See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/328236157

Heat Dissipation in Epoxy/Amine-Based Gradient Composites with Alumina

Particles: A Critical Evaluation of Thermal Conductivity Measurements

Article  in  Polymers · October 2018

DOI: 10.3390/polym10101131

CITATIONS READS

4 107

6 authors, including:

Matthias Morak Mario Gschwandl

Polymer Competence Center Leoben GmbH Polymer Competence Center Leoben GmbH
3 PUBLICATIONS   4 CITATIONS    9 PUBLICATIONS   10 CITATIONS   

SEE PROFILE SEE PROFILE

P.F. Fuchs Frank Wiesbrock

Polymer Competence Center Leoben GmbH Polymer Competence Center Leoben GmbH
29 PUBLICATIONS   86 CITATIONS    87 PUBLICATIONS   2,067 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Mimiflow View project

All content following this page was uploaded by Matthias Morak on 15 October 2018.

The user has requested enhancement of the downloaded file.

 polymers
Article
Heat Dissipation in Epoxy/Amine-Based Gradient
Composites with Alumina Particles: A Critical
Evaluation of Thermal Conductivity Measurements
Matthias Morak 1,2,† , Philipp Marx 1,3,† , Mario Gschwandl 1,2 , Peter Filipp Fuchs 1,2 ,
Martin Pfost 4 and Frank Wiesbrock 1,5, *
1 Polymer Competence Center Leoben GmbH, Roseggerstrasse 12, 8700 Leoben, Austria;
matthias.morak@pccl.at (M.M.); philipp.marx@pccl.at (P.M.); mario.gschwandl@pccl.at (M.G.);
peterfilipp.fuchs@pccl.at (P.F.F.)
2 Chair of Mechanics, Montanuniversitaet Leoben, Franz-Josef-Strasse 18, 8700 Leoben, Austria
3 Chair of Chemistry of Polymeric Materials, Montanuniversitaet Leoben, Otto Gloeckel-Straße 2,
8700 Leoben, Austria
4 Chair of Energy Conversion, TU Dortmund University, Emil-Figge-Strasse 50, 44227 Dortmund, Germany;
martin.pfost@tu-dortmund.de
5 Institute for Chemistry and Technology of Materials, Graz University of Technology, NAWI Graz,
Stremayrgasse 9, 8010 Graz, Austria
* Correspondence: frank.wiesbrock@pccl.at; Tel.: +43-3842-42962-42
† Both authors share the first authorship.




Received: 30 July 2018; Accepted: 8 October 2018; Published: 11 October 2018


Abstract: For the design of the next generation of microelectronic packages, thermal management
is one of the key aspects and must be met by the development of polymers with enhanced thermal
conductivity. While all polymer classes show a very low thermal conductivity, this shortcoming
can be compensated for by the addition of fillers, yielding polymer-based composite materials with
high thermal conductivity. The inorganic fillers, however, are often available only in submicron- and
micron-scaled dimensions and, consequently, can sediment during the curing reaction of the polymer
matrix. In this study, an epoxy/amine resin was filled with nano- and submicron-scaled alumina
particles, yielding a gradient composite. It was found that the thermal conductivity according to
laser flash analysis of a sliced specimen ranged from 0.25 to 0.45 W·m−1 ·K−1 at room temperature.
If the thermal conductivity of an uncut specimen was measured with a guarded heat flow meter,
the ‘averaged’ thermal conductivity was measured to be only 0.25 W·m−1 ·K−1 . Finite element
analysis revealed that the heat dissipation through a gradient composite was of intermediate speed
in comparison with homogeneous composites exhibiting a non-gradient thermal conductivity of 0.25
and 0.45 W·m−1 ·K−1 .

Keywords: epoxy/amine resin; alumina particles; thermal conductivity; thermal management;

gradient composite

1. Introduction
Thermal management is one of the key aspects in the design of reliable microelectronic
packages [1,2] as well as high-voltage machinery [3,4]. Geometry and materials have to be defined in a
way that the packages, insulations, multi-layer assemblies, etc. can withstand the application-specific
external or internal temperature loads. In microelectronics, the most critical load often originates
from an internally generated heat by active components such as a power metal-oxide-semiconductor
field-effect transistor (MOSFET) [5–8]. This type of silicon chip can produce high temperatures in short

Polymers 2018, 10, 1131; doi:10.3390/polym10101131 www.mdpi.com/journal/polymers

Polymers 2018, 10, 1131 2 of 13

times, which results in pronounced temperature gradients and high thermo-mechanical strains [9]. Due
to the ongoing demand for integrated functions and miniaturization, geometric changes of the design
are subject to limitations, which add additional importance to increasing the thermal conductivity of
the materials used [10].
Polymers are commonly used materials in power packages. Despite their broad versatility
in material properties and functionality, all polymer classes show one common physico-chemical
characteristic, namely, a very low thermal conductivity commonly in the range of 0.1 to
0.2 W·m−1 ·K−1 [4]. Very few polymers with dedicated structural motifs show a higher thermal
conductivity of 0.3 W·m−1 ·K−1 [11]. This characteristic can be overcome by the addition of fillers
with high thermal conductivity [4,12–14]. Due to the low heights of films in power packages or
insulation layers down to the µm range, such (inorganic) fillers are preferentially used in nano- and/or
submicron-scaled sizes. Aiming at high thermal conductivity of a composite material, highly priced
fillers with high thermal conductivity such as nanodiamonds (λ = 2200 W·m−1 ·K−1 ) [15], hexagonal
boron nitride (λ = 390 W·m−1 ·K−1 ) [16], and/or aluminum nitride (λ = 300 W·m−1 ·K−1 ) [17] may
be added; reasonably priced alternatives are comprised of silica (λ = 0.7 W·m−1 ·K−1 ) and alumina
(λ = 23 W·m−1 ·K−1 ) [18].
In cases of some inorganic fillers, nano-scaled particles are not available in quantities relevant
for industrial production and/or at reasonable prices; this is particularly true for nanodiamonds and
hexagonal boron nitride due to, among other things, their hardness. It must be taken into account
that particles with diameters above the nano-scale may sediment in a composite material during
the curing reaction, yielding a gradient composite with varying composition along the height scale
(e.g., increasing content of inorganic fillers from top to bottom) and, consequently, an analogously
varying thermal conductivity. The phenomenon of sedimentation eventually occurs to even higher
extent due to the agglomeration of (non-functionalized) particles, despite their initial homogeneous
dispersion in a polymer matrix [14].
Correspondingly, this study aimed at investigating the effect of particle sedimentation in an
epoxy/amine resin and the consequences on the thermal conductivity of and the thermal dissipation
through such a composite gradient. In order to trigger sedimentation of the particles, a mixture of
nano- and submicron-scaled alumina particles was used. The thermal conductivity was comparatively
quantified by laser flash analysis and a guarded heat flow meter. The experimental study was
complemented by modelling efforts and finite element analysis in order to detail the thermal properties
of gradient composites. With the help of numerical simulations based on thermal finite element models,
the thermal properties of various types of materials, including the gradient composites detailed in this
study, can be calculated in straightforward fashion, as this type of simulation models predominantly
considers the thermal flux through the material. As such, it inherently provides the opportunity to
calculate the temperature field without consideration of stresses, deformations, or electrical fields.

2. Materials and Methods

2.1. Materials
Bisphenol A diglycidyl ether (DGEBA) and diethylenetriamine (DETA) were purchased from
Sigma Aldrich (Vienna, Austria). The Al2 O3 nanoparticles (20–30 nm) and Al2 O3 submicron-particles
(200 nm) were bought from ABCR (Karlsruhe, Germany). The range of diameters of the particles
provided by the suppliers was verified by transmission electron microscopy (TEM and scanning
electron microscopy (SEM) measurements of polymer films containing one type of alumina particles;
dynamic light scattering (DLS) measurements failed to reproduce the diameters of the individual
particles due to aggregation of the non-functionalized alumina particles (Figure 1). All chemicals were
used as received.
Polymers
Polymers 2018,
2018, 10,
10, x1131
FOR PEER REVIEW 33 of
of 13
13

Figure 1. Dynamic light scattering (DLS) measurements of particles in water (left) as well as TEM/SEM
Figure 1. Dynamic light scattering (DLS) measurements of particles in water (left) as well as
images of the particles embedded in a polymer matrix (right), referring to the alumina nanoparticles
TEM/SEM images of the particles embedded in a polymer matrix (right), referring to the alumina
(top) and the alumina submicron particles (bottom). The ranges of diameters of 20–30 and 200 nm,
nanoparticles (top) and the alumina submicron particles (bottom). The ranges of diameters of 20–30
respectively, which were provided by the suppliers, could be verified by the TEM/SEM images; the
and 200 nm, respectively, which were provided by the suppliers, could be verified by the TEM/SEM
DLS measurements revealed aggregation of the particles.
images; the DLS measurements revealed aggregation of the particles.
2.2. Instrumentation
2.2. Instrumentation
The thermal diffusivity was measured on a LFA 467 Hyperflash system (Netzsch, Selb, Germany)
The thermalof diffusivity
at temperatures was
20, 40, 60, 80, 100,measured
120, and 140on ◦aC.LFA 467 Hyperflash
The samples were coated systemwith(Netzsch, Selb,
a thin graphite
Germany)
layer before at temperatures
the measurement of 20,in40, 60, 80,
order to 100,
avoid120, and 140 °C.
reflection. The The samples wereofcoated
measurements with a thin
the specific heat
graphite layer before the measurement in order to avoid reflection. The
capacity were performed in duplicate on DSC 6000 (PerkinElmer, Vienna, Austria) using sapphire measurements of the specific
heat capacitymaterial.
as reference were performed
The samples in duplicate
were heated on upDSC 6000
twice from −10 to 150 ◦Vienna,
(PerkinElmer, C with a Austria) using
heating rate of
sapphire − 1
10 K·min as .reference
The density material. The samples
was determined withwere heated upbalance
a hydrostatic twice from
XS205−10Dual to Range
150 °C (Mettler-Toledo
with a heating
rate
GmbH, of 10 K·min
Vienna, Austria) by weighing samples in air and water at 25 ◦ C. The
−1. The density was determined with a hydrostatic balance
thermalXS205 Dual Range
conductivity of the
(Mettler-Toledo GmbH, Vienna,
uncut gradient composites Austria) by
was measured onweighing
a DTC-300 samples
thermalinconductivity
air and watermeter at 25(TA
°C. The thermal
Instruments,
conductivity
New Castle, DE, of the
USA)uncut gradient
according to composites
the norm ASTM was measured on a DTC-300
E1530 at temperatures thermal
of 20, 40, 60,conductivity
80, 100, 120,
meter (TA◦ Instruments, New Castle, DE, USA) according to the norm
and 140 C, using specimens with diameters of 50 mm and heights of 10 mm. For the dispersion ASTM E1530 at temperatures
of
of 20,
the 40, 60, and
nano- 80, 100, 120, and 140 °C,
submicroparticles inusing specimens
the resins, with diameters
the Dissolver Dispermat of 50 AE mm 03 and heights of 10
(VMA-Getzmann
mm. For the dispersion of the nano- and submicroparticles in the
GmbH, Reichshof, Germany) was used (20 min, 5000 rpm). Particle sizes were measured in resins, the Dissolver Dispermat
aqueous
AE
media (5 mg/20 g) by dynamic light scattering using a Litesizer 500 (Anton Paar, Austria). sizes
03 (VMA-Getzmann GmbH, Reichshof, Germany) was used (20 min, 5000 rpm). Particle TEM
were
images measured in aqueous
were recorded on a FEImedia
Tecnai (5 12
mg/20 g) by dynamic
transmission electronlight scattering
microscope. using ameasurements
SEM-EDX Litesizer 500
(Anton Paar, Austria).
were performed TEM images
using a Tescan were recorded
Vega 3 scanning electronon a FEI Tecnai
microscope with an 12 energy
transmission
dispersive electron
X-ray
microscope.
spectrometerSEM-EDX
(EDX Oxford measurements
Instruments were performed
INKAx-act, (Highusing a Tescan
Wycombe, UK))Vega 3 scanning
attached. Electronelectron
energy
microscope
levels were set with an kV.
to 20 energy dispersive X-ray spectrometer (EDX Oxford Instruments INKAx-act,
(High Wycombe, UK)) attached. Electron energy levels were set to 20 kV.
2.3. Preparation of the Gradient Epoxy-Amine-Based Composite
2.3. Preparation of the Gradient Epoxy-Amine-Based Composite
For the preparation of the test specimens containing 20 wt % of Al2 O3 particles, 24.39 g of Al2 O3
nanoparticles and 24.39 of
For the preparation g of
theAl 2 O3specimens
test submicroparticles
containing were
20 wt added
% of toAl2174 g of DGEBA.
O3 particles, 24.39In order
g of Al2Oto3
achieve a homogenous
nanoparticles and 24.39dispersion,
g of Al2Othe mixture was stirred
3 submicroparticles werewith a high-shear
added to 174 gmixer at r.t. Subsequently,
of DGEBA. In order to
the particle-DGEBA
achieve a homogenous dispersion was mixed
dispersion, with DETA
the mixture in a mass
was stirred with ratio
a of 10.56:1 (molar
high-shear mixer ratio of
at r.t.
DGEBA:DETAthe
Subsequently, = 2.5:1) and stirred for
particle-DGEBA 5 min at was
dispersion r.t. For the removal
mixed with DETA of airinbubbles,
a mass the ratiomixture was
of 10.56:1
sonicated
(molar forof5 min,
ratio poured into= a2.5:1)
DGEBA:DETA moldandwith the targeted
stirred for 5 mingeometry, cured
at r.t. For the at r.t. for 4ofh,air
removal and post-cured
bubbles, the
mixture was sonicated for 5 min, poured into a mold with the targeted geometry, cured at r.t. for 4 h,
Polymers 2018, 10, 1131 4 of 13

at 120 ◦ C for 1 h. For the LFA measurements, a specimen with a diameter of 25 mm and a height
of 10 mm was cut into layers with a thickness of approximately 800 µm using a diamond-equipped
microtome. For the measurement of the “averaged” thermal conductivity of the bulk material, a
specimen with a diameter of 50 mm and a height of 10 mm was produced.

2.4. Simulation Model

In order to evaluate the effect of a material with gradual thermal conductivity distribution, a
simulation model using an uncoupled heat transfer analysis was employed. The term “uncoupled”
summarizes that the temperature field was calculated without consideration of the stresses,
respectively deformations, or any electrical fields; the model nonetheless can include conduction,
boundary convection, and boundary radiation. Since in the present case the material properties are
temperature-dependent, the heat transfer analysis is nonlinear. In order to consider the internal energy
within the calculation, a transient analysis was carried out, in which the time integration is performed
according to the backward Euler scheme [19]. The basic equation for an uncoupled heat transfer
analysis is the energy balance equation (Equation (1)) [20]
Z . Z Z
ρUdx = qds x + rdx (1)
Ω ∂Ω Ω
.
in which Ω is the volume of a solid material, ∂Ω the surface of the volume, ρ the density, U the rate of
the internal energy and where U = U (θ ) with the temperature θ, q the heat flux, and r the internal heat.
The internal energy depends on the temperature only, which is usually defined in terms of a
specific heat (Equation (2))
dU
c(θ ) = (2)
dθ
The equation for the thermal equilibrium can be derived from the energy balance equation
(Equation (1)) by means of the specific heat and the term of heat conduction governed by Fourier’s law
(Equations (3) and (4))
dθ
f = −k ∀ x ∈ Ω (3)
dx
dθ
f ∂Ω = −n·k· ∀ x ∈ ∂Ω (4)
dx
The overall energy balance equation (Equation (5)) [21] is obtained as
Z     Z  
∂θ ∂ ∂θ ∂θ
δθ ρc − · k· − q dx + δθ n·k· − qs dsx = 0 (5)
Ω ∂t ∂x ∂x ∂Ω ∂x

in which δΩ is the variational field satisfying the essential boundary conditions, ρ(θ ) the
temperature-dependent density, c(θ ) the temperature-dependent specific heat, k(θ ) = k(θ )I the
isotropic temperature-dependent thermal conductivity, q the heat added from arbitrary heat sources,
and qs = h(θ − θ0 ) the surface convection with the film coefficient h.
The simulation model itself consists of different parts considering the individual dimensions and
material parameters (Figure 2, Tables 1 and 2). Additionally, the measured data of the epoxy resin
(Figures 6 and 7) were assigned to the corresponding layers of the model.
Polymers 2018, 10, 1131 5 of 13
Polymers 2018, 10, x FOR PEER REVIEW 5 of 13

Figure 2.
Figure 2. Profile of
of the
the simulation
simulation model. Part 1 and part 3 are copper layers, part 2 is a silicon layer,
and
and part
part 44 is
is the
the epoxy-amine-based
epoxy-amine-based composite,
composite, which
which was
was divided
divided in
in eight
eightlayers.
layers.

Table 1. Parts and layer dimensions of the model.

Table 1. Parts and layer dimensions of the model.
Layer Length (mm) Width (mm) Height (mm)
Layer Length [mm] Width [mm] Height [mm]
Copper (layer
Copper (layer1)1) 9.5
9.5 9.5
9.5 11
Silicon 4.5 4.5 0.15
Silicon 4.5 4.5 0.15
Copper (layer 2) 4.5 4.5 0.05
Copper (layer 2)
Epoxy 4.5
9.5 4.5
9.5 0.05
2.2
Epoxy
Epoxy layers 9.5
0.3125 9.5
0.3125 2.2
0.25
Epoxy layers 0.3125 0.3125 0.25
Table 2. Material parameters according to literature data.
Table 2. Material parameters according to literature data.
g J
Material Temperature (◦ C) Thermal Conductivity ( mKkW
) Density ( mm3 ) Specific Heat ( kgK )
Temperature Thermal Conductivity 𝐠 𝐉
Material
Copper 25 10−1
4.0 ×𝐤𝐖 8.93 × 10−3
Density Specific
3.85 × 102
Heat
[°C] 𝐦𝐦𝟑 𝐤𝐠𝐊
𝐦𝐊
25 1.670 × 10−−11 2.329 × 10−3 5.769 × 102 2
Copper
Silicon 25
400
4.0 × 10
5.257 × 10−2
8.93 × 10−3
2.329 × 10−3
3.85 × 10
8.741 × 102
25 1.670 × 10−1 2.329 × 10−3 5.769 × 102
Silicon
400 5.257 × 10−2 2.329 × 10−3 8.741 × 102
In the field of microelectronics, a wide power range from mW up to GW exists. Hence, at the
beginning
In theof the of
field analysis, several values
microelectronics, for the
a wide powerheatrange
flux were
from tested,
mW upaiming
to GWtoexists.
find aHence,
suited value
at the
enabling to reach the temperature values that can occur during the power thermal cycle
beginning of the analysis, several values for the heat flux were tested, aiming to find a suited (PTC) testvalue
[22].
W
In final consequence,
enabling to reach thea temperature
surface heat flux of 75
values that wasoccur
areacan applied in a region
during withthermal
the power a radiuscycle
of 1 mm
(PTC)on test
top
ofPolymers
the silicon layer (Figure 3). This results in a heat flux of approx. 23.87 W · mm −2 .
2018, consequence,
[22]. In final a surface heat flux of 75
10, x FOR PEER REVIEW was applied in a region with a radius of 6 of113

mm on top of the silicon layer (Figure 3). This results in a heat flux of approx. 23.87 W·mm−2.
The entire model system was meshed with 8-noded linear heat transfer brick elements
(DC3D8), respectively, with 4-noded linear heat transfer tetrahedral elements (DC3D4), whereas the
total amount of elements of the epoxy layer was 112,664 tetrahedral elements. In order to
demonstrate the effects of the gradual heat conductive distribution, three simulations with different
material data of the epoxy-amine-based composite were performed: firstly, a simulation with the
layer-wise measured epoxy data (Figures 6 and 7), secondly, a simulation with the data gained from
Figure
Figure 3. 3. Area
Area ofof
thethe simulation
simulation model,
model, inin which
which thethe surface
surface heat
heat flux
flux was
was applied
applied (cp.
(cp. Figure
Figure 2).2).
the guarded heat flow meter (GHFM) measurements, and, finally, an idealized simulation, in which
it was
TheAsassumed
entire that system
initialmodel each of
temperature wasthe eightr.t.
meshed
condition, epoxy
with 23 layers
of 8-noded
°C was entails
linear thetransfer
heat
considered. highest thermal
brick
A cyclic conductivity
elements
thermal (DC3D8),
loading and
measured (bottom
respectively, with layer linear
4-noded 8 in Figures
heat 6 and
transfer 7). Since elements
tetrahedral the calibration
(DC3D4), of whereas
the filmthecoefficient
total was
amount
unloading was applied (Figure 4). In total, five loading cycles with a surface heat flux of 75 of
impossible,
of elements of several
the epoxyfilmlayer
coefficients weretetrahedral
was 112,664 evaluated.elements.
The overall behavior
In order remains the
to demonstrate the effects
same,
0.1 s, interrupted by five unloading cycles of 1 s (after the first four loadings), respectively 3 s (after
independent
of the gradualofheat theconductive
film coefficient, only three
distribution, the absolute
simulations temperature values are shifted
with different data ofand,
the fifth loading), were applied, yielding a total simulation time of 7.5 s. The material
thermal loading the
cycle
accordingly, the temperature difference between the simulation results is
epoxy-amine-based composite were performed: firstly, a simulation with the layer-wise measured more or less pronounced
with the abovementioned parameters was chosen in alignment to the power thermal cycle (PTC)
in dependence
epoxy of the
data (Figures film7),coefficient.
6 and secondly, aBoundary
simulationconvection was taken frominto account by defining a
test, which is one type of conventional test methodwith in thethe data
field ofgained the guarded
microelectronics [22,24].heat flow
film coefficient
meter (GHFM) of 10
measurements, from literature [23].
and, finally, an idealized simulation, in which it was assumed that
each of the eight epoxy layers entails the highest thermal conductivity measured (bottom layer 8 in
Polymers
Polymers 2018,
2018, 10,
10, xx FOR
FOR PEER
PEER REVIEW
REVIEW 66 of
of 13
13

Polymers 2018, 10, 1131 6 of 13

Figures 6 and 7). Since the calibration of the film coefficient was impossible, several film coefficients
were evaluated. The overall behavior remains the same, independent of the film coefficient, only
the absolute temperature values are shifted and, accordingly, the temperature difference between
Figure
Figure 3.
the simulation 3. Area
Area of
of the
results simulation
themore
is simulation
or less model,
model, in
in which
pronounced which the surface
surface heat
thedependence
in heat flux
flux was applied
wasfilm
of the applied (cp.
(cp. Figure
coefficient.Figure 2).
2).
Boundary
W
convection was taken into account by defining a film coefficient of 10 m2 K from literature [23].
As initial
As initial temperature
initial temperature condition,
temperature condition,
condition, r.t.r.t. of
r.t. of 23
23◦°C
of 23 °C was considered.
considered. A
was considered. cyclic
A cyclic thermal
cyclic thermal loading
thermal loading
loading and and
and
As C was A
unloading
unloadingwas
unloading was applied
wasapplied (Figure
applied(Figure
(Figure 4). In
4).4).InIn total,
total,
total, fivefive
five loading
loading
loading cycles
cycles
cycles withwith
with a surface
a surface
a surface heat
heatheat
fluxflux 75
of 75ofarea75of 0.1of
flux of W of
s,
0.1 s, interrupted
interrupted by fiveby
0.1 s, interrupted by five
unloadingunloading
cycles of
five unloading cycles of
1 s (after
cycles 1 s
of 1 the (after the
first the
s (after first
fourfirst four
loadings), loadings), respectively
respectively
four loadings), 3
3 s (after 3the
respectively s (after
fifth
s (after
the
the fifth
fifth loading),
loading), were
were applied,
were applied,
loading), applied, yielding
yielding ayielding aa total
total simulation
total simulation time of 7.5
simulation time of
of 7.5
s. The
time s.
s. The
thermal
7.5 thermal
Theloading
thermalcycle loadingwithcycle
loading the
cycle
with the
abovementionedabovementioned
parameters parameters
was chosen was
in chosen
alignment intoalignment
the power
with the abovementioned parameters was chosen in alignment to the power thermal cycle (PTC) to the
thermal power
cycle thermal
(PTC) cycle
test, which(PTC) is
test,
one which
test,type
which is one type
of conventional of
is one type oftestconventional
method intest
conventional test method
the method in the field
field of microelectronics of microelectronics
[22,24].
in the field of microelectronics [22,24].
[22,24].

Figure 4.
Figure4.
Figure Thermal
4.Thermal loading
Thermalloading and
loadingand unloading
andunloading steps.
unloadingsteps.
steps.

3.
3. Results
3. Results
3.1.
3.1. Concept and Preparation of the Gradient Composites
3.1. Concept
Concept andand Preparation
Preparation of of the
the Gradient
Gradient Composites
Composites
In
In this study, the thermal conductivity of aa gradient composite based on an epoxy-amine resin
In this
this study,
study, thethe thermal
thermal conductivity
conductivity of of a gradient
gradient composite
composite based
based on on an
an epoxy-amine
epoxy-amine resin resin
composed
composed of DGEBA and DETA was investigated. For the preparation of the test specimens, 10 wt %
composed of of DGEBA
DGEBA and and DETA
DETA was was investigated.
investigated. For For the
the preparation
preparation of of the
the test
test specimens,
specimens, 10 10 wt
wt %%
of
of alumina
alumina submicroparticles
submicroparticles (diameter
(diameter ofof 200
200nm)
nm) and
and 1010wtwt% %of alumina
of alumina nanoparticles
nanoparticles (diameter
(diameterof
of alumina submicroparticles (diameter of 200 nm) and 10 wt % of alumina nanoparticles (diameter
20–30
of nm) were were dispersed homogenously in the epoxythe resin usingresina high-shear amixer. Subsequently,
of 20–30
20–30 nm) nm) were disperseddispersed homogenously
homogenously in in the epoxy
epoxy resin using using a high-shear
high-shear mixer.mixer.
the resin
Subsequently,was cured
the at
resin r.t.
was for 4 h.
cured Due
at to
r.t. their
for 4 higher
h. Due weight,
to their the submicroparticles
higher weight, the were expected
submicroparticles
Subsequently, the resin was cured at r.t. for 4 h. Due to their higher weight, the submicroparticles
to sediment during
were the curing reaction (Figure 5), while the smaller nanoparticles were expected to
were expected
expected to to sediment
sediment duringduring the the curing
curing reaction
reaction (Figure
(Figure 5),5), while
while thethe smaller
smaller nanoparticles
nanoparticles
remain
were dispersed throughout the resin. Due to thethe sedimentation, a gradient of the particle distribution
were expected
expected to to remain
remain dispersed
dispersed throughout
throughout the resin. resin. Due
Due toto the
the sedimentation,
sedimentation, aa gradient
gradient of of the
the
was expected
particle to occur
distribution and,
was correspondingly,
expected to occur a gradient
and, of the thermal
correspondingly, conductivity
a gradient over
of thethe height
thermal
particle distribution was expected to occur and, correspondingly, a gradient of the thermal
of the specimen. Inthe
conductivity order to quantify the gradientInoforder the thermal conductivity, the curedthecomposite
conductivity over over the heightheight of of the
the specimen.
specimen. In order to to quantify
quantify thethe gradient
gradient of of the thermal
thermal
specimens
conductivity, with heights of 10 mm were cut into layers with a thickness of approximately 800 µma
conductivity, the the cured
cured composite
composite specimens
specimens with with heights
heights ofof 10
10 mmmm were
were cutcut into
into layers
layers with
with a
using a microtome.
thickness of The thermal
approximately 800 conductivity
μm using a of each individual
microtome. The layer was
thermal measuredofwith
conductivity each a individual
laser flash
thickness of approximately 800 μm using a microtome. The thermal conductivity of each individual
analysis
layer system.
layer was
was measured
measured with with aa laser
laser flash
flash analysis
analysis system.
system.

Figure
Figure5.
Figure 5.Schematic
5. Schematicoverview
overviewof
ofthe
theconcept
conceptand
andthethepreparation
preparationof
ofgradient
gradientcomposites
compositescontaining
composites containingaaa
containing
mixture
mixture of
of nano-
nano- and
and submicron-scaled
submicron-scaled particles.
particles.
mixture of nano- and submicron-scaled particles.
Polymers 2018, 10, 1131 7 of 13

3.2. Measurement of the Thermal Conductivity

The thermal conductivity is one key thermo-physical material parameter, quantifying the heat
transfer through a material. Several measurement techniques are available for the determination of
the thermal conductivity. In case of thin specimens, laser flash analysis (LFA) is considered the most
suitable approach for the determination of thermal conductivity, which is a contactless, non-destructive,
and transient measurement approach. A very short energy pulse (e.g., laser or light pulse) is applied
to heat the front surface of a plane-parallel specimen, while, on the rear surface, an infrared detector is
used to measure the temperature increase. Assuming a fully adiabatic system, the thermal diffusivity
a(T) of the sample can be calculated using the half-life period t1/2 of the temperature increase and the
sample height d (Equation (6)) [25]
d
a( T ) = 0.1388 (6)
t1/2
The LFA is an indirect measurement method for the thermal conductivity λ(T) as it measures the
thermal diffusivity a(T). Nonetheless, using a(T), the density ρ(T) and the specific heat capacity cp (T),
λ(T) can be calculated (Equation (7))
λ ( T ) = a ( T ) ρ ( T ) c p ( T ). (7)

Consequently, additional measurements are necessary, for differential scanning calorimetry

(DSC) for the specific heat capacity measurement as well as a hydrostatic balance for the density
measurement. Note that, the density is only characterized at room temperature, due to the negligible
change over temperature.
As mentioned above, the composites were cut into evenly dimensioned layers, all of which were
measured, unless their surfaces were too rough for uncorrupted LFA measurements (layers number 1
and 6). All layers show densities in the range of 1.33 to 1.38 g·cm−3 (Table 3). The LFA measurements
of the layers show a clearly visible gradient of the thermal diffusivity a(T) over the different layers
ranging from 0.20 to 0.35 mm2 ·s−1 at r.t. (Figure 6). The bottom layer (layer number 8 at position
h = 8.448 mm) shows a pronouncedly higher thermal diffusivity in comparison with the other layers.
Hence, it may be assumed that most of the (submicron-scaled) particles sedimented in the composite
material during the curing reaction. The results of the consecutive layers (layer numbers 7 to 2) as well
show a gradient of the thermal diffusivity.

Table 3. Position and density of the individual layers.

Height Position (mm) Layer (#) Density ρ (g·cm−3 )

1.592 2 1.364
2.767 3 1.376
3.856 4 1.364
4.961 5 1.367
7.177 7 1.368
8.448 8 1.372

The measurement of the specific heat capacity cp (T) by DSC yielded no gradient behavior. In the
temperature range from 20 to 140 ◦ C, cp (T) varied from approx. 1.0 to 1.6 J·g−1 ·K−1 (Figure 7 left).
The non-gradient behavior can be referred to the similar specific heat capacities of the filler Al2 O3
(cp (293 K) = 0.8 J·g−1 ·K−1 [26]) and the epoxy/amine resin (cp (293 K) = 1.2 J·g−1 ·K−1 [27]). As the
density and the specific heat capacity vary to a very small extent, only within an individual layer
at a given temperature, the thermal conductivity is quasi-proportional to the thermal diffusivity
(Equation (7)), and the thermal conductivity shows similar trends like the thermal diffusivity (Figure 7
right). The thermal conductivity of the bottom layer (layer number 8) is significantly increased, likely
due to the assumed filler agglomeration in that layer, while the consecutive layers display a clear
gradient to the topmost layer (layer number 2). In summary, the thermal conductivity at r.t. ranged
Polymers 2018, 10,
Polymers 2018, 10, 1131
x FOR PEER REVIEW 88 of
of 13
13
Polymers 2018, 10, x FOR PEER REVIEW 8 of 13

0.45 W·m−1 ·K−1

−1;; unfilled epoxy-amine resins exhibit a thermal conductivity of approx. 0.2 W·m−1 −1·K
−1
0.45 W·m
from 0.25 −1to·K0.45 W·m−1epoxy-amine
unfilled ·K−1 ; unfilledresins exhibit aresins
epoxy-amine thermal conductivity
exhibit a thermalofconductivity
approx. 0.2 of
W·m ·K−1
approx.
[14].
[14].W·m−1 ·K−1 [14].
0.2

Figure 6. (Left) Thermal diffusivity a(T) of the different layers according to laser flash analysis (LFA)
Figure 6. (Left) Thermal diffusivity a(T) of the different layers according to laser
laser flash analysis
analysis (LFA)
(LFA)
measurements. (Right) Gradient of the thermal diffusivity along the different layers.
measurements. (Right) Gradient of the thermal diffusivity
diffusivity along
along the
the different
different layers.
layers.

Figure 7.7.(Left)
(Left)Specific
Specific heat
heat capacity
capacity measurements
measurements fordifferent
for the the different layersdifferential
layers using using differential
scanning
Figure 7. (Left) Specific heat capacity measurements for the different layers using differential
calorimetry
scanning (DSC). (Right)
calorimetry (DSC). Calculated thermalthermal
(Right) Calculated conductivity of the of
conductivity different layerslayers
the different based on the
based on
scanning calorimetry (DSC). (Right) Calculated thermal conductivity of the different layers based on
measurements
the measurements of the
of thermal diffusivity,
the thermal the density,
diffusivity, and the
the density, andspecific heat heat
the specific capacity.
capacity.
the measurements of the thermal diffusivity, the density, and the specific heat capacity.
For the
For the determination
determination of of the
the averaged
averaged thermal
thermal conductivity
conductivity ofof such
such gradient
gradient composites,
composites, also
also
GHFM For the determination
measurements of a of the
non-slicedaveraged
specimenthermal
had conductivity
been carried of
out. such
This gradient
method is composites, also
well-established
GHFM measurements of a non-sliced specimen had been carried out. This method is
GHFM measurements
and commonly used of a non-sliced specimen had been carried out. This method is is
well-established and for the thermal
commonly conductivity
used measurement
for the thermal of polymers
conductivity [28–30].
measurement ofThe GHFM[28–
polymers a
well-established
stationary and
measurement commonly
method, used for
whereby the
thethermal
specimenconductivity
is placed measurement
between a hot of
and polymers
cold [28–
plate. A
30]. The GHFM is a stationary measurement method, whereby the specimen is placed between a hot
30]. The
fluxGHFM is a stationary measurement method, whereby
heat the specimen is placed betweenUsing
a hot
heat
and cold transducer
plate. A heatis flux
used, which
transducerdetermines the steady
is used, which determines the Q
flow through
steady heatthe
flowspecimen.
Q through the
and cold plate.
Fourier’s law, A heat flux
theFourier’s transducer
thermal conductivity is used,
can be which determines
calculated can
(Equationthe steady
(8)) heat flow Q through the
specimen. Using law, the thermal conductivity be calculated (Equation (8))
specimen. Using Fourier’s law, the thermal conductivity can be calculated (Equation (8))
Qd
λ( T ) = 𝑄𝑑
𝑄𝑑 (8)
𝜆 𝑇 =∆TA
𝜆 𝑇 = Δ𝑇𝐴
(8) (8)
Δ𝑇𝐴
in which∆T
in which ΔTisisthe
thetemperature
temperature difference between
difference the hot
between theandhotcold
and coldAplate,
plate, is the area
A isofthethearea
specimen’s
of the
in which
surface ΔTisisinthe
that temperature
contact with the difference
heating between
plates, and d the
is hot
the and cold
thickness of plate,
the A is the area of the
sample.
specimen’s surface that is in contact with the heating plates, and d is the thickness of the sample.
specimen’s
The GHFM surface that is in contact with the heating plates, and d isbulkthe thicknessofof0.25 theW sample.
·m−1−1·K −1
The GHFM measurements
measurementsreveal revealaathermal
thermalconductivity
conductivityofofthe the bulkmaterial
material of 0.25 W·m −1·K
−1
at r.t.The GHFM
(Figure 8). measurements
Apparently, reveal
over the a thermal
whole rangeconductivity
of of
temperaturesthe bulk
from material
20 to 140of◦ 0.25
C, the W·m ·K
thermal
−1
at r.t. (Figure 8). Apparently, over the whole range of temperatures from 20 to 140 °C, the thermal
at r.t. (Figure of
conductivity 8).the
Apparently, over reproduces
bulk material the whole range of temperatures
the thermal fromobserved
conductivity 20 to 140in °C,the
thetop thermal
layer
conductivity of the bulk material reproduces the thermal conductivity observed in the top layer
conductivity
(layer number of2)the
of bulk
the material
cut specimen reproduces
(see the thermal
hereinabove), conductivity
which is the layer observed
with thein the
lowest top layer
thermal
(layer number 2) of the cut specimen (see hereinabove), which is the layer with the lowest thermal
(layer numberThis
conductivity. 2) of the cut specimen (see hereinabove), which is Tt
themust
layerbewith the that
lowest thermal
conductivity. This measurement
measurement has has been
been performed
performed in in triplicate.
triplicate. Tt must be argued
argued that the
the overall
overall
conductivity. This
thermal conductivity measurement has been performed in triplicate. Tt must be argued that
tothe
the overall
thermal conductivityofofthe bulk
the material
bulk materialaccording
accordingto GHFMto GHFMmeasurementscorresponds
measurementscorresponds lowest
to the
thermal conductivity
thermal conductivity of of an
theindividual
bulk materiallayeraccording
of the to GHFM
gradient measurementscorresponds
material measured by LFA. to the
Hence, in
lowest thermal conductivity of an individual layer of the gradient material measured by LFA.
lowest measurements,
GHFM thermal conductivitythe of anthermal
overall individual layer ofofthe
conductivity an gradient materialmaterial
inhomogeneous measured is by LFA.
limited by
Hence, in GHFM measurements, the overall thermal conductivity of an inhomogeneous material is
Hence,
the in GHFM
lowest thermal measurements,
conductivity the overall
present thermal
in the conductivity ofThe
(gradient) an inhomogeneous material is
limited by the lowest thermal conductivity present in the material. heat dissipation
(gradient) material. from ‘hot
The heat dissipation
limited by the lowest thermal conductivity present in the (gradient) material. The heat dissipation
from ‘hot spots’ in vicinity to such gradient materials, consequently, cannot be discussed in
from ‘hot spots’ in vicinity to such gradient materials, consequently, cannot be discussed in
Polymers 2018, 10, 1131 9 of 13

Polymers 2018, 10, x FOR PEER REVIEW 9 of 13

spots’ in vicinity to such gradient materials, consequently, cannot be discussed in sufficient detail with
respect to detail
sufficient (only) with
the ‘averaged’
respect tothermal conductivity.
(only) the ‘averaged’Instead,
thermalthe gradient behavior
conductivity. Instead,ofthe
thegradient
thermal
conductivity
behavior must
of the also conductivity
thermal be considered. Foralso
must a more elaborate discussion
be considered. For a more of the thermal
elaborate dissipation
discussion of the
behavior,dissipation
thermal this behavior was modelled
behavior, and simulated
this behavior in thisand
was modelled study.
simulated in this study.

Figure 8. Comparison
Comparison ofof the guarded heat flow meter (GHFM) (representing a homogenized bulk
material) and the LFA measurements (representing the respective layers of a gradient composite
numbered 2 and 8) of the epoxy resin-based composite.

3.3. Modelling & Simulation of the Thermal Dissipation Behavior

3.3. Modelling & Simulation of the Thermal Dissipation Behavior
For the simulation of a gradual thermal conductivity distribution, a model with an uncoupled
For the simulation of a gradual thermal conductivity distribution, a model with an uncoupled
heat transfer analysis, including conduction, boundary convection, and boundary radiation, was
heat transfer analysis, including conduction, boundary convection, and boundary radiation, was
employed. A multi-layer model composed of copper, silicon, and an epoxy-based composite was
employed. A multi-layer model composed of copper, silicon, and an epoxy-based composite was
employed (Figure 2). For the epoxy material, eight layers were considered, the material parameters of
employed (Figure 2). For the epoxy material, eight layers were considered, the material parameters
which were taken from the measurements described above. A cyclic thermal loading and unloading
of which were taken from the measurements described above. A cyclic thermal loading and
with five loading cycles of 0.1 s each was applied (Figure 4). The heat was applied in a region with a
unloading with five loading cycles of 0.1 s each was applied (Figure 4). The heat was applied in a
radius of 1 mm on top of the silicon layer; the first four loading cycles were interrupted by unloading
region with a radius of 1 mm on top of the silicon layer; the first four loading cycles were interrupted
cycles of 1 s, while the last loading cycle was ended with an unloading cycle of 3 s. The temperature
by unloading cycles of 1 s, while the last loading cycle was ended with an unloading cycle of 3 s.
fields immediately after the last loading cycle (Figure 9 top) and 3 s after the last loading cycle (Figure 9
The temperature fields immediately after the last loading cycle (Figure 9 top) and 3 s after the last
bottom) show a gradual dissipation of the heat through the epoxy-based material. Expectedly, the
loading cycle (Figure 9 bottom) show a gradual dissipation of the heat through the epoxy-based
maximum temperature is reached in the region where the heat was applied and decreases throughout
material. Expectedly, the maximum temperature is reached in the region where the heat was
the epoxy. It can be clearly seen that the heat is transferred to the top of the epoxy structure in gradual
applied and decreases throughout the epoxy. It can be clearly seen that the heat is transferred to the
fashion, which evolves over time.
top of the epoxy structure in gradual fashion, which evolves over time.
For the evaluation of the (time-resolved) heat transfer, a path along the copper/silicon/epoxy
For the evaluation of the (time-resolved) heat transfer, a path along the copper/silicon/epoxy
materials with several examination points was defined, comprising the top of the silicon layer, the top of
materials with several examination points was defined, comprising the top of the silicon layer, the
the copper layer, and the top of each of the eight epoxy layers (Figure 10). For the calculations, three sets
top of the copper layer, and the top of each of the eight epoxy layers (Figure 10). For the calculations,
of material parameters for the epoxy layer were considered, namely (i) “averaged” thermal conductivity
three sets of material parameters for the epoxy layer were considered, namely (i) “averaged”
for a non-gradient composite material (reproducing the GHFM measurements); (ii) gradient thermal
thermal conductivity for a non-gradient composite material (reproducing the GHFM
conductivity according to the LFA measurements of individual layers of a gradient composite material;
measurements); (ii) gradient thermal conductivity according to the LFA measurements of individual
and (iii) high thermal conductivity for a non-gradient composite material (reproducing the LFA
layers of a gradient composite material; and (iii) high thermal conductivity for a non-gradient
measurement of the bottom layer number 8).
composite material (reproducing the LFA measurement of the bottom layer number 8).
From the temperature distribution along this path immediately after the last loading cycle
From the temperature distribution along this path immediately after the last loading cycle
(Figure 11 left), it is perceptible that the temperature on the silicon layer is, as expected, at its maximum
(Figure 11 left), it is perceptible that the temperature on the silicon layer is, as expected, at its
with the “averaged” GHFM-measured data, because of the low thermal conductivity considered for the
maximum with the “averaged” GHFM-measured data, because of the low thermal conductivity
bulk material. Correspondingly, the lowest temperature is obtained on the top-boundary of the epoxy
considered for the bulk material. Correspondingly, the lowest temperature is obtained on the
structure. Considering a non-gradient material with high thermal conductivity, the temperature is
top-boundary of the epoxy structure. Considering a non-gradient material with high thermal
highest on the top-boundary of the epoxy structure and lowest on the silicon layer. The heat convection
conductivity, the temperature is highest on the top-boundary of the epoxy structure and lowest on
of the simulation using the layer-wise measured data of the epoxy-amine-based composite is located
the silicon layer. The heat convection of the simulation using the layer-wise measured data of the
between the results of the simulation with the GHFM-measured data and the results with the idealized
epoxy-amine-based composite is located between the results of the simulation with the
GHFM-measured data and the results with the idealized epoxy data. Notably, 3 s after the thermal
loading (Figure 11 right), these effects become even more pronounced, revealing a temperature
Polymers 2018, 10, 1131 10 of 13

Polymers 2018, 10, x FOR PEER REVIEW 10 of 13

epoxy data. Notably, 3 s after the thermal loading (Figure 11 right), these effects become even more
difference
pronounced, of revealing
Polymers 2018, approximately
10, x FOR PEER 20 K on
REVIEW
a temperature the top-boundary
difference of the20epoxy
of approximately structure,
K on the between 13 the
10 of of
top-boundary the
layer-wise model, according to LFA measurements and the model with ‘averaged’ low
epoxy structure, between the layer-wise model, according to LFA measurements and the model with thermal
difference according
conductivity, of approximately
to GHFM 20measurements.
K on the top-boundary of the epoxy structure, between the
‘averaged’ low thermal conductivity, according to GHFM measurements.
layer-wise model, according to LFA measurements and the model with ‘averaged’ low thermal
conductivity, according to GHFM measurements.

Figure
Figure 9. (Top)
9. (Top) Resulting
Resulting temperaturefield
temperature fieldimmediately
immediately after
afterthe
thelast
lastloading
loading step forfor
step thethe
layer-wise
layer-wise
measured
measured epoxy
epoxy data.
data. (Bottom)
(Bottom) Resulting
Resulting temperature
temperature field
fieldat the
at end
the end of the
of last
the unloading
last unloading
Figure 9. (Top) Resulting temperature field immediately after the last loading step for the layer-wise step for for
step
the
measuredlayer-wise
the layer-wise
epoxy measured
measured epoxy data.
epoxy data.
data. (Bottom) Resulting temperature field at the end of the last unloading step for
the layer-wise measured epoxy data.

Figure
Figure 10. 10. Evaluation
Evaluation pointsonontop
points topofofthe
the silicon
silicon layer,
layer, on
ontop
topofofthe
thecopper layer,
copper andand
layer, on top of each
on top of each
of eight
of the the eight epoxy
epoxy layers.
layers.
Figure 10. Evaluation points on top of the silicon layer, on top of the copper layer, and on top of each
of the eight epoxy layers.
 Polymers 2018, 10, x FOR PEER REVIEW 11 of 13
Polymers 2018, 10, 1131 11 of 13
Polymers 2018, 10, x FOR PEER REVIEW 11 of 13

Figure 11. (Left) Temperature distribution immediately after the last loading cycle throughout the
defined path (Figure 10). (Right) Temperature distribution at the end of the last unloading cycle
Figure11.
Figure 11.(Left)
(Left)Temperature
Temperaturedistribution
distributionimmediately
immediatelyafter
afterthe
thelast
lastloading
loadingcycle
cyclethroughout
throughoutthe
the
throughout the defined path (Figure 10).
defined path (Figure 10). (Right) Temperature distribution at the end of the last unloading
defined path (Figure 10). (Right) Temperature distribution at the end of the last unloading cycle cycle
throughoutthe
throughout thedefined
definedpath
path(Figure
(Figure10).
10).
The comparison of the temperature evolution over time for all layers reveals that the
temperature difference between the three simulation results increases for each layer (Figure 12).
The
Thecomparison
comparison of the
of temperature evolution
the temperature over time
evolution for time
over all layers
for reveals that reveals
all layers the temperature
that the
Whilst at the top of the silicon layer, cyclic temperature lifts are observed (Figure 12 left) and
difference
temperature between the three
difference simulation
between results
the three increases
simulation for each
results layer (Figure
increases for each12). Whilst
layer at the
(Figure 12).
continuous heating occurs at the top of the epoxy structure (Figure 12 right). Especially at the
top of theatsilicon
Whilst the toplayer, cyclic
of the temperature
silicon lifts are
layer, cyclic observed (Figure
temperature lifts are12observed
left) and continuous heating
(Figure 12 left) and
silicone-epoxy interface, an enhancement of the thermal conductivity is noticeable after several
occurs at the top
continuous of theoccurs
heating epoxy atstructure
the top(Figure
of the12 right).structure
epoxy Especially at the silicone-epoxy
(Figure interface,
12 right). Especially at an
the
temperature lifts.
enhancement of the thermal conductivity is noticeable after several temperature
silicone-epoxy interface, an enhancement of the thermal conductivity is noticeable after several lifts.
temperature lifts.

Figure12.
Figure 12. (Left)
(Left) Temperature
Temperature distribution
distribution over
over the
thetotal
totalsimulation
simulationtime,
time,measured
measuredatat
thethe
toptop
of of
the
silicon layer. (Right) Temperature distribution over the total simulation
the silicon layer. (Right) Temperature distribution over the total simulation time measured at
measured at the the
Figure 12. (Left) Temperature distribution over the total simulation time, measured at the top of the
uppermostepoxy
uppermost epoxylayer.
layer.
silicon layer. (Right) Temperature distribution over the total simulation time measured at the
uppermost epoxy layer.
4.4.Summary,
Summary,Discussion
Discussionand andConclusions
Conclusions
An
An epoxy/amine-based
4. Summary, Discussion and gradient
epoxy/amine-based gradient composite
Conclusions composite was was prepared
prepared by by thethe addition
addition ofof nano-nano- and and
submicron-scaled
submicron-scaledalumina particles by
alumina particles byexpanding
expandingthe thecuring
curing time
time of the
of the polymer
polymer matrix
matrix to thetorange
the
range An epoxy/amine-based
of ahours.
few hours. The thermal gradient composite
conductivity was this
prepared by composite
the addition of nano- and
of a few The thermal conductivity withinwithin
this gradient gradient
composite rangedranged
from 0.25fromto0.25
0.45
submicron-scaled
toW·m
0.45−1·K
W−1 − 1 −
·mat ·r.t., 1alumina
K hence at r.t., byparticles
hence by expanding
by a factor the
of almost curing
2. The time
range of the polymer
of thermal matrix
conductivitiesto the range
could
a factor of almost 2. The range of thermal conductivities could be
beof a few hours.
measuredwithin The
withinthis thermal
thisstudy conductivity
studyonlyonlyby within
by performing this
performing laser gradient
laser flash composite
flash analysis
analysis on ranged from
individual 0.25
layers toof0.45
aa
measured on individual layers of
W·m−1·K−1 at
dedicatedly r.t., hence by a factor of conductivity
almost 2. Theof range of thermal conductivities could be
dedicatedly cut specimen. If the thermal conductivity of an uncut specimen was measured withaa
cut specimen. If the thermal an uncut specimen was measured with
measured within meter,
this study only by thermal
performing laser flash analysis on toindividual Wlayers
·m −1 Kof −1 a
guarded
guardedheat heatflow
flow meter, the
the‘averaged’
‘averaged’ thermal conductivity
conductivity waswasmeasured
measured be 0.250.25
to be W·m·−1 ·K−1, ,
dedicatedlyin
reproducing cut specimen. If the thermal conductivity of an uncut specimen was measured with a
reproducing ingood goodapproximation
approximationthe thelowest
lowestvalue
valueofofthermal
thermalconductivity
conductivityaccording
accordingtotolaserlaserflash
flash
guarded
analysis. heat
The flow
fact thatmeter,
these the
two ‘averaged’
types of thermal
measurement conductivity
techniques was measured
revealed to be
pronouncedly 0.25 W·m −1·K−1,
different
analysis. The fact that these two types of measurement techniques revealed pronouncedly different
reproducing
findings in good approximation the lowest value of thermal conductivity according to laser flash
findingswillwillbebesubject
subjectofoffurther
furtherstudies.
studies.Complementary
Complementaryfinite finiteelement
elementanalysis
analysisofofaamulti-layer
multi-layer
analysis. comprising
assembly The fact thatlayersthese of two types silicon,
of measurement techniques revealed pronouncedly different
assembly comprising layers of copper, silicon, and the epoxy/amine/alumina compositerevealed
copper, and the epoxy/amine/alumina composite revealed
findings
that will be subject of further studies. Complementary finite element analysis of a multi-layer
thatthe
theheat
heatdissipation
dissipation through
through a gradient
a gradient composite
compositewas was
of intermediate
of intermediatespeedspeed
in comparison with
in comparison
assembly
materials comprising
exhibiting layers
a non-gradientof copper, silicon, and
thermal conductivity the epoxy/amine/alumina
of 0.25 and 0.45 − 1
·m and composite
− 1
·K ,0.45 revealed
with materials exhibiting a non-gradient thermal conductivity ofW0.25 respectively,
W·m−1·Kif−1,
that
the the heat
layers with dissipation
the highest through conductivities
thermal a gradient composite
were was oftointermediate
adjacent the heat speed
source. Thesein findings
comparison of
respectively, if the layers with the highest thermal conductivities were adjacent to the heat source.
with
the materials
finite exhibiting
element analysis a non-gradient
supported thermaldata
the experimental conductivity
of the laser of 0.25
flash and 0.45 W·m−1·K−1,
analysis.
respectively, if the layers with the highest thermal conductivities were adjacent to the heat source.
Polymers 2018, 10, 1131 12 of 13

It may be concluded from this study that gradient composites are formed (autonomously)
within the range of a few hours if non-nanoscaled fillers are added to a polymer matrix, e.g., for
the enhancement of the thermal conductivity. This phenomenon must be taken into account in
particular if only submicron or micron-scaled fillers are (commercially) available. Correspondingly, the
thermal conductivity varies along (at least) one dimension of such a gradient composite. Such gradient
composites nonetheless enable faster heat dissipation compared to a homogeneous composite material
with a uniform thermal conductivity identical to the lowest one of the gradient composite, despite
eventually identical thermal conductivity according to guarded heat flow Meter measurements.

Author Contributions: M.M. and P.F.F. performed the modelling & simulation calculations, P.M. and M.G.
prepared the specimens and characterized them. P.F.F., M.P., and F.W. planned the study and discussed the results.
All authors jointly compiled the manuscript.
Funding: The research work was performed within the K-Project PolyTherm at the Polymer Competence
Center Leoben GmbH (PCCL, Austria) within the framework of the COMET-program of the Federal Ministry
for Transport, Innovation and Technology and the Federal Ministry for Digital and Economic Affairs with
contributions by the Montanuniversitaet Leoben and the TU Dortmund University. Funding is provided by the
Austrian Government and the State Government of Styria.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Coombs, C.F., Jr. Printed Circuits Handbook, 6th ed.; McGraw-Hill: New York, NY, USA, 2008;
ISBN 978-0071467346.
2. Liu, S.; Liu, Y. Modeling and Simulation for Microelectronic Packaging Assembly: Manufacturing, Reliability, and
Testing, 1st ed.; John Wiley & Sons (Asia) Pte Ltd.: Singapore, 2011; ISBN 978-0-470-82780-2.
3. Tanaka, Y.; Chen, G.; Zhao, Y.; Davies, A.E.; Vaughan, A.S.; Takada, T. Effect of additives on morphology and
space charge accumulation in low density polyethylene. IEEE Trans. Dielectr. Electr. Insul. 2003, 10, 148–154.
[CrossRef]
4. Pleşa, I.; Noţingher, P.V.; Schlögl, S.; Sumereder, C.; Muhr, M. Properties of Polymer Composites Used in
High-Voltage Applications. Polymers 2016, 8, 173. [CrossRef]
5. Bernardoni, M.; Cova, P.; Delmonte, N.; Menozzi, R. Heat management for power converters in sealed
enclosures. A numerical study. Microelectron. Reliab. 2009, 49, 1293–1298. [CrossRef]
6. Sokolov, A.; Liu, C.; Mohn, F. Reliability assessment of SiC power module stack based on thermo-structural
analysis. In Proceedings of the 19th International Conference on Thermal, Mechanical and Multi-Physics
Simulation and Experiments in Microelectronics and Microsystems (EuroSimE), Toulouse, France, 15–18
April 2018.
7. Benda, V. Progress in power semiconductor devices. Microelectron. Reliab. 2012, 52, 461–462. [CrossRef]
8. Yang, D.; Kengen, M.; Peels, W.; Heyes, D.; van Driel, W.D. Reliability modeling on a MOSFET power
package based on embedded die technology. Microelectron. Reliab. 2010, 50, 923–927. [CrossRef]
9. Pfost, M.; Boianceanu, C.; Lohmeyer, H.; Stecher, M. Electrothermal Simulation of Self-Heating in DMOS
Transistors up to Thermal Runaway. IEEE Trans. Electron. Devices 2013, 60, 699–707. [CrossRef]
10. Boettcher, L.; Karaszkiewicz, S.; Manessis, D.; Ostmann, A. Embedding of Power Semiconductors for
Innovative Packages and Modules. In Proceedings of the SMTA International 2014 Proceedings, Rosemont,
IL, USA, 28 September–2 October 2014; pp. 222–228, ISBN 978-1-63439-399-7.
11. Bolton, W.; Higgins, R.A. Materials for Engineers and Technicians, 5th ed.; Elsevier: Oxford, UK, 2010; ISBN
978-1856177696.
12. Eibel, A.; Marx, P.; Jin, H.; Tsekmes, I.-A.; Mühlbacher, I.; Smit, J.J.; Kern, W.; Wiesbrock, F. Enhancement
of the Insulation Properties of Poly(2-oxazoline)-co-Polyester Networks by the Addition of Nanofillers.
Macromol. Rapid Commun. 2018, 39. [CrossRef] [PubMed]
13. Reading, M.; Vaughan, A.S.; Lewin, P.L. An investigation into improving the breakdown strength and
thermal conduction of an epoxy system using boron nitride. In Proceedings of the 2011 Annual Report
Conference on Electrical Insulation and Dielectric Phenomena, Cancun, Mexico, 16–19 October 2011.
 Polymers 2018, 10, 1131 13 of 13

14. Marx, P.; Wanner, A.J.; Zhang, Z.; Jin, H.; Tsekmes, I.A.; Smit, J.J.; Kern, W.; Wiesbrock, F. Effect of Interfacial
Polarization and Water Absorption on the Dielectric Properties of Epoxy-Nanocomposites. Polymers 2017, 9,
195. [CrossRef]
15. Kidalov, S.V.; Shakhov, F.M. Thermal Conductivity of Diamond Composites. Materials 2009, 2, 2467–2495.
[CrossRef]
16. Weng, Q.H.; Wang, X.B.; Wang, X.; Bando, Y.; Golberg, D. Functionalized hexagonal boron nitride
nanomaterials: Emerging properties and applications. Chem. Soc. Rev. 2016, 45, 3989–4012. [CrossRef]
[PubMed]
17. Franco Júnior, A.; Shanafield, D.J. Thermal conductivity of polycrystalline aluminum nitride (AlN) ceramics.
Cerâmica 2004, 50, 247–253. [CrossRef]
18. Lide, D.R. CRC Handbook of Chemistry and Physics, 84th ed.; CRC Press: Boca Raton, FL, USA, 2003;
ISBN 978-0-8493048-4-2.
19. Dahmen, W.; Reusken, A. Numerik für Ingenieure und Naturwissenschaftler, 1st ed.; Springer: Berlin, Germany,
2008; ISBN 978-3540764922.
20. Eck, C.; Garcke, H.; Knabner, P. Mathematische Modellierung, 3rd ed.; Springer: Berlin, Germany, 2017;
ISBN 978-3662543344.
21. Dassault Systèmes: Simulia User Assistance 2017; Abaqus Documentation: Providence, RI, USA, 2017.
22. Durand, C.; Klinger, M.; Coutellier, D.; Naceur, H. Power Cycling Reliability of Power Module: A Survey.
IEEE Trans. Device Mater. Reliab. 2016, 16, 80–97. [CrossRef]
23. Chiozzi, D.; Bernardoni, M.; Delmonte, N.; Cova, P. A simple 1-D finite elements approach to model the
effect of PCB in electronic assemblies. Microelectron. Reliab. 2016, 58, 126–132. [CrossRef]
24. Lutz, J.; Schlangenotto, H.; Scheuermann, U.; De Doncker, R. Semiconductor Power Devices: Physics,
Characteristics, Reliability, 1st ed.; Springer: Berlin, Germany, 2011; ISBN 978-3-642-11124-2.
25. Parker, W.J.; Jenkins, R.J.; Butler, C.P. Flash Method of Determining Thermal Diffusivity, Heat Capacity, and
Thermal Conductivity. J. Appl. Phys. 1961, 32, 1679–1684. [CrossRef]
26. Schulz, B.; Hoffmann, M. Thermophysical properties of the system Al2 O3 -MgO. High Temp. High Press. 2002,
34, 203–212. [CrossRef]
27. Swier, S.; van Assche, G.; van Mele, B. Reaction Kinetics Modeling and Thermal Properties of Epoxy–Amines
as Measured by Modulated-Temperature DSC. I. Linear Step-Growth Polymerization of DGEBA and Aniline.
J. Appl. Polym. Sci. 2004, 91, 2798–2813. [CrossRef]
28. ISO 8301:1991. Thermal Insulation-Determination of Steady-State Thermal Resistance and Related Properties-Heat
Flow Meter Apparatus; International Organization for Standardization: Geneva, Switzerland, 1991.
29. ASTM E1530-11. Standard Test Method for Evaluating the Resistance to Thermal Transmission of Materials by the
Guarded Heat Flow Meter Technique; ASTM International: West Conshohocken, PA, USA, 2011.
30. Yoon, S.; Macphee, D.E.; Imbabi, M.S. Estimation of the thermal properties of hardened cement paste on the
basis of guarded heat flow meter measurements. Thermochim. Acta 2014, 588, 1–10. [CrossRef]

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).

View publication stats