thermally conductive polymer-based composites for electronic packaging applications
Post on 08-Nov-2014
39 Views
Preview:
DESCRIPTION
TRANSCRIPT
1
Thermally Conductive Polymer-Based Composites
for Electronic Packaging Applications
AHMED MAHMOUD
A thesis submitted in partial fulfillment
of the requirements for the degree of
BACHELOR OF APPLIED SCIENCE
SUPERVISOR: DR. HANI E. NAGUIB
Department of Mechanical and Industrial Engineering
2
Abstract
The objective of this thesis was to develop an electrically-insulative material with a
high thermal conductivity for electronic packaging applications. A heat-transfer
analysis study found that a thermal conductivity of at least 10 W/m.K would be
required in order to provide a viable alternative to the current electronic packaging
and heat sink assembly. To that end, a composite comprising a poly(p-phenylene
sulphide) matrix and Boron Nitride filler was investigated. The choice of matrix was
predicated upon the need to use a resin which has a low melt viscosity and excellent
affinity with the filler to ensure that air gaps within the composite are minimized,
while the choice of filler was based upon the favourable thermomechanical properties
of Boron Nitride, as well as encouraging previous research. Initial thermal
conductivity testing as well as SEM analysis using four different grades of Boron
Nitride singled out one polymorph – PTX60 – for further study; due to its spherical
shape, it allowed for maximization of filler-packing throughout the matrix, thereby
increasing the likelihood of filler-filler contact and reducing the percolation
threshold. At 7.8 W/m.K, however, the highest obtained thermal conductivity was
well below the goal of the thesis, as well as theoretically-predicted values. It was
conjectured that this was due to high interfacial resistance resulting from
incompatibility between the Boron Nitride and the matrix. This issue was addressed
by surface-treating PTX60 with a coupling agent, which was shown in SEM images to
enhance the adhesion of Boron Nitride to the matrix, thereby reducing phonon
scattering at interfacial barriers and lattice defects, and bringing the percolation
threshold within reach. The final result was a 13% improvement in thermal
conductivity, achieving a very remarkable value of 8.8 W/m.K using 74 wt% Boron
Nitride in a poly(p-phenylene sulphide) matrix. Methods to increase this value to the
desired 10 W/m.K, including use of multi-modal Boron Nitride filler and a novel
epoxy matrix, are explored at the end of the report.
3
Table of Contents
I Introduction............................................................................................................................... 1
1.1 Heat Transfer Analysis .............................................................................................................. 2
II Literature Review ...................................................................................................................... 5
2.1 Polymer Research..................................................................................................................... 5
2.2 Boron Nitride Research ............................................................................................................ 6
2.3 Percolation Theory and Previous Research on Fillers .............................................................. 7
III Phase One: Preliminary Research using Single-Phase and Hybrid Fillers .................................... 8
3.1 Methodology .......................................................................................................................... 10
3.2 Results .................................................................................................................................... 13
3.2.1 Single-Phase Composite Results ............................................................................ 13
3.2.2 Hybrid-Filler Composite Results ............................................................................. 16
3.3 SEM Analysis .......................................................................................................................... 18
3.2.1 Methodology .......................................................................................................... 18
3.3.2 SEM Micrograph Results ........................................................................................ 19
3.3.2.1 Single-Phase Composite Micrographs ................................................... 19
3.3.2.2 Hybrid-Filler Composite Micrographs ................................................... 21
3.4 Compression Molding Pressure Analysis ............................................................................... 24
IV Phase 2: Coupling Agents ........................................................................................................ 28
4.1 Background Research ............................................................................................................. 28
4.2 Methodology .......................................................................................................................... 30
4.3 Thermogravimetric Analysis (TGA)......................................................................................... 31
4.3.1 Methodology .......................................................................................................... 31
4.3.2 Results .................................................................................................................... 32
4.3.2.1 Calibration .............................................................................................. 32
4.3.2.2 PTX60 vs. PT110 ..................................................................................... 32
4.4 Design of Experiments ........................................................................................................... 36
4.5 Thermal Conductivity Results................................................................................................. 40
4.6 SEM Analysis .......................................................................................................................... 43
V Future Research Directions ...................................................................................................... 46
5.1 Multi-Modal hBN Fillers ......................................................................................................... 46
5.2 Ishida Method ........................................................................................................................ 47
5.2 Multifunctional Fillers ............................................................................................................ 48
VI Conclusion.............................................................................................................................. 49
1
Figure 1: An Aluminum heat sink and fan
I Introduction
As electronic devices are made smaller and more
powerful, they generate more heat in a smaller
volume of space, and so the need arises to develop
optimum heat dissipation solutions. Heat sinks are
used to transfer the heat generated by the
processor or chipset to the surrounding air; these
heat sinks are traditionally made out of a metal,
usually Aluminum, fabricated into a multi-fin arrangement to maximize heat dissipation.
However, Aluminum has a high electrical conductivity, which prompts the need for
micro-processor packaging solutions to prevent the heat sink from short-circuiting the
processor. [1] These packages are usually made of a copolymer, such as epoxy, which has a
high electrical resistivity and excellent processability, but also a low thermal conductivity
– in the range of 0.1 to 0.5 W/m.°K – which reduces the overall effective heat dissipation
of the heat sink. [2]
To address this problem, researchers have experimented with composites which
could exhibit both a high thermal conductivity and a low dielectric constant, the aim
being to develop a material which would perform the function of a heat sink, while
forgoing the need for electronic packaging. They started by embedding ceramic fillers,
such as alumina (Al2O3) [3], silica (SiO2) [4], aluminum nitride (AlN) [5][6][7], and boron
nitride (BN),[8][9][10] in a polymer matrix. Among these materials, hexagonal BN (hBN)
showed the greatest potential due to its high thermal conductivity (up to 400 W/mk) and
relatively low dielectric constant, compared with SiC, Al2O3, and AlN, which suffer from
high reactivity with moisture and a poor affinity with resin. Further, hBN exhibits
excellent resistance to oxidation and chemical corrosion. [11]
For the matrix, a wide array polymers have been explored by researchers to varying
success. Two polymers in particular have proven to be good candidates: poly(p-phenylene
sulfide) (PPS) and Low-Density Polyethylene (LDPE). Both exhibit similar thermal
2
properties but have slightly varying mechanical properties owing to differences in
molecular weight and degrees of cross-linking. [12] Though they are not as good as
epoxies in terms of affinity to BN, they are the best thermoplastic choices available since
they are easily processable.
The thermal conductivity of the polymer-filler composite can be maximized by
minimizing thermal barrier resistance along the heat flow path and forming a thermal
conductive network in the composites. Once a conductive network is formed by the filler,
the composite shows percolation behaviour, i.e., the thermal conductivity increases
steeply.
The objective of this thesis was to expand on current research on thermally-
conductive polymers in order to develop prototypes a composite with a ceramic filler, in
order to achieve a high thermal conductivity but low electrical conductivity.
1.1 Heat Transfer Analysis
Heat transfer across the heat sink can be modeled by a series of heat resistors between
the processor and ambient air
representing:
1. The epoxy packaging which
has a thermal conductivity
of ~0.3 W/m.K [13]
2. The thin layer of thermal
compound used to promote
heat conductivity from the
processor; a typical silver-
based thermal compound
has a thermal conductivity
of approximately 5 W/m.K
3. The Aluminum heat sink,
the thermal conductivity of which is 237 W/m.K
4. Heat convection from the heatsink
TAmbient
Rconvection
3
For this analysis, the following assumptions are made:
• The thickness of the epoxy packaging is 1 mm and that of the thermal compound is
0.5 mm
• The height of the heat sink is 10 cm
• The processor has the same dimensions as an Intel LGA775 socket processor: 0.6 in
x 0.6 in [14]
• Heat loss is by simple convection only, and therefore the heat transfer coefficient
is given by
ℎ =�
�∆�
where:
h = heat transfer coefficient, W/(m2K) q = heat flow in input or lost heat flow , W A = heat transfer surface area, m2
ΔT = difference in temperature between the solid surface and surrounding fluid area, which is ������� − ��������
A study by Rebay et al found that the heat transfer coefficient for a typical commercial
heat sink is roughly 49W/m2K. [15] This factor is a function of the geometry of the heat
sink and not the material from which it is made.
Therefore, calculating the effective heat resistivity across a commercial heat sink, we get:
��������� = ����� + �������� ����� + ����� ����,�� ��� + �������
=!����
"���� �+
!������� �����
"������� ����� �+
!���� ����
"���� ���� �+
1
ℎ�
=0.001
0.35 × (2.3 × 10+,)+
0.0005
5 × (2.3 × 10+,)+
0.1
237 × (2.3 × 10+,)+
1
49 × (2.3 × 10+,)
= 103.4 °2
3
4
If we are to replace the current arrangement with just a single heat sink that incorporates
the electronic packaging, only two resistivies exist: one due to the composite heatsink and
the other due to convection. Therefore:
��������� =!������
"������ �+
1
ℎ�
where Lcomposite is the height of the heat sink which, as stated above, is 10 cm.
Then the required k for the composite is:
"�������� =!������
(������� −1
ℎ���� �����) �
=0.1
(103.4 −1
49 × (2.3 × 10+,))(2.3 × 10+, )
= 29.63
5. 2
This figure approximates the thermal conductivity required by the composite which
would result in the same dissipative power as the current Aluminum heatsink
arrangement. However, due to time constraints, the target keff for this thesis is set to for
10 W/m.K, which is a more realistic expectation.
5
II Literature Review
There exists a substantial body of literature on attempts to develop thermally-conductive
composites for enhanced electronic packaging applications. This section summarizes the
research that has been done on the topic in the run-up of to experimental prototyping.
2.1 Polymer Research
Most studies were carried out using a polymer or epoxy for the matrix, the reason
being that they are affordable, easily processed and have a low dielectric constant [16].
However, almost all polymers have very poor thermal conductivities due to their
semicrystalline nature; specifically, in a spherulitic polymer, the chains fold into highly
ordered crystalline lamellae, but are separated by amorphous regions that disrupt phonon
heat transport (by scattering), therefore resulting in poor thermal conductivity. [17]
Therefore, it becomes important to choose a polymer which has a low viscosity so that it
is easily processed, has a high tendency to crystallize to promote thermal conductivity, a
high service temperature to withstand the heat generated by the processor, and most
importantly, it must allow for excellent filler dispersion. Research indicates that two
particular polymers: poly(p-phenylene sulfide) (PPS), Liquid Crystal Polymer (LCP), and
polyethylene (PE) are excellent candidates based on these criteria. [18][19] Whereas PPS is
chosen for its “outstanding mechanical properties, high thermal stability, and good
solvent resistance,” LCP and PE are equally as qualified for our application; though they
have slightly lower thermal conductivity than PPS, they can be stretched and aligned to
produce a high uni-directional thermal conductivity. A recently-published and critically-
acclaimed study by Shen el al. reported achieving thermal conductivities as high as 104
W/mK using high-quality ultra-drawn polyethylene nanofibres. The high thermal
conductivity achieved is attributed to “the restructuring of the polymer chains by
stretching, which improves the fibre quality toward an ‘ideal’ single crystalline fibre.”[20]
The methods used to achieve these results, however, are expensive and time consuming,
so the focus for now will be on embedding filler materials within a PPS matrix to the
desired thermal conduction network.
6
Figure 2: Hexagonal Boron Nitride Crystal Structure
2.2 Boron Nitride Research
Boron nitride exists in several allotropic forms; most commonly, it assumes one of
two structures: hexagonal (hBN, similar to graphite), and cubic (cBN, similar to
diamond). The hexagonal form is extremely soft and has a layered structure, where Van
der Waal’s forces hold sheets of covalently bonded boron and nitrogen atoms together.
As a consequence of the layered structure, hBN
crystals are anisotropic. This anisotropy is clearly seen
in the thermal conductivity: “The in-plane thermal
conductivity [for hBN] has been measured to be more
than 300 W/mK, and is attributed to good conduction
of phonons through boron nitride bonds, whereas the
through-plane conductivity has been measured to be
only about 3 W/mK, a significant difference, most likely due to poor phonon transfer
across layers.”[29]
Numerous studies have been carried out that
confirm the ability of Boron Nitride to show
percolation in a polymer matrix, thereby resulting in a
high thermal conductivity composite. One such study
by Sato et al was able to achieve a thermal conductivity
of 7 W/mK.
Using a polyimide matrix embedded with 60%
hBN.[30] On the other hand, Yung et al were able to
achieve 20 W/mK using multimodal (hybrid) fillers of
hBN and cBN.[31] Their results are shown in figure 3 on the right.
Much higher values for composite thermal conductivity have been obtained in other
studies, one of which is briefly touched on in section V.
Figure 3: Yung et al’s results for thermal conductivity of BN
resins with various filler contents.
7
2.3 Percolation Theory and Previous Research on Fillers
Previous studies indicate that embedding highly-conductive ceramic filler particles
in the polymer matrix is the most widely used method to develop a composite with high
thermal conductivity. The process which governs the enhancement in thermal
conductivity is called Percolation, which is a phenomenon in which the highly
conducting filler particles form at least one continuous low-resistance chain connecting
the opposite faces of the matrix. Percolation theory predicts that the thermal conductivity
of the composites increases according to a scaling law with increasing concentration of
the high thermal conductivity constituent after percolation. However, “when the
characteristic size of the particles in the nanocomposites is comparable to or smaller than
the phonon mean free path, the phonon scattering at interfaces between two materials
can introduce significant thermal resistance in the highly conductive phonon pathway.
Such interfacial thermal resistance can reduce the thermal conductivity of the
nanoparticle composites.” The thermal conductivity of the random composites thus
deviates significantly from the predictions of the percolation theory. [21][22]
In early studies on the topic, oxide fillers such as alumina (Al2O3) and silica (SiO2)
were used due to their affordability and processability in mixing procedures. However,
they have always yielded poor results inasmuch as the thermal conductivities of the oxide
filler composites were at most 1–3 W/mK. [23] Subsequent research focused on nitride
ceramics, particularly Aluminum Nitride (AlN) and Boron Nitride (BN), which are now
considered the best qualified fillers for thermally-conductive polymers. [24][25] Both
exhibit very high thermal conductivities, as evident in table 1, are affordable, and they
disperse well in polymers.
Table 1: Thermal Properties for Various Materials [26][27]
Material Density
(g/cc)
CTE
[ppm/°K]
Specific Heat
[J/kg-°K] @ 25°C
Thermal Conductivity
[W/m°K]
Boron Nitride 2.25 1.20 794 300+
Aluminum Nitride 3.26 4.10 734 260
Aluminum Oxide 3.98 7.1 798 30
Silica 2.20 0.5 689 1.4
Zinc Oxide 5.64 0 [28] 523 54
8
III Phase One: Preliminary Research using Single-Phase
And Dual ‘Hybrid’ Fillers
Based on the findings in the last section, it was decided to go forward with PPS and
Boron Nitride as the baseline constituents of the composite. This was primarily due to their
favourable properties as postulated by previous research.
In line with that strategy, the aim of this research phase is to gain an understanding of
how different amounts and types (polymorphs) of Boron Nitride embedded in a PPS matrix
would affect the thermal conductivity of the composite by conducting a battery of tests using
as many different compositions as possible.
Two different parameters known to affect thermal conductivity are going to be explored in
this section
1. Type of Boron Nitride: Four different polymorphs of hexagonal Boron Nitride
comprising different shapes and sizes were decided are going to be used in order to
compare their performance1. SEM images, along with physical properties, of these
polymorphs can be seen in the figure 4 below.
1 Obtained from Momentive Performance Materials, Inc.
Figure 4: SEM micrographs of the polymorphs of Boron Nitride which will be tested in this thesis
9
2. Amount of Boron Nitride in a thermal conductivity sample: Based on previous
research, we know that the higher the Boron Nitride content in the composite, the
high the thermal conductivity. However, there is both a theoretical and practical limit
as to how much Boron Nitride one can embed in a composite.
The theoretical limit can be easily determined by calculating the maximum theoretical
atomic packing factor (APF) of Boron Nitride spheres in the PPS matrix:
�67 =898:; <=ℎ>?> @9;A5>
898:; 7BB ACD8 E>;; @9;A5> =
FG
FH=
(163 )I�J
16�J√2= 0.74
This then tells us that the composition at which the Boron Nitride particles are closest
packed in the PPS matrix is 74 wt%. At this composition, however, the sample is
speculated to be too brittle to be viable. Therefore, an additional composition – 33
wt% Boron Nitride – will be tested in order to establish a meaningful relationship
between Boron Nitride content and Thermal Conductivity.
In addition to the above tests, we will explore the effect of using hybrid-filler composites,
containing 2 or more polymorphs of Boron Nitride, on thermal conductivity. The purpose
behind these tests is to attempt to maximize the thermal conductivity by combining the
positive effects of various shapes and sizes of Boron Nitride in the same matrix.
We will then conclude the section by analyzing a number of SEM micrographs of a
representative number of compositions in order to examine the morphologies of the
prototyped samples.
10
3.1 Methodology
For the thermal conductivity tests, a 2 cm diameter disc-shaped prototype is used. The
steps involved in making the fabricating this sample are as follows:
1. The constituents for the desired composition are weighed using a digital scale then
mixed thoroughly in a glass container.
2. The mixture is then placed in a compounder which melts the constituents and
then, using a pair of screws, extrudes the mixture into a long strand [32]. The
purpose of this process is to ensure that the filler particulates are properly
distributed and well dispersed within the polymer matrix
3. The strands are then inserted in a freeze mill which pelletizes them into a finely-
ground thoroughly dispersed powder. This is done by cooling the sample using
liquid nitrogen to make it sufficiently brittle and then hammering it into a powder;
the process is carried over three 2-minute cycles in order to give the hammer and
sample time to cool down before continuing to grind it down. The end result is a
thoroughly-mixed powder consisting of the polymer and filler(s)
4. The powder samples are then inserted in the mold and are compacted at room
temperature under 2000 psi to ensure that the powder solidifies in the mold before
moving on to the next sample. Otherwise, the power might be blown away as the
other samples are added.
11
5. Finally, when all samples are compacted, they are molded at 310°C and 2000 psi
for 15 minutes. The mold is then quenched in ice-cold water and the discs are
removed.
Two molds may be used which produce different sized samples: one is used for thermal
conductivity analysis, and is 1 cm in diameter and 2 cm thick. The other is used for SEM
inspection, and is the same diameter but only 1 cm thick in order to be easily shattered
during SEM preparation.
If the composition of the sample is over 45% BN or 20% CNT, the compounder is
unable to process the blend since it becomes too viscous, and it would therefore have to
be dry-blended directly using the compression molder. This entails weighing out and
mixing the samples in a glass container by shaking the container by hand – to ensure the
constituents are well dispersed – compacting them individually in the mold, and then
finally compressing them at 310°C, 2000 psi.
This method produces samples in which the filler is poorly dispersed in the polymer since
the compounding and freeze milling steps are forgone.
The final step is to then measure the thermal conductivity of the samples using a custom-
built analyzer which conforms to ASTM standard E-1255-04[33].
This is carried out by placing the test specimen under load between two steel bars having
the same diameter as the specimen. A heater at the top of the test stack is then turned on
and a temperature gradient is established across the stack; heat losses are minimized by
use of a thick insulation layer as well as a longitudinal guard having approximately the
same temperature gradient. At equilibr
measured temperature gradients in the respective specimens and the thermal
conductivity of the reference materials by using the following equation:
" =L
�
where,
Zi are positions measured from the top of Ti are temperatures at Zi kM is the thermal conductivity of the met
This calculation is carried out through software which monitors the temperatures across
the gradient using 9 thermocouples.
Figure 5: Schematic of the thermal conductivity analyze showing the comparative
Guard Heater
Section
Metal or
ceramic guard
shell at Tg
12
use of a thick insulation layer as well as a longitudinal guard having approximately the
same temperature gradient. At equilibrium, the thermal conductivity is derived from the
gradients in the respective specimens and the thermal
conductivity of the reference materials by using the following equation:
L, − LJ
�, − �J."M
2. (
�N − �O
LN − LO+
�P − �Q
LP − LQ)
are positions measured from the top of the column and given by figure 5
is the thermal conductivity of the metal reference bars
This calculation is carried out through software which monitors the temperatures across
the gradient using 9 thermocouples.
Schematic of the thermal conductivity analyze showing the comparative-guarded longitudinal heat
use of a thick insulation layer as well as a longitudinal guard having approximately the
, the thermal conductivity is derived from the
gradients in the respective specimens and the thermal
5
This calculation is carried out through software which monitors the temperatures across
guarded longitudinal heat flow system
13
3.2 Results
In line with the introductory discussion, 13 samples with different compositions were
prepared and tested for thermal conductivity. Six of those samples were single-phase
composites, meaning they contained only one polymorph of Boron Nitirde, while the
other four were hybrid (dual) composites, comprising two different polymorphs of Boron
Nitride, dispersed in a PPS matrix.
3.2.1 Single-Phase Composite Results
The purpose of the single-phase tests was two-fold:
1. To understand how different polymorphs of Boron Nitride affect the thermal
conductivity of BN-PPS composites: This is accomplished by testing 4 different
samples each containing the same concentration (33 wt%) of Boron Nitride but
with different polymorphs
2. To determine how much thermal conductivity increases with increasing Boron
Nitride concentration in the composite (from 33 wt% to 74 wt%)
0
1
2
3
4
5
6
7
8
9
Pure PPS PTX25 PTX60 PT371 PT110
Th
erm
al
Co
nd
uct
ivit
y (
W/m
.K)
Chart 1: Thermal Conductivity Results For Different Polymorphs of Boron Nitride
33 wt% Boron Nitride
74 wt% Boron Nitride
14
Looking at the chart 1 on the previous page, we observe that, for 33 wt% Boron Nitride
composites, three of the polymorphs, PTX25, PTX60 AND PT371, have an identical effect
on thermal conductivity, raising it only slightly above that of neat PPS. PT110, on the
other hand, performs relatively poorly, falling almost 30% short of the thermal
conductivities attained by the other polymorphs. This may be due to the fact that the
PT110 platelets are pulverized, or otherwise flattened across the width of the sample,
during compression molding, therefore diminshing the thermal network which forms
across the matrix and resulting in a breakdown of percolation. In fact, this is true of all
the other polymorphs, though to a lesser extent than in the case of PT110. Using a
modified version of the Maxwell-Garnett formula:
we can estimate that the theoretical thermal conductivity of a composite consisting of 33
wt% Boron Nitride within a PPS matrix to be 44 W/m.K, which is well above any of the
obtained results[46].
Generally speaking, this discrepancy between theoretical and experimental thermal
conductivities is due to a high thermal resistance at the interfacial boundaries between
PPS and the Boron Nitride particulates due to
the inherent incompatibility of the two
compounds. This issue can be resolved using
coupling agents, which will be addressed in
section III.
Another reason for the deviation of results from
theoretical predictions is due to the imperfect
dispersion of Boron Nitride in the PPS matrix.
Since the only method used to mix the two
Figure 6:
Schematic
showing (a)
theoretical, (b)
(likely) actual
packing of BN
particulate
fillers in PPS
matrix
15
components is manual shaking, there is usually an uneven distribution of matrix and filler
powders during compression molding. As a result, filler particulates tend to clump
together into clusters throughout the matrix with regions of pure PPS in between, rather
than percolating by arranging into a proper close-packed geometry as can be seen in
figure 6 on the previous page.
Referring back to chart 1 on page 13, we observe that increasing the Boron Nitride content
of the samples vastly improved their thermal conductivity. At 7.8 W/m.K, PTX60 achieves
the highest thermal conductivity attained so far, edging out PT371 by about 0.4 W/m.K.
The reason for this slight discrepenacy relates to differences in the microstructures of the
PTX60 and PT371 composites which will be explored in detail the next section using SEM
micrographs.
Since 74 wt% Boron Niride implies that most of the sample is
ceramic in nature, prototypes which were fabricated at this
composition very brittle and difficult to handle. In fact, all
samples containing 74 wt% PTX25 and PT110 fractured either
during molding or drilling, and so were rendered un-testable
(see figure 7a).
Though the value achieved for 74 wt% PTX60 is encouraging, it is not recommended to
develop a material with such a high content of ceramic since it would be very difficult to
machine into a multi-fin heatsink. In addition, it will be liable to fail due to fatigue early
on in its lifecycle as unstable cracks would propagate very quickly after initiating at sites
of stress concentration, such as at the corners of the fins.
Figure 7a: Cracked 74 wt% PTX25 sample
16
3.2.2 Hybrid Filler Composite Results
As was pointed out in the
previous section, one of the
two main reasons behind
the discrepancy between
empirical predictions of
thermal conductivity and
experimental results is due
to the coalescence of the
Boron Nitride particulates
in the matrix during
compression molding, leaving vast regions of poorly-conductive PPS across the
composite.
To address that issue, dual-phase PTX60:PTX25 hybrid fillers were attempted, predicated
on the understanding that PTX25 spheres are small enough to fit in the interstitial sites
between the larger PTX60 spheres, thereby achieving an ordered arrangement of filler
particulates in the matrix and forming the desired thermal conductivity network.
At 33 wt% Boron Nitride, three different samples with different ratios of PTX60 to PTX25
were prototyped and tested in order to determine the optimal proportion in the sample
which would ensure the highest thermal conductivity; these ratios are 2:1, 3:1 and 4:1
PTX60:PTX25. Looking at the results in chart 2 above, we notice that there is a weak,
though positive correlation between the ratio of PTX60:PTX25 and thermal conductivity.
Based on these results, we can hypothesize that PTX60 has a better thermal conductivity
in Boron Nitride-PPS composites than PTX25 since, in samples containing identical
amounts of boron nitride but with different ratios of PTX60 and PTX25, increasing the
proportion of PTX60 in the sample resulted in an increase in thermal conductivity of the
composite.
0
1
2
3
4
5
6
2:1
PTX60 : PTX25
3:1
PTX60 : PTX25
4:1
PTX60 : PTX25
15:1
PTX60 : PT110
Th
erm
al C
on
du
ctiv
ity
(W
/m.K
)
Chart 2: Thermal Conductivity Results For Hybrid-Filler BN Composites
17
We also conclude that the assumption made regarding better percolation was not fulfilled
since there is almost no improvement in thermal conductivity using hybrid fillers than
one-phase fillers, both of which achieve ~2 W/m.K at 33 wt% Boron Nitride, which is well
below our target of 10 W/m.K. The underlying reasons for this, and as to why PTX60
performs better than PTX25 will be explained using SEM micrographs in the next section
In addition to the PTX60:PTX25 hybrid-filler composites, one sample containing 74%
PTX60 and 5% PT110 – a 15:1 ratio of PTX60:PT110 – was tested. This was based upon the
assumption that the crushing of PT110 platelets during compression molding may serve to
disperse them throughout the matrix, thereby connecting together the PTX60
particulates in the matrix; this then forms an interconnected network of highly-
conductive Boron Nitride particulates throughout the matrix, leading to a high thermal
conductivity.
The result, however, indicates that this is not the case; at 5.6 W/m.K, the thermal
conductivity of the hybrid-filler sample is 40% less than the 74 wt% PTX60 sample, which
suggests that PT110 particulates have a detrimental effect on thermal conductivity. Likely
this is due to the high incompatibility between the PT110 platelets and PPS, leading to a
high thermal resistance at the interfacial boundaries. As was mentioned earlier, this
problem may be addressed by treating the Boron Nitride with a compatibilizing agent to
improve its bonding to the matrix. However, in the case of PT110, even if surface
treatment is carried out, the platlets will still tend to line up across the width of the
specimen, therefore providing asymmetric thermal conductivity in the composite.
A 15:1 PTX60:PTX25 sample was also prototyped, but due to excessive brittleness, it
shattered during extraction from the mold.
18
3.3 SEM Analysis
The Scanning Electron Microscope (SEM) is a powerful tool which can provide us with a
valuable insight into the microstructure of the prototypes , giving information about
dispersion, morphology, percolation and integrity of the sample. Faults such as cracks or
air pockets, which elevate thermal resistivity within the composite, can also be identified
on the micrographs.
3.3.1 Methodology
Two methods were used to obtain SEM samples for
analysis: The first was to compression-mold a thin disc
specially made for SEM analysis (see figure 7b); the
second was to cut slots into thermal conductivity
samples using a band saw after they were tested using
the thermal conductivity analyzer. After that, the following steps are carried out [34]:
1. The sample is dipped in liquid nitrogen for about one minute then, using a pair to
tongs, snapped in order to expose a clean fracture surface with good topology for
analysis.
2. Next, the sample is mount sample onto stub using glue
3. The surface is then polished with SiC sandpaper in order to enhance grain
boundaries
4. The metallic stub is covered with Carbon Black in order to insulate it so that it
doesn’t interfere with the electron beam
5. The sample is then sputter-coated with a 7.5 – 30 nm gold platinum layer at ~ 10 Pa
in order to impart electrical conductivity to the fractured surface. This results in a
higher rate of secondary and backscattered electron emissions which would yield a
higher resolution image. If the sample is not properly coated, the samples
accumulate charge when scanned by the electron beam, which can cause beam
Figure 7b: 74 wt% PT371 SEM sample
19
deflection and increased emissions of secondary electrons which cause scanning
faults and other image artifacts.
6. Finally, the samples are placed under the microscope, analyzed, and photographed
3.3.2 SEM Micrograph Results
As with before, analysis was carried out on both single-phase and hybrid-filler
composites. All samples were imaged at 50x, 500x and 1000x magnifications in order to
provide a cascade of morphology, from a macro overview to a high-resolution insight into
the grain structures of the composite constituents.
3.3.2.1 Single-Phase Composite Micrographs
SEM analysis was carried out on composites consisting of 74 wt% PTX25, PTX60 and
PT371, as well as 33 wt% PTX60 and PT371 in a PPS matrix.
The purpose behind analyzing these compositions is to show us how different
polymorphs of Boron Nitride disperse within the PPS matrix, how well they retain their
shape after compression molding, and to confirm the hypotheses postulated in section
3.2.2 regarding the thermal conductivity results obtained for these compositions.
SEM micrographs of the aforementioned compositions can be seen in the following two
pages.
20
Figure 8: 74% PTX25 in PPS matrix at 50X, 500X and 1000X respectively
Figure 9: 74% PTX60 in PPS matrix at 50X, 500X and 1000X respectively
Figure 10: 74% PT371 in PPS matrix at 50X, 500X and 1000X respectively
21
Figures 8, 9 and 11 show the Boron Nitride PTX25 and PTX60 spheres have largely
disintegrated during compression molding, resulting in a breakdown of their thermo-
mechanical properties.
This breadown occurs when Boron Nitride sheets slip past one another as they are
compressed under a high pressure, thereby deforming the hexagonal structure of the
Boron Nitride crystal. This results in steep changes in its crystallographic directions,
resulting in phonon scattering both within the Boron Nitride crystals and at the
interfacial barrier between Boron Nitride and PPS [31]. This scattering is the root cause of
the poor thermal conductivities obtained in the previous section. Therefore, the next
section (3.4) will attempt to address the issue of Boron Nitride pulverization during
compression molding by adjusting the process conditions
Figures 10 and 12 shows a similar pattern with the PT371 agglomerates broken down into
small flakes randomly dispersed throughout the matrix; these flakes are not only poorly
conductive relative to their intact counterparts, but they are also spaced apart in the
matrix, and therefore not forming a proper 3D thermal conductivity network.
3.3.2.2 Hybrid-Filler Composite Micrographs
As with the single phase samples, SEM analysis was conducted for two compositions, 15:1
PTX60: PT110 (74% PTX60, 4% PT110), and 15:1 PTX60:PTX25, which can be seen in figures
13 and 14 on the next page.
Figure 11: 33% PTX60 in PPS matrix at 500X Figure 12: 33% PT371 in PPS matrix at 500X
22
Figure 13: 74% PTX60 + 5% PT110 in PPS matrix
micrographs at 50X, 500X and 1000X respectively
Figure 14: 74% PTX60 + 5% PTX25 in PPS matrix
micrographs at 50X, 500X and 1000X respectively
23
In both samples, it is clear that, despite good dispersion, there is severe breakdown of the
Boron Nitride particulates in the matrix. This is due to the fact the matrix is inundated
with copious amounts of Boron Nitride (79 wt% in total), which results in their
immediate pulverization as they press up against each other during compression
molding. This is particularly apparent in figure 14, where it can be seen that both
polymorphs of Boron Nitride have completely broken down. On the other hand the
sample in figure 13 shows that some particulates retained their integrity during
compression molding. Nevertheless, the disintegration of was so extensive, that we
cannot consider either composition viable.
24
3.4 Compression Molding Pressure Analysis
As was mentioned in the last section, Boron Nitride particulates undergo extensive
damage due to the high pressure (2,000 psi) sustained during compression molding. This
prompts us to attempt to understand how we can address this issue; particularly, how
varying this pressure may produce better percolation.
Pressure plays a huge role in determining the final morphology of compression-molded
samples. In the case of ceramic-polymer composites, there are two dimensions to the
effect of pressure:
1. On the polymer matrix: The effect of molding pressure on the thermo-
mechanical properties of mouldable polymers has been extensively studied and is
widely understood. One paper by Parasnis and Ramini summarized this effect very
well; by conducting compression molding runs on samples compressed at different
pressures, they found that the stiffness and crystallinity of the molded polymer
increased as the pressure is increased. Past a certain pressure, these characteristics
decline rapidly, thereby diminishing the thermo-mechanical integrity of the
molded samples. That is to say, they found that there is a narrow pressure band in
which the crystallinity and stiffness of the polymer are maximized. For thermal
conductivity of ceramic-polymer composites, achieving crystallinity is not crucial,
but it may improve the obtained results due to the fact that neatly-crystallized
polymer speherulites tend to conduct heat better than amorphous polymers with
randomly-distributed strands.[43]
2. On the ceramic filler: As noted previously, damage sustained by the Boron
Nitride particles can result in a breakdown of their thermomechanical properties,
rendering them far less conductive than they would be if they were intact.
In order to understand how different moulding pressures affect the above mentioned
parameters, prototypes consisting of 33 wt% PTX60 in PPS were compression molded at 4
different pressures – 500, 1000, 1500 and 2000 psi – and analyzed using SEM.
25
Figure 15: 33% PTX60 in PPS matrix compression-molded at 500 psi – 50X
Figure 17: 33% PTX60 in PPS matrix compression-molded at 1,500 psi – 50X Figure 18: 33% PTX60 in PPS matrix compression-molded at 2,000 psi – 50X
Figure 16: 33% PTX60 in PPS matrix compression-molded at 1,000 psi – 50X
26
Figure 19: 33% PTX60 in PPS matrix compression-molded at 500 psi – 500X Figure 20: 33% PTX60 in PPS matrix compression-molded at 1,000 psi – 500X
Figure 21: 33% PTX60 in PPS matrix compression-molded at 1,500 psi – 500X Figure 22: 33% PTX60 in PPS matrix compression-molded at 2,00 psi – 500X
27
Two sets of images were taken of the samples compressed at the 4 pressures, one at 5ox
magnification to give a macroscopic view of the morphology of the samples (figures 15
through 18), and the other at 500x magnification to provide information on Boron Nitride
integrity after compression molding (figures 19 through 22).
The following observations are made:
• At low pressures, large microvoids, cracks and air pockets form throughout the
cross-section of the sample. These are and are a result of insufficient pressure
during compression molding, and are extremely detrimental to the thermal
conductivity of the final composite since the air in these gaps has a high thermal
resistivity. Therefore, we conclude that higher pressure yields a higher thermal
conductivity based on the 50x magnification micrographs.
• On the other hand, looking at figures 19 through 22, we notice that, as pressure
increases, there is a more pronounced breakdown of boron nitride particulates,
leading to loss of percolation and therefore diminishing the thermal conductivity
of the composite. Based on that observation, we conclude that the lower the
pressure, the better the percolation, and therefore the better the thermal
conductivity.
To strike a balance between these contradictory observations, we would have to pick one
of the two middle pressures – 1000 psi or 1500 psi. From a qualitative standpoint, it would
seem that the 1000 psi treatment is better since it produces fewer microvoids, while at the
same time preserving the integrity of the Boron Nitride sphere better than at 1,500 psi.
Therefore, it was decided from this point onward to apply a 1,000 psi compression
molding pressure to all thermal conductivity prototypes.
28
IV Phase Two: Coupling Agents
For a given filler concentration, there are three methods to increase the thermal
conductivity of the composite. They are (i) maximizing conductive paths, (ii) minimizing
thermal resistance in each conductive path [35], and (iii) decreasing the thermal contact
resistance at the filler–matrix interface. In this section, we will focus on the third method,
attempting so with the use of a coupling agent (also known as compatibilizer).
The value of the third method stems from the tendency for gaps or other flaws to occur at
the filler–matrix interface due to the insufficient affinity between filler and matrix. Such
interfacial flaws cause a high thermal resistance at the interface, thus reducing the
thermal conductivity of the composite [36, 37]. This thesis will attempt to use surface
treatment of the filler to improve the affinity between filler and matrix, thereby
significantly increasing the thermal conductivity of the composite. In particular, this
thesis will focus on using silane coupling agents for surface treatment. Silane acts as a
bridge to connect the ceramic filler and the polymer matrix together, because it has two
different chemical structures at the two ends of the molecule. One end is chemically
reactive with the polymer; the other end is chemically reactive with the surface of the
ceramic filler [37].
4.1 Background Research
Several studies have been conducted which used a variety of silane coupling agents. For
example, Yung and Yue [37] describe using a coupling agent – KBM-430 – to treat a
polymorph of hexagonal Boron Nitride not different from PTX60.
In surface-treating the Boron Nitride with the coupling agent, they first mixed the filler
and the coupling agent of 1 wt% of filler weight in a small amount of isopropyl alcohol at
80°C for 2 h. Next, they carried out vacuum drying to remove the solvent at 85°C for 24 h,
exposing the filler to ambient air for 4 h. Finally, they heat treated the filler at 100/120°C
for 4 h. The treated fillers were then stored in vacuum dryer.
29
Next, they embedded the
treated Boron Nitride in
an epoxy matrix and
conducted thermal
conductivity
measurements.
In their results, they
noted that increasing the
concentration of the
coupling agent increased the thermal conductivity up until 1%, which they found to be
the optimal concentration of coupling agent. Further increasing the coupling agent, they
found, led to a thick coating on the BN filler, which eventually become thermal barrier
causing the thermal conductivity to decrease. This is evident in figure 23 above, which
summarizes the results of their experiments.
The value of this study is that it provides a starting point for this section of the thesis. An
initial investigation found that KBM-430 may be difficult to procure, so a similar silane
agent, 3-aminopropyltriethoxysilane (SCA100), was obtained from Struktol.
Other coupling agents used by other researchers were also explored, such as Styrylethyl
trimethoxysilane (KH-560) [38] and [Methoxy(polyethyleneoxy) propyl]trimethoxysilane
[35] but they were both deemed unsuitable for this thesis as they could not sustain the
operating conditions under which the prototypes are produced. Therefore, the focus of
this section of the thesis will be solely on using SCA1100 as a surface treatment coupling
agent.
Finally, in addition to PPS, LDPE will be explored as a matrix polymer in future
experiments when using treated Boron Nitride. This is in accordance with research
Figure 23: Graph
showing the strong
correlation between
use of coupling
agents and thermal
conductivity. Note
that an increase in
the amount of
coupling agent does
not necessarily
imply an increase in
matrix-filler
adhesion. In fact, it
may have a
detrimental effect as
is the case here [37]
30
conducted by Harrison et al, which found a high affinity between silane-treated BN and
LDPE [10].
4.2 Methodology
The following steps are carried out in preparing a sample of Boron Nitride surface treated
with SCA1100 coupling agent:
1. A 50:50 mixture of water an ethanol is prepared in a beaker
2. The desired amount of coupling agent is added to the water-ethanol mixture using
a micropipette
3. The desired amount of Boron Nitride powder is added to the mixture of water,
ethanol and coupling agent
4. The mixture is stirred at the required temperature on a magnetic stir plate
5. The stirred mixture is filtered, leaving behind treated Boron Nitride powder (i.e.
Boron Nitride which has bonded to the coupling agent)
6. The Boron Nitride is dried on the hot plate overnight at 60°C
Treated Boron Nitride
50:50 water/ethanol solution Coupling Agent Boron Nitride
Stir on hot plate Filter out treated boron nitride
Treated Boron
Nitride
31
4.3 Thermal Gravimetric Analysis
Thermogravimetric Analysis (TGA) is a type of test which measures weight changes in a
material as a function of temperature (or time) under a controlled atmosphere. Its
principal uses include measurement of a material's thermal stability and composition.
For this thesis, TGA was carried out on treated and untreated samples of PTX60 and
PT110 in order to compare the effect of surface treatment on both polymorphs, as well as
to ensure that the coupling agent has adhered sufficiently to the Boron nitride particles.
4.3.1 Methodology
A TA Instruments Q-50 TG analyzer was used to conduct the TGA tests. The analyzer
itself consists of a high-precision balance with a platinum pan loaded with 10-20mg of a
powder of the composition to be tested. The sample is placed in a small electrically
heated oven with a thermocouple to accurately measure the temperature. The
atmosphere is then purged with Nitrogen in order prevent oxidation or other undesired
reactions.
Though the TGA machine can reach temperatures up to 1000°C, a temperature of 800°C
was deemed sufficient for the testing. A ‘ramp’ profile was picked, in which the furnace
heats the sample gradually at 20°C/minute until it reaches the 800°C mark.
After the data are logged, curve smoothing and other operations are then carried out by
the software in order to identify the points of inflection and produce an accurate graph of
percentage weight loss vs. temperature (or time).
32
4.3.2 Results
4.3.2.1 Calibration
As a proof-of-concept, LDPE was
analyzed for weight loss by
ramping the TGA up to 800°C.
Since LDPE’s degradation
temperature is only 120°C, one
would expect it to completely burn
off during the test, which was
confirmed as can be seen in figure
24 on the right. This test was
carried out to ensure the TGA is
properly calibrated.
4.3.2.2 PTX60 vs. PT110
Next, treated and untreated samples of PTX60 were analyzed in order to compare weight
loss results before and after surface treatment using 8 parts of SCA1100 coupling agent per
100 parts of PTX60.
Figure 24: TGA results for Pure LDPE
Figure 25: TGA results for Untreated PTX60 Figure 26: TGA results for PTX60 treated with SCA100
33
The same was carried out for PT110 samples before and after treatment with SCA1100
The fractional-loss results for treated and untreated BN upon heating to 800°C are shown
in chart 3 below. The amounts of volatile/decomposable material on the as-received,
acetone and HNO3 treated BN particles were negligible (~0.1% weight loss), indicating
that treatments involving acetone and acids did not result in a coating on the particle. On
the other hand, the amounts of weight loss were much higher (averaging out around -
0.82%) with treated PTX60 powder, indicating that the silane treatment resulted in a
coating on the particle, which decomposed during heating. This is exactly what we are
looking for in a coupling agent.
Figure 27: TGA results for Untreated PT110 Figure 28: : TGA results for PT110 treated with SCA100
-1.00%
-0.80%
-0.60%
-0.40%
-0.20%
0.00%
0.20%
PTX60 PT110 Untreated
Pe
rce
nta
ge
We
igh
t C
ha
ng
e
Chart 3: TGA results
34
The results for PT110, however, were slightly more ambiguous, with two of the samples
recording a weight gain and one recording a negligible loss. There are several reasons in
existing literature which may explain this unexpected weight gain. The first is that one of
the constituents in the sample may have oxidized during the experiment, leading to a
chemical reaction, which forms a compound that increases the weight of the sample [44].
This, however, is unlikely since the furnace is purged with nitrogen gas for the duration of
the run, which expels oxygen, thereby preventing any chemical reactions. Further, if
oxidation was an issue with the TGA, we would’ve observed the same weight gain when
testing the treated PTX60 samples.
Another possible explanation for the weight gain
is due to a phenomenon called baseline drift, in
which the TGA machine becomes decalibrated
over time due to changes in buoyancy of the purge
gas as its density decreases with increasing
temperature [45].
Regardless of the validity of this hypothesis is true, it was made clear from the TGA charts
that there is no perceptible weight loss in any of the treated PT110 samples, indicating
that SCA1100 failed to coat the PT110 platelets during the treatment process. This is likely
due to an incompatibility between the functional groups on the coupling agent and the
PT110 platelets.
Figure 29: Baseline Drift in TGA
35
By using SCA1100 solutions of three different concentrations, different amounts of coating
resulted. The three amounts were: 3 parts of SCA1100 per 100 parts of Boron Nitride (3 pph
for short), 8 pph (the recommended dosage per Struktol, the supplier of SCA1100), and 15
pph. These treatments were carried out for both PTX60 and PT110, and the results are
plotted in chart 4 below.
As can be seen, the higher the silane concentration in the solution, the greater the weight
loss, indicating an improved coating of the Boron Nitride. This was true for both PTX60
and PT110, though much more pronounced with the former.
These results imply that PTX60 has good reactivity with SCA1100 as it bonds well to it,
smoothing out its surface, thereby mitigating the interfacial thermal resistance that it
forms with PPS when the two are compression molded into a composite. These results,
however, do not imply that a higher amount of SCA1100 will ensure a higher thermal
conductivity in the a composite containing treated PTX60.
The next section addresses the issue of how to determine what the optimal amount of
coupling agent should be, amongst other factors.
-0.80%
-0.60%
-0.40%
-0.20%
0.00%
0.20%
0.40%
3 8 15
We
igh
t C
ah
ng
e
Amount of SCA1100 (parts of SCA1100 per hundred parts of Boron Nitride)
Chart 4 : TGA Results for PTX60 vs. PT110 vs. Amount of Silane Used
PTX60
PT110
36
4.4 Design of Experiments
Design of Experiments (DOE) is a systematic approach to investigating, understanding
and optimizing system or process. It comprises a series of structured tests in which
planned changes are made to the input variables – in this case, changes in process
conditions or amount of coupling agent, for example – of a process. The effects of these
changes on a pre-defined output – in this case, thermal conductivity – are then assessed
using statistical software such as MATLAB.
DOE is widely used in multivariate data analysis since it allows a judgement on the
significance to the output of input variables acting alone, as well input variables acting in
combination with one another.
Traditional ‘one-factor-at-a-time’ (OFAT) testing always carries the risk that the
experimenter may find one input variable to have a significant effect on the output while
failing to discover that changing another variable may alter the effect of the first (i.e.
some kind of dependency or interaction). This is because the temptation is to stop the
test when this first significant effect has been found. In order to reveal an interaction or
dependency, 'one change at a time' testing relies on the experimenter carrying the tests in
the appropriate direction. However, DOE plans for all possible dependencies in the first
place, and then prescribes exactly what data are needed to assess them i.e. whether input
variables change the response on their own, when combined, or not at all. In terms of
resource the exact length and size of the experiment are set by the design (i.e. before
testing begins).
Three types of DOE exist which may be used in our experiments: 1) Full Factorial DOE,
which accounts for all possible combinations of input variables, 2) Fractional Factorial
DOE, which exploits the sparsity-of-effects principle to expose information about the
most important features of the problem studied, while using a fraction of the effort of a
full factorial design in terms of experimental runs and resources, and 3) the Taguchi
Method, which is similar to the Fractional Factorial method, but relies on analysis of
variance (ANVOA) to define the list of experiments needed to be run.
37
For this thesis, it was determined that a Taguchi orthogonal-array DOE would be the
most efficient; two such arrays were to be constructed, one for samples with an LDPE
matrix, and the other for samples with a PPS matrix.
The first step in detailing these arrays was to list all control factors known (or suspected)
to have an effect on thermal conductivity of the resultant composite2:
1. Coupling Agent Treatment (stirring) Condition: Three stirring conditions
would be studied: 10 minutes at room temperature, 20 minutes at 60°C, and 20
minutes at 60°C
2. Amount of Coupling Agent: Per the recommendation of the supplier of SCA1100,
Struktol, three dosages of coupling agent would be attempted: 3 parts of SCA1100
per 100 parts of Boron Nitride (3 pph), 8 pph, and 15 pph.
3. Weight % of treated PTX60 in a thermal conductivity Sample: Six potential
levels would be considered: 5%, 10%, 25%, 33.3%, 50% and 74%. This allows us to
gauge the effect of coupling agents across a wide spectrum of different
compositions.
In summary, 3 control factors were identified: the first and second have three design
levels, and the third has six design levels. If we were to run a full factorial DOE against all
these factors, we would have to test 54 prototypes, which would be too expensive and too
time-consuming. Therefore, using MINITAB, we determine we arrive at the following
testing arrays: (1) An L9 DOE for samples consisting of a PPS matrix, and (b) An L12 DOE
for samples with an LDPE matrix. These arrays allow us to minimize the number of
experimental runs while still gaining information of how different control factors affect
the thermal conductivity independently and collectively.
A graphical summary of all the chosen experiments can be seen in figures 30 and 31 on the
next two pages.
2 The type of coupling agent was also identified as a control factor during the literature review phase, however, since no coupling agents were found that can withstand the process conditions currently used in prototyping samples, only SCA1100 is considered.
38
8 pph SCA 1100 +
PTX60
10 wt%
25 wt%
33 wt%
50 wt%
74 wt%
5 wt%
SCA 1100 + PTX60 at 60° for
10 minutes
3 pph SCA 1100 +
PTX60
15 pph SCA 1100
+ PTX60
33 wt%
33 wt%
Coupling agent
treatment condition
Amount of
Coupling Agent
wt% of agent-treated
PTX60 in a thermal
conductivity sample
SCA 1100 + PTX60 at 60° for 20
minutes
SCA 1100 mixed with PTX60 at
room temperature for 10 minutes
8 pph SCA 1100 +
PTX60
8 pph SCA 1100 +
PTX60
50%
33 wt%
33 wt%
50%
Type of Coupling Agent
3-aminopropyltriethoxysilane
(SCA1100)
Styrylethyl trimethoxysilane
[Methoxy(polyethyleneoxy)
propyl] trimethoxysilane
LDPE matrix
Figure 30: Summary of all compositions to be tested, using an LDPE matrix, as per a Taguchi L-27 DOE
39
Note the red cells in figure 30 on the previous page. These represent experimental
runs which have failed since they were carried out using LDPE in pellet form. The
samples produced using this method were extremely brittle as the LDPE pellets did
not melt during the compounding, but rather got embedded within the PPS.
Nevertheless, the information lost from the DOE due to the failure of these samples is
not severe This is because there is an additional array (fig 31) which carries out
similar tests, the results of which may, with caution, be extended to make conjectures
about samples with an LDPE matrix. This is explored in the next section.
Figure 31: Summary of all compositions to be tested, using a PPS matrix, as per a Taguchi L-27 DOE
40
1.7
1.9
2.1
2.3
2.5
2.7
2.9
Th
erm
al
Co
nd
uct
ivit
y (
W/m
.K)
4.5 Thermal Conductivity Results
Chart 5 above summarizes the
results of all experiments runs
at the same treatment condition of
10 minute stirring at 60°C for prototypes
with a PPS matrix. Immediately, we notice
that there is an almost linear relationship
between Boron Nitride content and thermal
conductivity, with the highest being 8.8 W/m.K
at 74 wt% Boron Nitride – very close to the desired
goal of 10 W/m.K. This figure is 13% higher that the
thermal conductivity obtained for untreated 74 wt%
Boron Nitride, indicating the coupling agent has
achieved its purpose of improving of the interface
between matrix and particles, thereby minimizing filler–matrix thermal contact
resistance. This improvement was also observed at the 33 wt% Boron Nitride
0
1
2
3
4
5
6
7
8
9
10
0% 10% 20% 30% 40% 50% 60% 70% 80%
Th
erm
al
Co
nd
uct
ivit
y (
W/m
.K)
wt% of PTX60 Boron Nitride
Chart 5: Thermal Conductivity Results: Treated vs. Untreated PTX60 in PPS matrix
Untreated PTX60
Treated with 8 pph SCA1100
Treated with 3 pph SCA1100
Treated with 15 pph SCA1100
41
composition, where the treated sample performed almost 40% better than its untreated
counterpart.
Next, we look at the effect of amount of coupling agent on thermal conductivity; this
comparison was carried out on 33 wt% Boron Nitride prototypes, and the results of these
tests are shown in the inset on chart 5.
Given that all three concentrations of SCA1100 yielded conductivities above that achieved
with untreated PTX60, we conclude that all silane treatments were effective in increasing
the thermal conductivity of the composite, though to varying degrees. As the amount of
silane coating on the particles increased, the thermal conductivity was enhanced more,
until the silane concentration reached 8 pph. Further increase of the silane concentration
to 15 pph caused the thermal conductivity to decrease. Hence, silane 8 pph treatment was
the most effective. This means that the coating resulting from the silane treatment must
be sufficiently thick in order for the treatment to be fully effective. The coating serves as
an interlayer at the filler–matrix interface, thereby improving the quality of the interface.
However, if the coating is too thick, the interlayer will become less effective or even a
thermal barrier, thereby decreasing the thermal conductivity.
42
Turning our attention to LDPE samples, we will now be able to understand the effect
of treatment conditions on thermal conductivity, as well as compare the effectiveness
of LDPE and PPS as matrices for our composites.
Looking at chart 6 above, we observe that the conditions under which PTX60 is
treated have a minimal effect on thermal conductivity regardless of composition.
That is to say, stirring PTX60 and SCA1100 at 60°C for 20 minutes does not achieve
any better a coating than does stirring them for 10 minutes at room temperature.
Therefore, to save time and energy, it is recommended for future research to carry
out the stirring at the latter condition.
We can also observe from chart 6 that PPS is a marginally better matrix than LDPE as
it manages to achieve, on average, an 8% higher thermal conductivity. This is likely
due to the fact that the functional groups on SCA1100 bond better to PPS than they do
with LDPE. Further, LDPE has less tendency to crystallize than PPS, therefore
diminishing the thermal conductivity of the matrix in the final composite.
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
60° for 10
mins
60° for 20
mins
22° for 10
mins
60° for 10
mins
60° for 10
mins
60° for 20
mins
22° for 10
mins
60° for 10
mins
LDPE PPS LDPE PPS
33% PTX60 50% PTX60
Th
erm
al
Co
nd
uct
ivit
y (
W/m
.K)
Chart 6: Thermal Conductivity Results: Different Treatment Conditions
43
Referring back to Design of Experiments, notice how, using a Taguchi DOE, we were
able to reduce the number of experiments by concluding from the LDPE results in
chart 6 that the treatment conditions are inconsequential, therefore forgoing the
need to test PPS samples at each process condition.
4.6 SEM Analysis
In order to study the effect of coupling agents on the morphology of the samples, four
prototypes were sectioned for SEM analysis, of which two comprised on a PPS matrix
embedded with:
1. 33 wt% PTX treated with 8 pph SCA1100
2. 33 wt% PTX treated with 8 pph SCA1100
And the other two comprised an LDPE matrix embedded with
3. 33 wt% PTX treated with 8 pph SCA1100
4. 53 wt% PTX treated with 8 pph SCA1100
44
Figure 32 shows SEM micrographs of the cross-section of Boron Nitride-PPS composites
containing as-received BN particles (fig 32a) and silane-treated BN particles (fig 32b).
Figure 32a shows cracks, bare particles, and holes caused by particle pull-out due to the
poor interface between particles and matrix. Figure 32b shows fewer cracks, less bare
particles and less holes, indicating the improvement of the interface. This is in line with
the thermal conductivity results in section 4.5 which show that the treated Boron Nitride
samples performed substantially better than their untreated counterparts.
Figures 33 and 34 above show PPS composite samples embedded with treated 33% (fig
33) and 50% (fig 34) PTX60 Boron Nitride. In both cases, we observe that the Boron
Nitride particles have been coated with a layer of Silane. This improves the matrix-
Figure 32: SEM micrographs of a PPS matrix embedded with 33 wt% PTX60 (a) as-received, (b) treated
with SCA1100 at 60°C for 10 minutes. 500X magnification.
(a) (b)
Figure 33: 33 wt% surface-treated PTX60 in PPS matrix Figure 34: 50 wt% surface-treated PTX60 in PPS matrix
45
filler bonding and reduces the steep interfacial directional changes which were
previously resulting in severe phonon scattering leading to poor thermal
conductivity.
Figures 35 show LDPE composite samples embedded with treated 33% (fig 35) and
50% (fig 36) PTX60 Boron Nitride. In those cases, we observe the same coating layer
on the Boron Nitride particulates, but far less pronounced than in the PPS images,
indicating that the preparation of the treated Boron Nitride was not as effective as it
was in the previous cases. As such, it is recommended that these samples be retested
using a new batch of PTX60 treated with 8 pph SCA1100.
Finally, we note from figure 37, which shows
a low magnification image of an LDPE-PTX60
composite that the composites consisting of
an LDPE matrix produce an excellent finish
with very few microvoids, gaps or
imperfections as is the case with PPS matrix
samples (see figures 15 through 18).
Figure 35: 33 wt% surface-treated PTX60 in LDPE matrix Figure 36: 50 wt% surface-treated PTX60 in LDPE matrix
Figure 37: 33 wt% surface-treated PTX60 in LDPE
matrix. Low magnification
46
V Future Research Directions
It would seem that this research project has reached its limits, with the highest thermal
conductivity – still well short of predicted values – being obtained at the maximum 74
wt% Boron Nitride (with coupling agent). However, expanding on the results obtained in
this thesis, much work can be done in order to further close the gap between
experimental values and theoretical predictions. This section outlines three such
methods.
5.1 Multi-Modal hBN fillers
So far, the focus has been on using hexagonal Boron Nitride (hBN) embedded in
PPS in order to produce a composite with appreciable thermal conductivity. Another
form of Boron Nitride exists, however, which may be used in tandem with hBN to
improve upon the current results: Cubic boron nitride (cBN), a diamond-like allotrope of
Boron Nitride, exhibits excellent mechanical and thermal properties due to its
microstructure. It is synthesized by heating powdered hBN to 1500°C at a pressure in
excess of 60 kbars. This process transforms the atomic structure from hexagonal to cubic,
making the material hard and of high abrasivity.
hBN and cBN are often referred to as ‘modes’ of Boron Nitride. When compounded with a
resin and compression molded, they are equally good in terms of thermal conductivity
[25]. However, due to its hardness and abrasivity, cBN must be used with caution, and in
small amounts, as it may damage the compression molding.
In order to ensure the formation of a near-perfect conductive network throughout
the composite, maximum packing of the Boron Nitride in the matrix must be carried out.
To achieve a high packing density composites, several researchers have found the use of
large size particles with multi- modal particle size distribution and low aspect ratio with
smooth surface texture to be very most effective [9][25][31]. The theory behind the
superiority of multi-modality relies on the idea of combined use of whiskers (hBN) and
47
particles(cBN) enhances the formation of a thermally conductive networks throughout
the composite[29]. The large particle size, on the other hand, is desired to minimize the
scattering of phonons due to the interfacial thermal barrier. Moreover, the use of large
particle size tends to form fewer thermally resistant junctions of the polymer layers than
the small particle size at the same filler content. Finally, the smooth surface texture
ensures that there is good adhesion between the BN and resin.
5.2 Ishida Method
Perhaps one of the best and most encouraging papers written on the topic of thermally-
conductive polymers is by Ishida and Rimdusit, who managed to obtain a thermal
conductivity of 32.5 W/m.K by embedding 88 wt%
hexagonal BN (grade HCJ48)3 in as-synthesized
bisphenol-A-methylamine type polybenzoxazine.
They note that “the remarkably high value was
obtained using the well-recognized concept of
thermal management in composite materials by
maximizing the formation of conductive networks
while minimizing the thermal barrier resistance
along the heat flow path. The concept was
accomplished by using
highly thermally conductive filler with a matrix
resin which has low melt viscosity and good adhesion to the filler. In addition, a large
particle size with multimodal particle size distribution was used. Boron nitride and
polybenzoxazine have properties that meet all these requirements and thus exhibit a very
high thermal conductivity value [...] Boron nitride-filled polybenzoxazine has many
outstanding properties which makes it suitable for an application as a molding compound
3 supplied by Advanced Ceramics, http://www.advceramics.com/
Figure 38: Ishida and Rimdusit's results for BN-saturated
polybenzoxazine
48
for the electronic packaging industry and other applications with high thermal
conductivity.” [9]
5.3 Multifunctional Fillers
As was pointed out in section 3.3, the
likely reason why there is a large gap
between the obtained and the
expected thermal conductivity values
is that, instead of agglomerating
together in a close-packed
arrangement to form a thermal
conductive network, the BN filler
particulates tend to cluster together in
a random arrangement, thus leaving
gaps in the conductive path.
These gaps may be bridged using carbon nanotubes (CNT), carbon fibres or graphene
[40] as diagrammed in figure 39. One study, conducted by Agarwal et al using various
compositions and orientations of carbon fiber and CNT fillers in a polycarbonate matrix
concluded that “composites containing hybrid mixtures of two fillers show higher
thermal conductivity due to increased contacts between the microfibers that are
promoted by the presence of the nanofibers, especially when microfibers are aligned.” [41]
Other potential fillers include nano-scaled graphene platelets (NGPs), which can be
obtained from American Elements [42] and carbon fibres, both of which exhibit excellent
thermomechanical properties.
Figure39:
Diagram
showing
thermal
conductive
pathways in
the matrix
49
VI Conclusion
A thermal conductivity of 8.8 W/m.K was obtained using composite consisting of 74 wt%
Boron Nitride embedded in a poly(p-phenylene sulphide) matrix. This remarkably high
value was obtained using the well-recognized concept of thermal management in
composite materials by maximizing the formation of conductive networks while
minimizing the thermal barrier resistance along the heat flow path. This was
accomplished by (1) using a resin which has a low melt viscosity and excellent affinity
with the filler to ensure that air gaps within the composite are minimized, (2) surface
treating the filler with a coupling agent, which improved the adhesion of Boron Nitride to
the matrix, thereby reducing phonon scattering at interfacial barriers and lattice defects
and bringing the percolation threshold within reach, (2) utilizing an optimized
compression molding pressure to preserve the thermomechanical integrity of Boron
Nitride, and (3) exploiting the spherical shape of PTX60-grade Boron Nitride to maximize
the packing of filler throughout the matrix, thereby increasing the likelihood of filler-filler
contact, and reducing the percolation threshold.
Though LDPE has shown far better affinity to Boron Nitride than PPS in SEM images, BN-
LDPE composites seem to have a high percolation threshold, and therefore would require
the use of a different coupling agent, or Boron Nitride contents in excess of 50 wt% in
order to achieve the desire goal of 10 W/m.K. However, a better-established method
would be the use of epoxy resins such as polybenzoxazine, combined with multi-modal
Boron Nitride filler, which, together, have shown potential to reach up to 32 W/m.K.
50
Acknowledgements: The author thanks Dr. Naguib for his continuous supervisory support and Dr. Leung for his superb mentoring, as well as Mr. Khan for his guidance and suggestions.
References
[1] Li et al, “Design and performance evaluation of microprocessor packaging capacitors using integrated capacitor-via-plane model”, IEEE Trasnscations on Advanced Packaging, Vol23, Issue: 3 pp 361-367, Aug 2000
[2] Sato et al, “Thermally conductive composite films of hexagonal boron nitride and polyimide with affinity-enhanced interfaces”, Journal of Materials Chemistry, vol.20,no.14, pp.2749-2752,2010
[3] Droval, G.; Feller, J. F.; Salagnac, P.; Glouannec, P. Polym. AdV. Technol. 2006, 17, 732–745
[4] Lee, W. S.; Yu, J. Diamond Relat. Mater. 2005, 14, 1647–1653
[5] Zhu etl al, “Study on the Properties of the Epoxy-Matrix Composites Filled with Thermally Conductive AlN and BN Ceramic Particles”, Journal of Applied Polymer Science Volume 118, Issue 5, pp 2754–2764, 5 December 2010
[6] Xu, “Increasing the thermal conductivity of boron nitride and aluminum nitride particle epoxy-matrix composites by particle surface treatments”, Composite Materials Research Laboratory, State University of New York at Buffalo, Buffalo, NY 14260-4400, USA
[7] Lee et al, “Thermally conductive and electrically insulating "EVA composite encapsulants for solar photovoltaic (PV) cell”, eXPRESS Polymer Letters, Vol.2, No.5 (2008) 357–363
[8] K. Sato, H. Horibe, and H. Horibe, “Thermally conductive composite films of hexagonal boron nitride and polyimide with affinity-enhanced interfaces,” Journal of Materials Chemistry, vol. 20, no. 14, pp. 2749–2752, 2010.
[9] Ishida et al, “Very high thermal conductivity obtained by boron nitride-filled polybenzoxazine”, Thermochimica Acta, vol 320, Issues 1-2, 2 pp 177-186, November 1998
[10] Harrison et al, “Polyethylene/Boron Nitride Composites for Space Radiation Shielding”, Journal of
Applied Polymer Science, Volume 109, Issue 4, pages 2529–2538, 15 August 2008
[11] Li et al, “Enhanced Thermal Conductivity of Polyimide Films via a Hybrid of Micro- and Nano Sized Boron Nitride” J. Phys. Chem. B, 2010, 114 (20), pp 6825–6829
[12] Gabellini and Maroaes, “Miscibility and morphology of poly p-phenylene sulphide–liquid crystal polymer blends”, Journal of Applied Polymer Science, volume 60, Issue 1, pages 21–27, 4 April 1996
[13] Author Unknown, JOURNAL OF MATERIALS SCIENCE 12 (1977), http://www.springerlink.com/content/r15008m241448215/fulltext.pdf
[14] http://www.dimensionsguide.com/processor-dimensions/
[15] Reebay et al, “Experimental Study of the convective heat transfer coefficient in electronic cooling”, Quantitative Infrared Thermography, University of Laval
[16] Xu, “Increasing the thermal conductivity of boron nitride and aluminum nitride particle epoxy-matrix composites by particle surface treatments”, Composite Materials Research Laboratory, State University of
New York at Buffalo, Buffalo, NY 14260-4400, USA
[17] Chung et al, “Role of Phonon Dispersion in Lattice Thermal Conductivity Modeling”, Transactions of
the ASME, Vol. 126, JUNE 2004
[18] L. D'Ilario and A. Martinelli, “Glass transition and the origin of poly(p-phenylene sulfide) secondary crystallization”, The European Physical Journal: Soft Materials and Biological Processes, vol 19, Number 1, 37-45, DOI: 10.1140/epje/e2006-00009-4
[19] Xu et al, “Synthesis and characterization of poly(p-phenylene sulfide sulfone/ketone) copolymer” POLYMER BULLETIN, Volume 54, Numbers 4-5, 251-261, DOI: 10.1007/s00289-005-0392-3
51
[20] Shen S, Henry A, Tong J, Zheng R, Gang Chen G. “Polyethylene nanofibres with very high thermal conductivities”, Nature Nanotechnology. 7 March 2010.
[21] Tian and Yang, “Phonon Transport and Thermal Conductivity Percolation in Random Nanoparticle Composites” CMES, vol.24, no.2, pp.123-141, 2008
[22] Martin et al. “Field-structured composites for efficient, directed heat transfer” Journal of Applied Physics, Volume: 106 Issue:8 Oct 2009
[23] Sato et al, “Thermally conductive composite films of hexagonal boron nitride and polyimide with affinity-enhanced interfaces”, JOURNAL OF MATERIALS CHEMISTRY,vol.20,no.14,pp.2749-2752,2010
[24] Neelakanta, Perambur S. Ph.D., C. Eng., "Handbook of Electromagnetic Materials, Monolithic and Composite Versions and their Applications", pp. 25-26.
[25] Droval et al, “Electrothermal Behavior of Conductive Polymer Composite Heating Elements Filled with Ceramic Particles”, Journal of Thermophysics and Heat Transfer 0887-8722 vol.22 no.4, Oct 2008
[26] Ali, “Ceramic-Polymer Composite Material and its Use in Microelectronics Packaging", US Patents Office, August 1998
[27] Raman and Meneghetti, “Boron nitride finds new applications in thermoplastic compounds”, Plastics, Additives and Compounding, Volume 10, Issue 3, May-June 2008, Pages 26-29, 31
[28] L.I. Demkina “A New System for Calculating The Coefficient of Thermal Expansion of Silicate Glasses”, Glass and Ceramics, Volume 17, Number 10, 503-510, DOI: 10.1007/BF00669607
[29] Raman and Meneghetti, “Boron nitride finds new applications in thermoplastic compounds”, Plastics,
Additives and Compounding, Volume 10, Issue 3, May-June 2008, Pages 26-29, 31
[30] Sato et al, “Thermally conductive composite films of hexagonal boron nitride and polyimide with affinity-enhanced interfaces”, Journal of Materials Chemistry, vol.20,no.14, pp.2749-2752,2010
[31] Yung et al, “Enhanced Thermal Conductivity of Boron Nitride Epoxy-Matrix Composite Through Multi-Modal Particle Size Mixing” Journal of Applied Polymer Science, vol 106, Issue 6, pages 3587–3591, 15 December 2007
[32] Burbank, “Twin Screw Compounding”, Rubber World, Jul 1997 http://www.thefreelibrary.com/Twin+screw+compounding-a019677610
[33] “Standard Test Method for Thermal Conductivity of Solids by Means of the Guarded-Comparative-Longitudinal Heat Flow Technique”, Developed by Subcommittee: E37.05, Book of Standards Volume: 14.02
[34] Jeffree, C. E.; Read, N. D. (1991). "Ambient- and Low-temperature scanning electron microscopy". In Hall, J. L.; Hawes, C. R.. Electron Microscopy of Plant Cells. London: Academic Press. pp. 313–413. ISBN 0123188806.
[35] R. Ruth, K. Y. Donaldson and D. P. H. Hasseiman, J. Amer. Ceram. Soc. 75, 2887 (1992).
[36] Jung et al, “Preparations and thermal properties of micro- and nano-BN dispersed HDPE composites”, Thermochimica Acta, Volume 499, Issues 1-2, 20 February 2010, Pages 8-14
[37] KC Yung and Yue, “Thermal Management for Boron Nitride Filled Metal Core Printed Circuit Board” Journal of Composite Materials, December 2008 vol. 42 no. 24 2615-2627
[38] Lee et al, “Thermally conductive and electrically insulating "EVA composite encapsulants for solar photovoltaic (PV) cell”, eXPRESS Polymer Letters, Vol.2, No.5 (2008) 357–363
[39] Yen Ng et al, “Thermal Conductivity of Boron Nitride-Filled Thermoplastics: Effect of Filler Characteristics and Composite Processing Conditions”, Polymer Composites, Volume 26, Issue 6, pages 778–790, December 2005
[40] “'Cool' graphene might be ideal for thermal management in nanoelectronics”, Nanowerk Spotlight
http://www.nanowerk.com/spotlight/spotid=4641.php
[41] Argwal et al, “Thermal Conductivity of Polymer Nanocomposites Made With Carbon Nanofibers”, Polymer Engineering & Science, Volume 48, Issue 12, pages 2474–2481, December 2008
[42] http://www.americanelements.com/cgnf.html
52
[43] Ramani NC, "Analysis of the Effect of Pressure on Compression Moulding of UHMWPE", J. Mat. Science: Mat. In Medicine, 9: 165-172, 1998 [44] Thermogravimetric Analysis Theory, Operation, Calibration and Data Interpretation, TA Instruments Presentation, Kadine Mohamed [45] Determining Volatiles in Polyethylene Terephthalate Using the Q5000 IR Thermogravimetric Analyzer, R. Bruce Cassel Ph.D . TA Instruments [46] Koledintseva et al, “Maxwell Garnett Model for Dielectric Mixtures Containing Conducting Particles at Optical Frequencies”, Progress In Electromagnetics Research, PIER 63, 223–242, 2006
top related