800, TA
SC
Instruments) in controlled-force mode. The force was ramped from an initial force of 0.01N to 18N at a rate of 1N/min, at 298K and in an atmosphere of air. Coeffcient of
M AN U
thermal expansion (CTE) measurements [44] were measured using a DMA (Q800, TA Instruments) in multi-frequency (stress) mode. 0.4 mm thick rectangular samples were ramped from a temperature of 273K to 423K at a rate of 3K/min, with a force of 0.05N, and a frequency of 10Hz. The initial force was 0.01N, the initial amplitude was 0.1µm,
TE D
and the measurements were conducted in an atmosphere of air. Data for CTE measurements was taken from 283K to 303K. X-Ray Photoelectron Spectroscopy (XPS).
EP
2.3.4
XPS spectra were recorded with an X-Ray photoelectron spectrometer (Axis
AC C
Ultra, Kratos Analytical LTD) using an Al Kα monochromatic radiation X-ray source (1486.7eV). Data was collected and analyzed using a Casa XPS and Vision Data processing program. High-resolution XPS spectra were collected with a takeoff angle of 90°. Vacuum condition in the chamber was held at 1.9 × 1034 Torr for the C 1s, O 1s, N 1s, B 1s and Si 2p levels with pass energy of 20 and 0.1 eV step size. The binding energies were calibrated using C 1s peak energy as 285.0 eV.
8
ACCEPTED MANUSCRIPT 2.3.5
Fourier Transform Infrared Spectroscopy - Attenuated Total Reflectance (FTIRATR). The infrared spectra were obtained with an FTIR spectrometer (Alpha-P,
Bruker), using single-bounce ATR with a diamond crystal, and by collecting and
RI PT
averaging 24 scans at the range of 400 to 4000 cm-1 with a resolution of 4 cm-1. A heavy glass pane was used to press the samples against the ATR crystal, and liquid silane was dripped directly onto the crystal. To calculate degree of conversion for the acrylic
SC
polymer, ATR curves were converted to FTIR curves, and the degree of conversion was
equation: 5' = 100 × 61 −
7'89:&;<= >'&?9=8 = >=: @ (7) 7'89:&;<= 9'7'9=8 = >=:
Viscosity measurements.
TE D
2.3.6
M AN U
then calculated in reference to the normalized peak of the monomer using the following
Qualitative viscosity measurements were modeled after flow cup measurements such as Ford cup viscosity [45]. Liquid materials were poured into a 1mL syringe
EP
without a plunger, and then allowed to flow freely. This process was video-recorded and then analyzed in slow-motion in order to determine the amount of time taken for the
AC C
materials to flow. 2.3.7
Thermogravimetric Analysis (TGA). Thermogravimetric analysis was performed using a high-resolution TGA (Q500,
TA instruments). The samples were heated at a rate of 30°C/min to 600°C at a resolution of 4°C in a nitrogen atmosphere. 2.3.8
High-Resolution Scanning Electron Microscopy (HR-SEM).
9
ACCEPTED MANUSCRIPT Field emission HR-SEM (Gemini Ultra Plus, Zeiss) with a low voltage of 1eV and a working distance of 4.9 mm was used to observe the surface of the synthesized composites. The specimens were not coated, and the SEM images were magnified up to x10000.
RI PT
3. Results and Discussion. 3.1.Surface chemistry characterization.
SC
The hBN surface chemistry was evaluated by both FTIR-ATR and XPS. FTIRATR analysis revealed that hexagonal boron nitride has two extremely prominent peaks
M AN U
at approximately 800 cm-1 and 1360 cm-1 (Figure 1), which can be ascribed to its inplane and out-of-plane ring stretching vibrations, respectively [46,47]. In addition to these peaks, a shoulder can be seen at approximately 925 cm-1.
When the MPTS treatment was applied, the two main peaks shifted to yield
TE D
peaks at 787 cm-1 and 1358 cm-1. In addition, a new peak appeared at 1727 cm-1, implying a carbonyl group on the surface, which matches the carbonyl peak of the MEMO silane (shown in the supplementary information). This finding suggests that the
EP
silane was covalently bonded to the hBN surface, with the peak's small size corresponding to the small amount of hBN edge defects available for bonding [47].
AC C
Slight shoulders can be seen adjacent to the main 1358 cm-1 hBN peak at about 1000 – 1150 cm-1, a region typically associated with Si-O [48]. This further supports the assumption that a covalent bond was formed between the silanol and the hBN surface. FTIR-ATR analysis of the carbodiimide-treated (CDI) hBN showed a shift of the two main peaks to 774 cm-1 and 1347 cm-1. Furthermore, three new peaks formed at 1651 cm-1, 1564 cm-1, and 1193 cm-1 (as a shoulder). The peak at 1193 cm-1 can be ascribed to the C-N stretch of the amide [49], while the peaks at 1651 cm-1 and 1564
10
ACCEPTED MANUSCRIPT cm-1 are slightly more ambiguous. These peaks can be attributed either to the amide carbonyl group (1651 cm-1: C=O stretch, 1564 cm-1: N-H bend) or to the C=C double bond, however they more likely indicate the amide carbonyl, as the carbonyl peak is usually stronger than that of the double bond. The fact that there is only one carbonyl
RI PT
peak indicates that only the amine groups on hBN's surface reacted, since they are both more prevalent and more basic than the hydroxyl groups [37]. In similar experiments, in which pyridine was used as an activating basic agent or when the reaction was run for
SC
96 hours (instead of 24), the same results were found, indicating that the free hydroxyl
1 0.9 0.8 0.7 0.6 0.5
1740
1700
1660
1620
0.3 0.2 0.1 0 4000
1600
1575
3600
TE D
0.4
M AN U
group on the hBN is not reactive under the conditions explored [50,51].
3200
1550
2800
EP
CDI
2400
MPTS
2000
1600
1200
800
400
Untreated
AC C
Figure 1. FTIR-ATR plots of untreated hBN (dotted), MPTS hBN (light gray) and CDI hBN (black). The insets display enlargements of the regions between 1550 - 1600 cm-1, and 1620 - 1740 cm-1
The results of all XPS tests were similar to one another (Figure 2), with the exception of the MPTS-treated sample which showed two additional peaks resulting from the added Si (2p). The untreated hBN exhibited a carbon peak at its surface,
11
ACCEPTED MANUSCRIPT possibly as a result of either a small amount of carbon-containing contaminant, or an
RI PT
artifact caused by the sample holder
CPS
N (1s)
B (1s)
O (1s)
C (1s)
1000
800 CDI
600
400
200
M AN U
1200
SC
Si (2p)
MPTS
0
Untreated
Figure 2. XPS survey plots of untreated hBN (dotted), MPTS hBN (light gray) and CDI
TE D
hBN (black)
The deconvoluted C (1s) peaks provided further insight into the material's surface. Both the MPTS- and CDI-treated hBN showed a C=O peak within their
EP
deconvoluted C (1s) peaks, suggesting the addition of a carbonyl group to the surface of the hBN. This information is best viewed in percentages, as shown in Table 1.
AC C
Table 1. Deconvoluted C(1s) peaks of all hBN samples (in precentages) Adventitious
C-O bond
C=O bond
Untreated
80.79%
19.21%
-
MPTS
71.21%
17.80%
10.99%
CDI
72.43%
25.44%
2.13%
12
ACCEPTED MANUSCRIPT From these results, it can be concluded that the MPTS was more successful than CDI as a surface treatment, especially when considering that the carbonyl group is a much larger section of the CDI treatment than of the MPTS treatment. Figure 3 shows the results of a thermogravimetric analysis performed for the
RI PT
different hBN powders. The pristine hBN showed no more than a ~0.5% weight loss
when heated to 600°C. The MPTS-treated hBN, however, showed a 1.6% decrease in weight, suggesting that an organic layer had been previously present. The main
SC
temperature decrease was recorded at approximately 441.2°C, and was confirmed by the
M AN U
derivative weight change curve. Since this temperature is above the boiling point of the original MEMO silane, we concluded that the organic layer was the result of a covalent bond between the silane and the hBN surface, rather than simple adsorption. These findings concur with other thermogravimetric findings of silane-treated hBN [52,53]. The CDI-treated hBN showed a 4.8% weight decrease, with a main peak on the
TE D
derivative weight change curve at approximately 177°C, again hinting at a new covalent
AC C
EP
bond between the acrylic acid and the hBN surface.
13
ACCEPTED MANUSCRIPT 101
99 98 97
RI PT
Weight percent (%)
100
96 95 94 150
225
300
375
Temperature (°C) MPTS
525
600
Untreated
M AN U
CDI
450
SC
75
Figure 3. Weight percent loss vs. temperature plots of untreated hBN (dotted), MPTS hBN (light gray) and CDI hBN (black)
The combined results of the chemical analyses allow for an estimate of the
TE D
chemical modifications made to the pristine hBN, using the two methods discussed. A
AC C
EP
schematic representation can be seen in Scheme 1.
Pristine hBN
CDI hBN
MPTS hBN
Scheme 1. Schematic representation of the chemical modifications made to pristine hBN
3.2.Uncured monomer mixture characterization.
14
ACCEPTED MANUSCRIPT Viscosity measurements using a modified "Ford Cup" method (measured in seconds) were performed on the neat monomer as well as on 5% and 10% (w/w) hBNloaded samples. Higher concentrations of hBN proved too viscous to be poured into the syringe used. Table 2 details the viscosities found.
Untreated hBN
MPTS hBN
CDI hBN
0.69s
5wt%
0.86s
0.82s
0.83s
10wt% 1.26s
1.10s
1.30s
M AN U
SC
0wt%
RI PT
Table 2. Viscosity measurement of different formulations of treated and untreated hBN
From Table 2 it is clear that the addition of hBN to the neat acrylic monomer increased the viscosity of the mixture. The smaller increases in viscosity seen with
TE D
MPTS-modified samples possibly indicate superior compatibility between the filler and the matrix when compared to CDI. The monomer's viscosity is an important factor in applications that rely on the use of thin liquids, and as such, future research in the field
EP
is required to minimalize this increase in viscosity with the addition of hBN [54,55].
AC C
Dispersion longevity was evaluated visually for 1wt% hBN dispersions in the uncured monomer. The dispersions were sonicated for 45 minutes, set aside for various amounts of time, and photographed periodically. Figure 4 depicts the changes in the dispersions over a 12-day period.
15
SC
RI PT
ACCEPTED MANUSCRIPT
M AN U
Figure 4. Treated and untreated hBN dispersions in an uncured monomer over the course of 12 days
Based on these visual results, the MPTS treatment seemed to allow for the least amount of precipitation. The CDI treatment showed slightly better dispersion than
TE D
untreated hBN. A similar approach to stabilizing mixtures has been used extensively with CNTs, wherein functional groups are attached to surface defects on the CNTs. This method has been shown to achieve enhanced stability of dispersion in polymeric
AC C
EP
matrices [56–58].
3.3.Degree of Conversion
16
ACCEPTED MANUSCRIPT 1 0.9 0.8 0.7 0.6 0.5
0.3 0.2
0 4000
1800
1700
3600
1600 3200
Monomer
1500
1400
2800
1300
2400
2000
1600
1200
800
400
SC
0.1
RI PT
0.4
No accelerator
Accelerator, surface
Accelerator, bulk
M AN U
Figure 5. FTIR transmittance of the neat monomer and the polymer when cured under different conditions. In the inset, the region between 1300 and 1800 cm-1 is enlarged. The acrylic monomer was radiation-cured two separate ways: using only 2,2Dimethoxy-2-phenylacetophenone, and using the two photoinitiators and the curing
TE D
accelerator. In order to neutralize the two dominant hBN peaks in the FTIR graph, the polymer was cured without hBN. When the curing accelerator was used, both the surface and the bulk of the cured polymer were examined for degree of conversion
EP
calculations. Figure 5 depicts the FTIR graphs of the polymers cured in different
AC C
conditions. In addition, Table 3 shows the examination of the double bond peak at ~1406 cm-1 using the carbonyl peak at ~1721 cm-1 as an internal standard [59,60]. Table 3. Absolute and normalized intensties of carbonyl and double bond peak areas of different curing procedures Carbonyl peak ~1721
C=C peak ~1406 cm-1
cm-1 Absolute
Degree of conversion
Normalized
Absolute
Normalized
(%)
17
ACCEPTED MANUSCRIPT (%)
Int.
(%)
Monomer
3.0187
100
1.2003
39.8
-
No acc.
3.4247
100
0.5354
15.6
60.7
Acc.,
1.5864
100
0.0799
5
87.3
0.2997
100
0.0054
1.8
Surface Acc., Bulk
RI PT
Int.
95.5
SC
The cured polymers showed a distinctly higher degree of conversion when the curing
M AN U
accelerator was used, both at the surface and within the material, going from approximately 60% of conversion to more than 95%. 3.4.Cured composites characterization.
0.8
TE D
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
5
10
AC C
0
EP
Thermal Conductivity (W/mK)
0.9
15
20
25
30
35
40
%hBN by weight CDI
MPTS
Untreated
Figure 6. Thermal conductivity of Acrylic/hBN composites. For clarity, values are shown without error bars Figure 6 depicts the thermal conductivities of untreated and treated hBNs as a function of hBN weight fraction. Within the compositions studied, the influence of the surface
18
ACCEPTED MANUSCRIPT treatments seems negligible. However, as the hBN content increased in the synthesized composites, the MPTS-treated hBN showed a slight increase in thermal conductivity as compared to the other treated hBNs. This may be as a result of the enhanced compatibility between the MPTS and the polymeric matrix [17], however the difference
RI PT
in values is not pronounced enough to show a clear trend. The fact that all plots in
Figure 6 can be fitted rather well with linear trend lines in this region suggests that the hBN content in these compositions is too low to fully achieve thermally conductive
SC
paths in the composite material, thus limiting their thermal conductivity [14].
M AN U
Predictions regarding thermal conductivity were made with three known models (Maxwell, Nielsen, Agari & Uno), while using calculated volume fractions and the data gathered on pristine hBN. Figure 7 shows the thermal conductivity of unmodified hBN/acrylic composites from 1% by weight up to 35%. As can be seen in the graph, the Maxwell model, which does not take particle shape into account, is less accurate for the
TE D
non-spherical hexagonal boron nitride particles. The Nielsen model, on the other hand, achieved better results using
= 10 and
"
= 0.637. The shape parameter, A,
EP
signifies that the hBN particles are non-spherical, and have a higher aspect ratio. The values found in these calculations are close to those given by other authors [61]. The
AC C
maximum packing fraction,
",
is in accordance with three-dimensional random close
orientation of irregularly shaped fillers [62]. In order to plot the Agari-Uno model, the factors
%
= 1.1;
= 0.9193 were calculated using measurements made for the
untreated hBN. The very high
%
value appears to hint that the polymer has a closer
pack after the addition of hBN, while the high
value signifies that hBN is able to
efficiently form conductive paths when sufficient loading is used in the formulation. These findings indicate that the thermal conductivity of photopolymerizable composites does not behave differently from that of other thermally polymerized materials. 19
ACCEPTED MANUSCRIPT 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0
0.2
0.4
0.6
0.8
1
SC
Weight fraction of hBN
RI PT
Thermal Conductivty (W/mK)
1
Maxwell Model
Agari-Uno Model
Measured Values
M AN U
Nielsen Model
Figure 7. Thermal conductivity of composites containing different hBN concentrations according to three different models (Maxwell, Nielsen, and Agari-Uno) and the experimental values measured for untreated hBN. The graphs are cut-off at an arbitrary
TE D
1 W/mK value to better show models in contrast with the experimental results Because most thermal conductivity measurements are based on calculations
EP
rather than direct measurement, the results are prone to large errors. Here, two independent models fit well with the experimental results, providing much more
AC C
credence to the thermal conductivity measurement method used. In addition, the applied models have been proven to be rather effective for a variety of composite materials. Nielsen's model, for example, has been used for calculations involving boron nitride nanosheets in SBR rubber [35], hBN and secondary carbon-based fillers in poly(phenylene sulfide) [63], hBN-filled thermoplastics [64], hBN in several typed of epoxy [61], and aluminum nitride filled epoxy [20], to name a few. Agari and Uno's model has also been proven to be effective, and examples can be found for PVC and polyethylene filled with either carbon black or graphite [65], silicon-reinforced HDPE 20
ACCEPTED MANUSCRIPT [66], polystyrene filled with aluminum nitride [67], and epoxy filled with several ceramic fillers [12]. Although HRSEM images could not conclusively show improved adhesion between the matrix and the treated fillers, the CDI treatment did seem to allow for a
RI PT
more homogenic composite, while the MPTS treatment and the untreated hBN seem to create similar composite materials. HRSEM images are provided in the supplementary
SC
information (Figure 12).
The coefficients of thermal expansion of the different composites were
M AN U
investigated using DMA. The experiments described the change in sample dimension as a function of the original length, and the CTE can be then be calculated from this information. All DMA curves had two wide-ranged regions, corresponding to the glassy state and to the rubbery state, and a transition between the two [68]. However, even
TE D
within the same region, the slope was found to change as the temperature rose, which is indicative of the expected CTE change.
Because the CTE curves change continuously, the values reported in Figure 8
EP
are extracted from the range of ~ 283K – 303K, which is in the glassy region for all composites. As the filler concentration increased, the CTE generally decreased, which is
AC C
to be expected with a ceramic filler (ceramics tend to have low CTEs, while polymers usually have higher CTEs [12,69]). One of the most common models for calculating CTEs is the Schapery model
[66,70,71], which assumes that the composites are isotropic and applies extremum values based on energy principles. The model establishes upper and lower bounds for CTE composites. It is clear from Figure 8 that the measured values are well within Schapery's bounds, with the single exception of 35% by weight hBN composites, which
21
ACCEPTED MANUSCRIPT all seem to be just under the lower bound. These deviations may be the result of the hBN platelets orienting themselves perpendicular to the testing direction, as their CTE is about one order of magnitude smaller for transverse testing. 80
RI PT
60 50 40 30 20 10 0 0%
5%
10%
15%
20%
25%
CDI
MPTS
30%
M AN U
hBN content in % by weight
SC
CTE (μm/mK)
70
Untreated
Upper bound
35%
40%
Lower bound
Figure 8. CTE of the different composites fabricated at 298K compared to the
TE D
theoretical Schapery's model
The mechanical properties of the fabricated composites were investigated using their stress-strain curves, generated from DMA measurements. Overall, the samples
EP
showed trends similar to one another when their Young's Moduli were examined using a
AC C
linear trend line at the beginning of their stress-strain curves (Figure 9).
22
ACCEPTED MANUSCRIPT 7000 6000
E (MPa)
5000 4000 3000
1000 0 5%
10%
15%
20%
25%
hBN content in % by weight CDI
MPTS
Untreated
30%
35%
40%
SC
0%
RI PT
2000
M AN U
Figure 9. Modulus of elasticity of the different composites at varying concentrations Small amounts of boron nitride within polymeric matrices have been shown to improve mechanical strength [69]. As a result, the slight increase in Young's modulus for the 5% by weight composite seen in Figure 9 is to be expected. The decreased
TE D
modulus at 15% by weight is most likely the result of hBN aggregates slipping over one another, causing the well-documented "lubricating" effect [69]. The second increase,
EP
starting with approximately 25% by weight composites, is most likely due to the increased particulate matter, leading to an overall stronger, more brittle material. When
AC C
examining the stress-strain curves of the highest-filled composites for each series, all materials showed a higher Young's modulus, as expected. The untreated hBN composites exhibited embrittlement, showing small amounts of strain before failure. Both treated hBN composites, however, showed much higher elongation, i.e. improved ductility of both composites. Stress-strain curves for all 35wt% compositions and for the neat acrylic are presented in the supplementary information (Figure 13). 4. Conclusions.
23
ACCEPTED MANUSCRIPT To the best of our knowledge, this is the first investigation of thermal conductivity of a photopolymerizable thermosetting material. We synthesized photocurable composites through the addition of a small amount of curing accelerator and radiation polymerization with high loadings of thermally conductive hBN. The physical
RI PT
properties of the polymerized composites were not significantly affected by hBN
covalent functionalization using silane (MPTS) and carbodiimide reaction (CDI).
However, the effects of this functionalization on the prepolymer stability and viscosity
SC
were prominent. These finding suggest that functionalization may be critical when
considering the shelf life of the examined compositions. Overall, the addition of hBN
M AN U
resulted in a significant enhancement of thermal conductivity (from 0.19 W/mK up to 0.84 W/mK), while still allowing the composite to cure using only irradiation. In addition, the mathematical calculations fit well with the relevant models (Nielsen, Agari & Uno).
TE D
5. Supplementary information.
Further information regarding the FTIR-ATR results and peak assignments for
EP
MPTS, the thermal conductivity results for each composition of acrylic/hBN (including omitted error bars), the HRSEM images of the 35wt% compositions, and the stress-
AC C
strain curves for all 35wt% compositions as well as for the unfilled material are included in the supplementary information section. 6. References.
[1]
H. Ishida, S. Rimdusit, Very high thermal conductivity obtained by boron nitridefilled polybenzoxazine, Thermochim. Acta. 320 (1998). doi:10.1016/S00406031(98)00463-8.
[2]
R. Mahajan, G. Chrysler, Cooling a Microprocessor Chip, Proc. IEEE. 94 (2006) 24
ACCEPTED MANUSCRIPT 1476–1486. doi:10.1109/JPROC.2006.879800. [3]
G.-W. Lee, M. Park, J. Kim, J.I. Lee, H.G. Yoon, Enhanced thermal conductivity of polymer composites filled with hybrid filler, Compos. Part A Appl. Sci.
[4]
RI PT
Manuf. 37 (2006) 727–734. doi:10.1016/j.compositesa.2005.07.006. H.H. Chiang, C.A. Hieber, K.K. Wang, A unified simulation of the filling and postfilling stages in injection molding. Part I: Formulation, Polym. Eng. Sci. 31
[5]
SC
(1991) 116–124. doi:10.1002/pen.760310210.
K.P. Karunakaran, S. Suryakumar, V. Pushpa, S. Akula, Low cost integration of
M AN U
additive and subtractive processes for hybrid layered manufacturing, Robot. Comput. Integr. Manuf. 26 (2010) 490–499. doi:10.1016/j.rcim.2010.03.008. [6]
A. Bjorneklett, L. Halbo, H. Kristiansen, Thermal conductivity of epoxy adhesives filled with silver particles, Int. J. Adhes. Adhes. 12 (1992) 99–104.
[7]
TE D
doi:10.1016/0143-7496(92)90030-Y.
R.G. Humfeld, D.A. Dillard, Residual stress development in adhesive joints to
thermal
cycling,
J.
Adhes.
65
(1998)
277–306.
EP
subjected
doi:10.1080/00218469808012250. G. Viana, M. Costa, M. Banea, L. da Silva, A review on the temperature and
AC C
[8]
moisture degradation of adhesive joints, Proc. Inst. Mech. Eng. Part L J. Mater. Des. Appl. 0 (2016) 1–14. doi:10.1177/1464420716671503.
[9]
H. Lipson, M. Kurman, Fabricated: The new world of 3D printing, 2013.
[10] C. Zweben, Ultrahigh-thermal-conductivity packaging materials, Semicond. Therm. Meas. Manag. Symp. 2005 IEEE Twenty First Annu. IEEE. (2005) 168– 174. doi:10.1109/stherm.2005.1412174. 25
ACCEPTED MANUSCRIPT [11] A. a Balandin, S. Ghosh, W. Bao, I. Calizo, D. Teweldebrhan, F. Miao, C.N. Lau, Superior Thermal Conductivity of Single-Layer Graphene 2008, Nano Lett. 8 (2008) 902–907. doi:10.1021/nl0731872. [12] C.P. Wong, R.S. Bollampally, Thermal Conductivity , Elastic Modulus , and
RI PT
Coefficient of Thermal Expansion of Polymer Composites Filled with, 74 (1999) 3396–3403.
Future
Directions,
Proc.
IEEE.
94
M AN U
doi:10.1109/JPROC.2006.879796.
SC
[13] R. Prasher, Thermal Interface Materials: Historical Perspective, Status, and (2006)
1571–1586.
[14] Z. Han, A. Fina, Thermal conductivity of carbon nanotubes and their polymer nanocomposites:
A
review,
Prog.
Polym.
Sci.
36
(2011)
914–944.
doi:10.1016/j.progpolymsci.2010.11.004.
TE D
[15] T.-L. Li, S.L.-C. Hsu, Enhanced thermal conductivity of polyimide films via a hybrid of micro- and nano-sized boron nitride., J. Phys. Chem. B. 114 (2010)
EP
6825–9. doi:10.1021/jp101857w.
[16] A. Yu, P. Ramesh, M.E. Itkis, E. Bekyarova, R.C. Haddon, Graphite
AC C
Nanoplatelet-Epoxy Composite Thermal Interface Materials, J. Phys. Chem. C. 111 (2007) 7565–7569. doi:10.1021/jp071761s.
[17] I. a. Tsekmes, R. Kochetov, P.H.F. Morshuis, J.J. Smit, Thermal conductivity of polymeric composites: A review, 2013 IEEE Int. Conf. Solid Dielectr. (2013) 678–681. doi:10.1109/ICSD.2013.6619698. [18] Y. Xu, D.D.L. Chung, Increasing the thermal conductivity of boron nitride and aluminum nitride particle epoxy-matrix composites by particle surface
26
ACCEPTED MANUSCRIPT treatments,
Compos.
Interfaces.
7
(2000)
243–256.
doi:10.1163/156855400750244969. [19] P. Bujard, Thermal Conductivity of Boron Nitride Filled Epoxy Resin: Temperature Dependence and Influence of Sample Preparation, Intersoc. Conf.
RI PT
Therm. Phenom. Fabr. Oper. Electron. Components I-THERM ’88. (1988) 41– 49.
SC
[20] P. Bujard, J.P. Ansermet, Thermally conductive aluminium nitride-filled epoxy resin for electronic packaging, Fifth Annu. IEEE Semicond. Therm. Temp. Meas.
M AN U
Symp. (1989) 126–130.
[21] S. Ino, T. Shiobara, Thermal Conductivity Of Molding Compounds For Plastic Packaging, IEEE Trans. Components Packag. Manuf. Technol. Part A. 17 (1994) 527–532. doi:10.1109/95.335037.
TE D
[22] Z. Pu, J.E. Mark, J.M. Jethmalani, W.T. Ford, Effects of Dispersion and Aggregation of Silica in the Reinforcement of Poly(methyl acrylate) Elastomers,
EP
Chem. Mater. 9 (1997) 2442–2447. doi:10.1021/cm970210j. [23] K. V. Wong, A. Hernandez, A Review of Additive Manufacturing, ISRN Mech.
AC C
Eng. 2012 (2012) 1–10. doi:10.5402/2012/208760. [24] R. Singh, Process capability study of polyjet printing for plastic components, Evol. Ecol. 25 (2011) 1011–1015. doi:10.1007/s12206-011-0203-8.
[25] J.G. Leprince, P. Leveque, B. Nysten, B. Gallez, J. Devaux, G. Leloup, New insight into the “depth of cure” of dimethacrylate-based dental composites, Dent. Mater. 28 (2012) 512–520. doi:10.1016/j.dental.2011.12.004. [26] A.C. Shortall, How light source and product shade influence cure depth for a 27
ACCEPTED MANUSCRIPT contemporary
composite,
J.
Oral
Rehabil.
32
(2005)
906–911.
doi:10.1111/j.1365-2842.2005.01523.x. [27] J.V. Koleske, Radiation Curing of Coatings, ASTM International, 2002.
RI PT
[28] C. Decker, UV-radiation curing chemistry, Pigment Resin Technol. 30 (2001) 278–286. doi:10.1108/03699420110404593.
[29] P. Gloeckner, Radiation Curing (European Coatings Tech Files), 2008.
SC
[30] S. Adanur, Y. Arumugham, Characteristics of Ultraviolet Cured Glass-Epoxy Textile Composites: Part 1: Experimental Procedures and Testing, J. Ind. Text.
M AN U
32 (2002) 93–106. doi:10.1106/152808302031090.
[31] J.J. Park, D.M. Park, J.H. Youk, W.R. Yu, J. Lee, Functionalization of multiwalled carbon nanotubes by free radical graft polymerization initiated from photoinduced
surface
groups,
Carbon
N.
Y.
48
(2010)
2899–2905.
TE D
doi:10.1016/j.carbon.2010.04.024.
[32] F.H. Gojny, J. Nastalczyk, Z. Roslaniec, K. Schulte, Surface modified multi-
EP
walled carbon nanotubes in CNT/epoxy-composites, Chem. Phys. Lett. 370
AC C
(2003) 820–824. doi:10.1016/S0009-2614(03)00187-8. [33] R. Kochetov, Thermal and Electrical Properties of Nanocomposites, Including Material Processing, 2012.
[34] P.C. Irwin, Y. Cao, A. Bansal, L.S. Schadler, Thermal and mechanical properties of polyimide nanocomposites, 2003 Annu. Rep. Conf. Electr. Insul. Dielectr. Phenom. (2003) 1–4. [35] X. Wu, H. Liu, Z. Tang, B. Guo, Scalable fabrication of thermally conductive elastomer / boron nitride nanosheets composites by slurry compounding, 28
ACCEPTED MANUSCRIPT Compos.
Sci.
Technol.
123
(2016)
179–186.
doi:10.1016/j.compscitech.2015.12.010. [36] C.W. Nan, R. Birringer, D.R. Clarke, H. Gleiter, Effective thermal conductivity of particulate composites with interfacial thermal resistance, J. Appl. Phys. 81
RI PT
(1997) 6692–6699. doi:10.1063/1.365209.
[37] M.T. Huang, H. Ishida, Investigation of the boron nitride / polybenzoxazine
doi:10.1002/(SICI)1099-0488(19990901)37.
SC
interphase, J. Polym. Sci. Part B Polym. Phys. 37 (1999) 2360–2372.
M AN U
[38] J. Yu, X. Huang, C. Wu, X. Wu, G. Wang, P. Jiang, Interfacial modification of boron nitride nanoplatelets for epoxy composites with improved thermal properties,
Polymer
(Guildf).
53
(2012)
471–480.
doi:10.1016/j.polymer.2011.12.040.
TE D
[39] J. Hou, G. Li, N. Yang, L. Qin, M.E. Grami, Q. Zhang, N. Wang, X. Qu, Preparation and Characterization of Surface Modified Boron Nitride Epoxy Composites with Enhanced Thermal Conductivity, RSC Adv. 4 (2014) 44282–
EP
44290. doi:10.1039/C4RA07394K.
AC C
[40] E. Verdonck, G. Dreezen, Thermal Conductivity Measurements of Conductive Epoxy Adhesives by MDSC, Therm. Libr. Appl. Br. TA312. (n.d.). http://www.tainstruments.co.jp/application/pdf/Thermal_Library/Applications_B riefs/TA312.PDF.
[41] C. Maxwell, James, A treatise on electricity and magnetism, (1873). [42] L.E. Nielsen, Thermal conductivity of particulate-filled polymers, J. Appl. Polym. Sci. 17 (1973) 3819–3820. doi:10.1002/app.1973.070171224.
29
ACCEPTED MANUSCRIPT [43] Y. Agari, T. Uno, Estimation on thermal conductivities of filled polymers, J. Appl. Polym. Sci. 32 (1986) 5705–5712. doi:10.1002/app.1986.070320702. [44] P.F. Rios, S. Kenig, R. Cohen, A. Shechter, The effect of carbon nanotubes on the thermal expansion isotropy of injection molded carbon fiber reinforced
RI PT
thermoplastics, Polym. Compos. 34 (2013) 1367–1374. doi:10.1002/pc.22551.
[45] ASTM International, Standard Test Method for Viscosity by Ford Viscosity Cup,
SC
(2010).
[46] R. Geick, C.H. Perry, G. Rupprecht, Normal Modes in Hexagonal Boron Nitride,
M AN U
Phys. Rev. 146 (1966) 543–547. doi:10.1103/PhysRev.146.543.
[47] K. Sato, H. Horibe, T. Shirai, Y. Hotta, H. Nakano, H. Nagai, K. Mitsuishi, K. Watari, Thermally conductive composite films of hexagonal boron nitride and polyimide with affinity-enhanced interfaces, J. Mater. Chem. 20 (2010) 2749.
TE D
doi:10.1039/b924997d.
[48] M.A. Villegas, J.M.F. Navarro, Characterization of BaOa-SiOa glasses prepared
EP
via sol-gel, J. Mater. Sci. 23 (1988) 2464–2478. doi:10.1007/BF01111904. [49] M. Habibi, R. Amrollahi, M.H.S. Alavi, Polymerization of Acrylic Acid by a 4kJ
AC C
plasma focus device, (2012) 1–4.
[50] B. Neises, W. Steglich, Simple Method for the Esterification of Carboxylic Acids,
Angew.
Chemie
Int.
Ed.
English.
17
(1978)
522–524.
doi:10.1002/anie.197805221. [51] C.A.G.N. Montalbetti, V. Falque, Amide bond formation and peptide coupling, Tetrahedron. 61 (2005) 10827–10852. doi:10.1016/j.tet.2005.08.031. [52] K. Kim, M. Kim, J. Kim, Fabrication of UV-curable polyurethane acrylate 30
ACCEPTED MANUSCRIPT composites containing surface-modified boron nitride for underwater sonar encapsulant
application,
Ceram.
Int.
40
(2014)
10933–10943.
doi:10.1016/j.ceramint.2014.03.092. [53] K. Kim, M. Kim, Y. Hwang, J. Kim, Chemically modified boron nitride-epoxy
RI PT
terminated dimethylsiloxane composite for improving the thermal conductivity, Ceram. Int. 40 (2014) 2047–2056. doi:10.1016/j.ceramint.2013.07.117.
SC
[54] B. Derby, Inkjet Printing of Functional and Structural Materials: Fluid Property Requirements, Feature Stability, and Resolution, Annu. Rev. Mater. Res. 40
M AN U
(2010) 395–414. doi:10.1146/annurev-matsci-070909-104502.
[55] S.D. Hoath, O.G. Harlen, I.M. Hutchings, Jetting behavior of polymer solutions in drop-on-demand inkjet printing, J. Rheol. (N. Y. N. Y). 56 (2012) 1109. doi:10.1122/1.4724331.
coupling
TE D
[56] P.C. Ma, J. Kim, B.Z. Tang, Functionalization of carbon nanotubes using a silane agent,
Carbon
N.
Y.
44
(2006)
3232–3238.
EP
doi:10.1016/j.carbon.2006.06.032.
[57] P.C. Ma, J.K. Kim, B.Z. Tang, Effects of silane functionalization on the
AC C
properties of carbon nanotube/epoxy nanocomposites, Compos. Sci. Technol. 67 (2007) 2965–2972. doi:10.1016/j.compscitech.2007.05.006.
[58] P.-C. Ma, N.A. Siddiqui, G. Marom, J.-K. Kim, Dispersion and functionalization of carbon nanotubes for polymer-based nanocomposites: A review, Compos. Part A
Appl.
Sci.
Manuf.
41
(2010)
1345–1367.
doi:10.1016/j.compositesa.2010.07.003. [59] Application Note AN # 103 Determination of the degree of cure of a varnish, n.d.
31
ACCEPTED MANUSCRIPT [60] S. Imazato, J.. McCabe, H. Tarumi, A. Ehara, S. Ebisu, Degree of conversion of composites measured by DTA and FTIR, Dent. Mater. 17 (2001) 178–183. doi:10.1016/S0109-5641(00)00066-X. [61] R.F. Hill, P.H. Supancic, Thermal Conductivity of Platelet-Filled Polymer
RI PT
Composites, J. Am. Ceram. Soc. 85 (2002) 851–857. doi:10.1111/j.11512916.2002.tb00183.x.
(1986) 125–140. doi:10.1002/pc.750070302.
SC
[62] D.M. Bigg, Thermally conductive polymer compositions, Polym. Compos. 7
M AN U
[63] M.O. Khan, Thermally Conductive Polymer Composites for Electronic Packaging Applications, University of Toronto, 2012.
[64] C. Raman, A. Murugaiah, B. Xiang, R. Roden, Thermally conductive but electrically insulating plastics for thermal management applications, Proc. 12th
TE D
Automot. Compos. Conf. (2012) 23.
[65] Y. Agari, T. Uno, Thermal Conductivity of Polymer Filled With Carbon
EP
Materials: Effect of Conductive Particle Chains on Thermal Conductivity., J. Appl. Polym. Sci. 30 (1985) 2225–2235. doi:10.1002/app.1985.070300534.
AC C
[66] T.K. Dey, M. Tripathi, Thermal properties of silicon powder filled high-density polyethylene
composites,
Thermochim.
Acta.
502
(2010)
35–42.
doi:10.1016/j.tca.2010.02.002.
[67] S. Yu, P. Hing, X. Hu, Thermal conductivity of polystyrene–aluminum nitride composite, Compos. Part A Appl. Sci. Manuf. 33 (2002) 289–292. doi:10.1016/S1359-835X(01)00107-5. [68] E.S.A. Rashid, K. Ariffin, C.C. Kooi, H.M. Akil, Preparation and properties of 32
ACCEPTED MANUSCRIPT POSS/epoxy composites for electronic packaging applications, Mater. Des. 30 (2009) 1–8. doi:10.1016/j.matdes.2008.04.065. [69] G. Wypych, Handbook of Fillers (3rd Edition), 2010.
on
Energy
Principles,
J.
Compos.
doi:10.1177/002199836800200308.
Mater.
RI PT
[70] R.A. Schapery, Thermal Expansion Coefficients of Composite Materials Based 2
(1968)
380–404.
SC
[71] A.M. Drews, Control of thermal expansion coefficient of a metal powder
AC C
EP
TE D
M AN U
composite via ceramic nanofiber reinforcement, University of Akron, 2009.
33