1 s2.0 S0079670016000216 Main

Progress in Polymer Science 59 (2016) 41–85
Contents lists available at ScienceDirect
Progress in Polymer Science

journal homepage: www.elsevier.com/locate/ppolysci
Thermal conductivity of polymer-based composites:

Fundamentals and applications
Hongyu Chen a,∗ , Valeriy V. Ginzburg b , Jian Yang c , Yunfeng Yang a , Wei Liu a ,
Yan Huang a , Libo Du a , Bin Chen a
a
Dow Chemical (China) Investment Co. Ltd., Shanghai 201203, People’s Republic of China
b
The Dow Chemical Company, Midland, MI 48674, USA
c
The Dow Chemical Company, Freeport, TX 77541, USA
a r t i c l e i n f o a b s t r a c t
Article history: Thermal management is critical to the performance, lifetime, and reliability of electronic
Available online 26 March 2016 devices. With the miniaturization, integration and functionalization of electronics and
the emergence of new applications such as light emitting diodes, thermal dissipation
Keywords: becomes a challenging problem. Addressing this challenge requires the development of
Thermally conductive polymer composites novel polymer-based composite materials with enhanced thermal conductivity. In this
Modeling review, the fundamental design principles of highly thermally conductive composites were
Chain orientation
discussed. The key factors influencing the thermal conductivity of polymers, such as chain
Liquid crystalline polymers
structure, crystallinity, crystal form, orientation of polymer chains, and orientation of
Thermally conductive fillers
Microstructure control ordered domains in both thermoplastics and thermosets were addressed. The properties of
thermally conductive fillers (carbon nanotubes, metal particles, and ceramic particles such
as boron nitride or aluminum oxide) are summarized at length. The dependence of ther-
mal conductivity of composites on the filler loading, filler aggregate morphology and overall
composite structure is also discussed. Special attention is paid to recent advances in control-
ling the microstructure of polymer composites to achieve high thermal conductivity (novel
approaches to control filler orientation, special design of filler agglomerates, formation
of continuous filler network by self-assembly process, double percolation approach, etc.).
The review also summarizes some emerging applications of thermally conductive poly-
mer composites. Finally, we outline the challenges and outlook for thermally conductive
polymer composites.
© 2016 Elsevier Ltd. All rights reserved.
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2. Modeling of thermal conductivity in composites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.1. Constitutive (micromechanical) modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.2. Finite element modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.3. Interfacial thermal resistance modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
∗ Corresponding author. Tel.: +86 21 38513128; fax: +86 21 58954173.

E-mail address: [email protected] (H. Chen).
http://dx.doi.org/10.1016/j.progpolymsci.2016.03.001
0079-6700/© 2016 Elsevier Ltd. All rights reserved.
42 H. Chen et al. / Progress in Polymer Science 59 (2016) 41–85
3. Thermal conductivity measurement techniques. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .51

3.1. Steady-state methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.1.1. Guarded hot plate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.1.2. Axial flow method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.1.3. Heat flow meter method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.1.4. Pipe method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.2. Transient methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.2.1. Flash method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.2.2. Transient hot wire method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.2.3. Transient plane source method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4. Thermal conductivity of thermoplastics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.1. Effect of chain structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.2. Effect of crystallinity and crystal form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.3. Effect of orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.3.1. Orientation of crystalline/semicrystalline polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.3.2. Orientation of amorphous polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.3.3. Orientation of liquid crystalline polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.4. Polymer blends by engineered interchain interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5. Thermal conductivity of thermosets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.1. Effect of liquid crystal domain size and content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.2. Effect of curing condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.3. Effect of orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6. Thermally conductive fillers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.1. Metallic fillers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.2. Carbon based materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.3. Ceramic fillers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
7. Influence of the filler morphology and properties on thermal conductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
7.1. Effect of filler loading level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
7.2. Effect of filler shape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
7.2.1. One-dimensional fillers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
7.2.2. Platelet-like fillers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
7.2.3. Spherical fillers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
7.3. Size effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
7.4. Hybrid fillers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
7.4.1. Mixture of fillers with different sizes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
7.4.2. Mixture of fillers with different shapes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
7.4.3. Mixture of different types of fillers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
7.5. Surface treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
8. Microstructure control of heat conductive polymer composites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
8.1. Filler orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
8.1.1. Orientation by shear or extensional flow during processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
8.1.2. Orientation by external fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
8.1.3. Orientation of carbon nanotube by chemical vapor deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
8.2. Filler agglomerates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
8.3. Formation of continuous filler network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
8.3.1. Molding of the mixture of filler and polymer powders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
8.3.2. In situ polymerization in porous heat conductive filler network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
8.3.3. Self-assembly process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
8.3.4. Double percolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
9. Applications of thermally conductive polymer composites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
9.1. LED devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
9.2. Electronic packaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
9.2.1. Single chip packages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
9.2.2. 3D chip stack assembly and packaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
9.2.3. Automotive electronic control units (ECUs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
9.3. Batteries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
9.4. Solar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
10. Conclusions and future prospects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
H. Chen et al. / Progress in Polymer Science 59 (2016) 41–85 43
Symbols
Cu copper
A cross sectional transfer area CuNWs copper nanowires
Cp specific heat capacity per unit volume DC direct-current
Dint interparticle distance DDE 4,4 -diaminodiphenylethane
k or TC thermal conductivity DDM diffuse mismatch model
kB bulk thermal conductivity DGEBA diglycidyl ether of bisphenol A
kc thermal conductivity of the composite DGEBP diglycidyl ether of 4,4 -biphenol
keff effective heat conductivity DGEDH diglycidyl ether of 1,4-diaminobenzene-bis
kf thermal conductivity of filler (4-hydroxy-benzaldehyde)
kp thermal conductivity of polymer matrix DGEHAP diglycidyl ether of 4-((4-
ij thermal conductivity tensor hydroxybenzylidene)amino)phenol
h interfacial thickness DGEHH diglycidyl ether of hydrazine-bis(4-
J heat flux hydroxybenzaldehyde)
L conduction path length DGETAM diglycidyl ether of terephthalylidene-bis-
l phonon mean free path (4-amino-3-methylphenol)
L11 , k33 geometrical factors DGETMBP diglycidyl ether of 3,3 ,5,5 -tetramethyl-
Mn number average molecular weight 4,4 -biphenol
q rate of heat transfer DSC differential scattering calorimetry
r filler aspect ratio ECU electronic control units
Rksm interfacial thermal resistance between EMA effective medium approximations
SWCNT and matrix EMC epoxy molding compound
T0 constant temperature EMT effective medium theory
T temperature difference EN European Standard
TC|| TC along the orientation direction EVA poly(ethylene vinyl acetate)
TC⊥ TC perpendicular to the draw direction EVs electric vehicles
˛ thermal diffusivity FD finite difference
ı kf /kp Fe3 O4 ferroferric oxide
volume fraction of fillers FEA finite element analysis
draw ratio FeCr Fe–Cr steel particles
interfacial thermal conductivity GBN glycidyl methacrylate-butyl acrylate-
phonon velocity maleic anhydride
density GNSs graphene nanosheets
electrical conductivity H3 BO3 boric acid
hBN hexagonal boron nitride
Abbreviations HDPE high density polyethylene
1-D one-dimensional HEVs hybrid electric vehicles
2-D two dimensional ISO International Organization for Standardiza-
3-D three-dimensional tion
a-PT amorphous polythiophene ITRS International Technology Roadmap for
ABS poly(acrylonitrile-butadiene-styrene) Semiconductors
copolymer LABNs linear assembles of BN nanosheets
AgNWs silver nanowires LADFs linearly aligned diamond fillers
AgNWs@SiO2 silica-coated silver nanowires LCP liquid crystalline polymers
Al aluminum LDPE low density polyethylene
Al2 O3 aluminum oxide LED light-emitting diode
AlN aluminum nitride MG model Maxwell–Garnett approximation
ASTM American Society for Testing and Materials MPP mesophase pitch
BACE bisphenol A cyanate ester MWCNTs multi-walled carbon nanotubes
BeO beryllium oxide Ni nickel
BN boron nitride PA 6 polyamide-6
BNNT boron nitride nanotubes PA66 polyamide-6,6
BNNTs boron nitride nanotubes PAA poly(acrylic acid)
BZE 4-(4-oxylanylbutoxy)benzoic acid-1,4 - PAN polyacrylonitrile
phenylene ester PAP poly(N-acryloyl piperidine)
c-PBO crystalline polybenzobisoxazole PBO poly(p-phenylene benzobisoxazole)
c-PE crystalline polyethylene PBT poly(butylene terephthalate)
CNF carbon nano fibers PC polycarbonate
CNT carbon nanotube PCM phase change materials
CTE coefficient of thermal expansion PDMS poly(dimethyl siloxane)
1. Introduction
PE polyethylene Thermal dissipation is critical to the performance,

PEEK polyetheretherketone lifetime, and reliability of electronic devices. With the
PEI polyetherimide miniaturization, integration and functionalization of elec-
PET poly(ethylene terephthalate) tronics and the emergence of new applications such
PI polyimide as three-dimensional chip stack architectures, flexible
PIB polyisobutylene electronics, and light emitting diodes, thermal dissipa-
PIPD polyhydroquinone diimidazopyridine tion becomes a challenging problem. Thus, new highly
PMDA pyromellitic dianhydride thermally-conductive materials are needed to help address
PMMA polymethylmethacrylate this challenge [1,2].
POM polyoxymethylene For one-dimensional, steady state heat flow, the heat
POSS polyhedral oligosilsesquioxane conduction rate is expressed by Fourier’s equations (1) and
PP polypropylene (2) [2]:
PPS polyphenylene sulfide
PPSU polyphenylsulfone T
q = kA (1)
PS polystyrene L
PSU polysulfone or
PT polythiophene q dT
J= = −k (2)
PTFE polytetrafluoroethylene A dx
PU polyurethane
where q is rate of heat transfer (W); J is heat flux (W/m2 ), k is
PV photovoltaic
thermal conductivity W/(m K); A is cross sectional transfer
PVC polyvinyl chloride
area (m2 ); T is temperature difference (◦ C); L is conduc-
PVDC polyvinylidene chloride
tion path length (m).
PVDF polyvinylidene difluoride
In solids, heat may be transported by charge carri-
PVF polyvinyl formal
ers (such as electrons and holes) or by phonons (energy
PVP polyvinylpyrrolidone
quanta of atomic lattice vibrations). For insulators and
PVPh poly(4-vinyl phenol)
semiconductors the thermal conductivity is dominated
PWD pine wood dust
by the contribution from phonons, while in metals the
RVE representative volume elements
electronic contribution greatly outweighs the phonon con-
SAN styrene–acrylonitrile copolymer
tribution. For most polymers, the primary mechanism of
SB styrene/polybutadiene copolymer
heat conduction is through phonons [3].
SEBS hydrogenated styrene/butadiene copoly-
The thermal conductivity (TC) of polymers (usually rep-
mer
resented by Greek letter or Latin lowercase k), can be
Si3 N4 silicon nitride
obtained from the Debye equation, i.e.
SiC silicon carbide
SiO2 silica Cp l
k= (3)
SiO2 @CNTs SiO2 coated carbon nanotubes 3
SiO2 @MWCNTs SiO2 coated multiple wall carbon where Cp is the specific heat capacity per unit volume,
nanotubes is the phonon velocity, and l is the phonon mean free
sPS syndiotactic poly(styrene) path. For most polymers, l is extremely small due to scat-
SWCNT single-wall carbon nanotube tering with other phonons, defects, and grain boundaries.
TiB2 titanium boride Therefore, most polymers have a low TC in the range of
TIMs thermal interface materials 0.1–0.5 W/(m K), which is not sufficient for many appli-
TME4 4-(oxylanylmethoxy)benzoic acid-4,4 - cations that require high heat conduction [4]. However,
[1,4-butane-diylbis(oxy)]bisphenol ester polymers have the advantages of good processability, light
TME6 4-(oxylanylmethoxy)benzoic acid-4,4 - weight, low water absorption, high electrical resistivity,
[1,6-hexane-diylbis(oxy)]bisphenol high voltage breakdown strength, corrosion resistance,
ester and most importantly, low cost. Polymer-based thermally
TME8 4-(oxylanylmethoxy)benzoic acid-4,4 - conductive materials are therefore often desired in many
[1,8-octane-diylbis(oxy)]bisphenol ester applications.
TPEE thermoplastic polyester elastomer Typically, heat conductive fillers, such as aluminum
TPU urethane base TPE oxide, boron nitride, aluminum nitride, silicon nitride,
TSV through silicon via diamond, graphite, metal particles, carbon nanotube,
UHMWPE ultrahigh molecular weight polyethylene graphene, etc., are introduced into polymers to increase TC.
VA-MWCNT vertically aligned multiwalled carbon For applications that require both high TC and electrically
nanotube insulating properties, electrically insulating fillers, such as
VDOS vibrational density of states aluminum oxide, boron nitride, aluminum nitride, etc. can
VDP vertical double percolation be used, while graphite, metal particles, carbon nanotube,
WAXD wide angle X-ray diffraction and graphene are often utilized in applications where elec-
ZnO-NPs zinc oxide nano particles trical insulation is not required.
The thermal conductivity of polymer composites is a interaction of polymers and fillers also need to be care-
function of the TC of both polymers and fillers. However, fully understood and controlled. Furthermore, not only the
the importance of the TC of polymers is often neglected TC, but also other properties and processability need to be
in the design of thermally conductive polymer compos- considered and balanced.
ites because it is usually quite low compared to that of In this review, key factors that influence the TC of
the filler. The TC of polymers is particularly important at composites are discussed. Special attention is given to ther-
low filler loadings, as the thermally conductive fillers are mally conductive nanocomposites. Despite the fact that
separated by polymer matrix, which acts as a thermal bar- there has been an excellent review on the TC of car-
rier and becomes rate-limiting in the thermal conduction bon nanotubes and their polymer composites [3], they
pathway. are still included in this review, but in less detail. Novel
The addition of thermally conductive inorganic fillers is techniques for microstructure control of heat conductive
indispensable for achieving high TC. The filler type, loading composites are reviewed. Various applications of thermally
level, filler size, and filler shape have a strong influence conductive polymer composites are summarized. Finally,
on the TC of polymer composites. Moreover, the spatial the challenges and outlook for thermally conductive poly-
arrangement of the fillers and the orientation of fillers are mer composites are discussed.
also important to the TC. Many types of filler have non-
spherical shapes (plate-like or needle-like). They can be 2. Modeling of thermal conductivity in composites
oriented during processing, which leads to composites with
anisotropic thermal conductivity (e.g., if plate-like fillers Due to the almost infinite number of possible mate-
are strongly aligned, the “in-plane” TC would be larger than rial compositions (the choice of polymer matrix, the type
the normal or “through-plane” TC). In some applications, an of thermally conductive fillers, the filler weight fraction,
anisotropic TC is not preferred or the orientation direction etc.), it is imperative to find ways to limit the number
of the fillers is not in the desired direction. Novel techniques of possible experiments and to be able to estimate the
are being developed for obtaining desired filler orientation. properties of the candidate materials. Thus, theory and
Often creating a continuous filler network is the key to modeling play important role in the process of screening
obtaining high TC in composite structures. Network for- potential compositions and selecting the more promising
mation, however, usually takes place at high filler loading ones. First, modeling (such as molecular dynamic simu-
levels. This can lead to poor processability, poor mechani- lations) can predict “ideal” thermal conductivity of many
cal properties, and high cost. Studies are ongoing to develop “pure” materials (carbon nanotube, graphite, ceramic par-
composites with high TC at a lower filler loading level by ticles, and various polymers). Over the course of this
carefully controlling filler spatial arrangement. review, we will often reference such studies. Second, using
The interaction of filler with polymer matrix is also models, one can determine – often in a qualitative fash-
important to the TC. Poor interface between filler and ion, sometimes semi-quantitatively or quantitatively – the
polymer could lead to high thermal interfacial resistance dependence of composite thermal conductivity on the
and reduce the TC. Improving polymer-filler interfacial matrix TC, filler TC, and filler volume fraction (“loading”).
interaction can increase the overall composite TC quite sub- Other factors that influence the overall composite TC – and
stantially [4]. are often neglected in the models – include the filler mor-
The TC is determined by the structure and properties of phology (dispersed individual particles; aggregates; fractal
polymer and fillers, as well as the morphology of the com- clusters; percolated network, etc.), anisotropy (especially
posites and the interaction of polymers and fillers, as shown in the case of non-spherical filler particles), properties
in Fig. 1. When we design thermally conductive polymer of filler–matrix interfaces, and effects of processing his-
composites, a holistic approach to satisfy all the applica- tory. Sometimes, these effects are difficult to predict from
tion requirements is needed. It is not sufficient just to select first principles, and then modeling should be combined
appropriate polymers and fillers. The morphology and the with experimental inputs (such as electron microscopy) to
provide explanation of observed trends and design subse-
quent experiments.
Interaction Broadly, composite models can be divided into two
(filler surface treatment)
classes – analytical micromechanical models and finite
element simulations. Analytical micromechanical models
(also known as constitutive equations) provide closed-
form expressions for the composite TC as functions of the
TCs of the matrix and filler materials. Usually, these models
Thermal assume ideal filler-matrix interfaces (no thermal resistance
Conductivity for phonons to go through) and idealized morphology (e.g.,
uniform dispersion of filler particles). As a result, those
Polymer Type Morphology
models can provide very quick estimates of the depend-
Filler (orientation of fillers and ence of composite TC on both filler and matrix TCs and the
(size, shape, type, polymers, ordered structure, filler loading; however, the predictions are usually qualita-
hybrid filler) filler arrangement/packing)
tive. Finite element models usually offer more opportunity
Fig. 1. Thermal conductivity of composites depends on multiple factors to calculate composite TC fairly accurately by taking
that need to be considered when designing new materials. into account realistic composite morphology and possible
interfacial resistance; however, they are often time-

consuming and – until recently – have been used mainly to
refine and test predictions of constitutive models. Below,
we review both approaches in more detail.
2.1. Constitutive (micromechanical) modeling
The simplest (and crudest) approach – that often works

as the first approximation – is to consider two limiting
cases: linear mixing rule (also known as the parallel model)
and inverse mixing rule (often referred to as the series
model). It can be rigorously shown that all linear proper-
ties (TC, electrical conductivity, tensile modulus, etc.) are
bracketed by these two values, with the parallel model
providing the upper bound, and series model providing
the lower bound. Formally, the parallel model is based
on the assumption that the temperature gradient is uni- Fig. 2. Effective thermal conductivity of composites as a function of filler
loading, calculated using MG model. Different curves correspond to dif-
form and the heat flux is the weighted sum of heat fluxes ferent filler thermal conductivities. [8], Copyright 2010. Reproduced with
through the matrix and filler domains. The series model permission from John Wiley & Sons.
is based on the assumption that the heat flux is uni-
form, and the temperature gradient is the weighted sum Maxwell–Garnett [6] derived the expression for the
of temperature gradients through the matrix and filler conductivity k for suspension of spherical non-interacting
domains. particles, as defined in Eq. (6):
In the parallel model, filler and matrix contribute
independently to the overall thermal conductivity, in pro- 3(ı − 1)
kc = kp 1+ (6)
portion to their relative volume fractions (see Eq. (4)). 2 + ı − (ı − 1)
kc = (1 − )kp + kf (4) where ı = kf /kp . Dependence of the composite TC on the

filler loading, as computed on the basis of the MG model, is
shown in Fig. 2. The ratio of filler TC to matrix TC is varied
Here, kc , kp and kf are the TC of the composite, polymer
from 10 to 300. It can be seen that at low filler loadings,
matrix and filler respectively, and is the volume fraction
the dependence of composite TC on filler content is nearly
of fillers. The parallel model maximizes the contribution
linear.
of the filler phase and assumes perfect contact between
Later, Fricke [9] generalized Maxwell–Garnett’s model
particles in a fully percolating network implicitly. This
for the case of the spheroidal inclusions. Again, the suspen-
model has some relevance to the case of continuous fiber
sion has to be diluted enough so that suspended particles
composites in the direction parallel to fibers, but gener-
are considered to be thermally isolated one from another.
ally results in very large overestimation for other types of
We can re-write Fricke’s equation to make it consistent
composites.
with Maxwell-Garnett’s formula (see Eq. (7)).
The series model is most applicable for composites
where the fillers are dispersed in a matrix and there is no

n(ı − 1)
percolation even at high filler volume fractions. Within this kc = kp 1+ (7)
n − 1 + ı − (ı − 1)
model, the composite thermal conductivity is given by
Here
−1
(1 − ) ˇ(ı − 1)
kc = + (5) n= (8)
kp kf (ı − 1) − ˇ
1
2 1

Most of the experimental results were found to fall ˇ= (ı − 1) + (9)
somewhere between these two types of models. However, 3 1 + L11 (ı − 1) 1 + L33 (ı − 1)
the series model is usually closer to the experimental data Constants L11 and L33 are geometrical factors depend-
as compared to the parallel model [5], leading to the devel- ent on the filler aspect ratio r (which is defined as the size
opment of many different models derived from the basic divided by thickness, so it is always larger than unity).
series model. These models generally introduce some more
complex weighted averages on thermal conductivities and r2 r
L11 = − cosh−1 r
volume fractions of particles and matrix. The models often − 1)
2(r 2 2(r 2 − 1)
3/2 (10)
include semi-theoretical fitting parameters and rely on L33 = 1 − L11
so-called effective medium approximations. Two com-
monly used effective medium approximation methods are For a sphere, r = 1 and the geometrical factors are all
the Maxwell–Garnett (MG) model [6] and the Bruggeman equal to 1/3. In this case, the shape factor n becomes equal
model [7]. to 3, which recovers Maxwell–Garnett’s formula.
Fig. 3. Bruggeman model prediction of composite TC as a function of par-

ticle loading (volume fraction) with fixed filler TC and various matrix TC
(kp ) ranging from 0.2 to 1.0 W/(m K).
Fig. 4. Schematic of well-dispersed aggregates in composite materials. A

Both original and modified Maxwell–Garnett models thermally conductive path is identified with a solid line.
assume that temperature experiences no discontinuity
at the filler surface. This assumption works well on a TC of particle clusters, and the MG model is used otherwise
macroscopic scale, but breaks down on much smaller char- to approximate the effective TC of composites [13]. The
acteristic length scales where the thermal energy carriers schematic representation of this combination approach is
(electrons or phonons) scatter at the interface. Kapitza [10] shown in Fig. 4.
suggested that temperature at the interface undergoes a Apart from the above models, other theoretical
change proportional to the normal component of the heat approaches [14] include equations by Hashin and Shtrik-
flux through the interface. Nan [11] further generalized the man [15,16], Hamilton and Crosser [17,18], Hatta and Taya
equation for TC of composite material to include the impact [19], Agari [20], Cheng and Vachon [21] as well as by Lewis
of the interfacial resistance in the form of an effective- and Nielsen [22–24]. These models also appear to reason-
medium theory. ably fit most of the experimental data for composites based
Despite many successful cases where the above on isotropic particles as well as short fibers and flakes with
approach is applied, the basic assumption of separated limited aspect ratio, up to loadings of about 30% in volume.
particles in the effective medium approach is not valid in Analytical models are often used as the first-pass esti-
principle for highly filled composites, where contacts are mate for the dependence of TC on the filler volume fraction,
likely to occur, possibly leading to thermally conductive as well as filler and matrix TC. These models have two types
paths [12]. For those types of composites, the so-called of uncertainties. First, morphology in real applications is
Bruggeman model [7] provides a much better descrip- different from the special, idealized cases used in model
tion. In this model, mean field approach is used to capture as assumptions (monodisperse spherical fillers; perfectly
the interactions among the randomly distributed fillers. aligned or perfectly isotropic monodisperse cylinders, etc.).
There is no limitation on the concentration of inclusions for Even in these cases, the models are usually only semi-
Bruggeman model. As a result, it is widely used for particle quantitative and often are unable to capture effects such
percolation in suspensions, i.e. particle clustering effect. For as interfacial thermal resistance (discussed in more details
a binary composite containing matrix and spherical fillers, in Section 2.3). For these and other reasons, more accurate
the Bruggeman model yields the following implicit Eq. (11) analysis often requires the use of numerical approaches
for kc , discussed next.

kf − kc kp − kc
+ (1 − ) =0 (11) 2.2. Finite element modeling
kf − 2kc kp + 2kc
In the illustration (Fig. 3), we demonstrate how the com- In the previous section, we described various analytical
posite TC depends on the matrix TC and the filler loading or closed-form expressions evaluating TC of composites as
as predicted by the Bruggeman model. Assuming a high- a function of their composition and the characteristics of
TC inorganic filler such as boron nitride (kf ∼ 300 W/(m K)), the matrix and the fillers. While those approaches are very
we can clearly see that composite TC depends substantially useful for quick evaluation of the trends expected upon
on the matrix TC, especially at low and intermediate filler changes in filler loading or filler type, they are limited to
volume fractions. And at high loadings (above 50%), the some very specific situations (e.g., monodisperse spheri-
composite TC approaches the filler TC. This dependence will cal fillers in a homogeneous matrix or perfectly aligned
be discussed more extensively in subsequent sections. ellipsoidal fillers in a homogeneous matrix). In general,
For composites with low filler loading, Bruggeman and however, composite morphology could be more compli-
Maxwell–Garnett models give very similar predictions. cated, in which case numerical approaches, such as finite
For composites with higher filler loading and in those element analysis (FEA) or finite difference (FD) modeling,
systems where thermally conductive fillers form perco- could be better suited. Here, we provide several examples
lated pathways, the MG model significantly under-predicts of the FEA modeling of TC in polymer-based composites.
the composite thermal conductivity, while the Bruggeman Comprehensive description of the FEA approach to com-
model can describe experimental data reasonably well. posite TC is certainly beyond the scope of this paper –
Hence, Bruggeman model is used to predict the effective we refer the reader to several recent books and reviews
that provide a general description of the method and its

application to TC [25–29]. In general, within FEA-based
approaches, one considers the morphology of a composite
material and attempts to numerically solve the stationary
equation (Eq. (12)) of heat transfer in solids,
∇ ((r)∇ T (r)) = 0 (12)
subject to the boundary condition,

∂T
= −J (13)
∂n
where ∂T/∂n denotes the normal derivative at the bound-
ary. If we assume, for example, that heat flux J is applied at
the X = 0 boundary, while the boundary X = X0 is maintained
at constant temperature T0 , and the heat flow through the
XY and XZ boundaries can be neglected, then effective heat
conductivity of the sample can be computed as
JX0
eff = (14)
T (x = 0) − T0
where the brackets indicate averaging over y and z in
the x = 0 plane. If the composite is anisotropic, similar cal-
culations could be done in other directions to compute the
principal components of the symmetric tensor kij (the sym-
metry of the tensor follows from the Onsager reciprocal
relations [30,31]). In most cases, polymer-based compos-
ites turned out to be either isotropic (so that only one
TC number is needed) or having a preferential filler plane
orientation (so that two TC numbers, in-plane and through-
plane, are required). While there certainly are many cases
with more complex anisotropy, these will not be discussed
here.
Fig. 5. Composite thermal conductivity in HDPE/Al binary mixtures as
One of the early numerical studies of TC in function of Al filler volume fraction. Experimental data are compared with
polymer–inorganic composites was performed by Kum- predictions using various micromechanical models, as well as FEA simu-
lutas and co-workers [32]. They investigated composites lations. [32], Copyright 2007. Reproduced with permission from Elsevier
of high-density polyethylene (HDPE) and spherical alu- Ltd.
minum metal particles. The particles were dispersed
uniformly and homogeneously in the matrix, and there Bruggeman model which is expected to work better at
was no anisotropy or directionality. The authors took high filler loadings.
electron microscopic images of composites correspond- Annapragada and co-workers [35] studied TC of partic-
ing to various filler loadings, and used finite difference ulate composites with polydisperse spherical fillers. They
modeling in two dimensions to compute TC, assuming developed packing algorithms to generate representative
that k(LDPE) = 0.543 W/(m K) and k(Al) = 204 W/(m K). volume elements (RVE), both in two and three dimen-
Numerical results were then compared to experimental sions. The predicted TC was shown to be fairly close to the
data, as well as to a number of theoretical models (Fig. 5). experimental value, and the approach certainly could be
It was observed that the numerical predictions (squares) generalized to other composites with spherical fillers. Liang
and the experimental data (triangles) were in a fairly good and Li [36,37] studied TC of polypropylene filled with hol-
agreement at high filler loadings (>15 vol.%); at lower filler low glass latex particles; in this case, TC of the composite
loadings, numerical predictions were substantially higher is lower than that of the pure matrix. As shown in Fig. 6,
than the measured values. There can be a number of expla- the predicted TC agrees with the measured data reasonably
nations for this discrepancy. For example, the model did well, both when two-dimensional and three-dimensional
not consider the role of interfacial resistance [33,34]; also, FEA models were used. Again, the role of thermal resistance
the digitization of the images could have led to incorrect at particle interface was not considered.
assignment to some pixels in the regions crowded with Nayak et al. [38] investigated TC of epoxy/pine wood
filler particles. On the other hand, it was shown that con- dust (PWD) composites; again, the filler TC (k(PWD) =
stitutive models, such as Cheng and Vachon [21], predict 0.068 W/(m K)) was less than that of the matrix
TC fairly well at filler loading of less than 10 vol.%. This (k(epoxy) = 0.363 W/(m K)), so the overall TC should
is consistent with the notion that the effective medium decrease as the filler loading is increased. The authors used
theories lose accuracy at higher filler volume fractions as three-dimensional (3D) sphere-in-cube lattice model,
the fillers tend to form large aggregates. Interestingly, the varying the effective sphere radius to change the volume
authors did not attempt to compare their results with the fraction. The problem was solved using ANSYS software.
Fig. 6. Comparison between experimental data (line) and FEA simulations

(2D – filled squares; 3D – open circles) for the polypropylene/hollow glass
particle composites. [37], Copyright 2007. Reproduced with permission Fig. 8. FEA simulations of relative thermal conductivity of polymer/CNT
from Elsevier Ltd. composites as function of CNT volume fraction. Different curves corre-
spond to different interfacial thermal conductivities. [34], Copyright 2007.
Reproduced with permission from Elsevier Ltd.
conduction, making polymer/CNT nanocomposites reason-

ably good conductors, even if not on the same scale as
metals. However, TC increases in polymer/CNT nanocom-
posites are not as significant as the electrical conductivity
increases, despite the similarity between the two trans-
port properties. Moniruzzaman and Winey [39] reviewed
a number of experimental studies on polymer–CNT ther-
mal conductivity, concluding that in general, the composite
TC can exhibit about two- to three-fold increase over
the matrix TC; this is a substantial increase but nowhere
Fig. 7. Thermal conductivity of epoxy/pine wood dust (PWD) compos-
near what is seen for electrical conductivity. Accord-
ites as function of the filler volume fraction. Triangles are FEA simulation ingly, several computational and theoretical studies have
results, and squares are experimental data. [38], Copyright 2010. Repro- been performed to explain this discrepancy. For example,
duced with permission from Elsevier Ltd. Shenogina et al. [46] simulated the heat flow between two
nearby nanotubes in a polymer matrix using ABAQUS finite
Fig. 7 shows the dependence of TC on the PWD volume element software. They concluded that the rate of heat
fraction; simulations and experiment are in a reasonable flow was fairly independent of the distance between the
agreement, although the non-monotonic behavior of the nanotubes, thus suggesting either absence or weakness
simulated TC is not fully explained and could be an artifact of percolation effects. Singh et al. [34] used element-free
of the lattice model. Galerkin method to numerically compute effective TC of
The examples above considered composites with spher- polymer/CNT composite as function of CNT volume frac-
ical fillers, monodisperse or polydisperse. Anisotropic tion and interfacial thermal resistance. The nanotubes were
fillers provide further complexity since both positional assumed to be perfectly aligned, and the calculated TC
and orientational distribution of fillers needs to be con- was along the direction of their alignment. As shown in
sidered. In recent years, many studies have been devoted Fig. 8, if the interfacial TC is equal to the nominal CNT
to polymer–carbon nanotube (CNT) composites [3,39–42]. thermal conductivity of about 3000 W/(m K), the com-
Indeed, with nominal TC of CNTs being in excess of posite TC increases rapidly with CNT concentration and
1000 W/(m K), one can expect to produce high-TC compos- could become an order of magnitude higher than that of
ites with relatively small CNT volume fractions. It has been the matrix (and thus, be on the order of 1–5 W/(m K)) at
shown [39,43,44] that for polymer/CNT nanocomposites, moderate CNT volume fractions (15–20%). However, if the
their electrical conductivity, , increases dramatically (by interfacial TC is relatively low (comparable to that of the
5–6 orders of magnitude) as the CNT volume fraction is matrix), the increases in TC are more modest, in line with
changed from about 0.2 volume percent to about 1 vol.%. experimental observations.
Indeed, according to percolation theory, high-aspect ratio More recently, Li et al. [47] used 3D finite element mod-
fillers such as CNT form continuous network at very low eling of polymer/single-wall-carbon nanotube (SWCNT)
concentrations (for the dependence of percolation thresh- composite with disordered nanotubes, and compared
old on aspect ratio, see, e.g., review by Bicerano et al. [45]). results with experimental data on two model systems,
This percolated network accounts for most of the charge epoxy/SWCNT and polyolefin/SWCNT (Fig. 9). The results
Fig. 9. (a) A finite element model of a randomly distributed SWCNT/polymer composite. The volume fraction of SWCNTs is 22%. (b) A temperature
distribution of a steady-state thermal analysis of a randomly distributed SWCNT/polymer composite. The volume fraction of SWCNTs is 22%. (c) Calculated
and experimental thermal conductivity (kc ) of SWCNT/epoxy composites as function of SWCNT volume fraction. (d) Same as (c) but for polyolefin/SWCNT
composites. Here, Rksm is the interfacial thermal resistance between SWCNT and matrix. [47], Copyright 2012. Reproduced with permission from Elsevier
Ltd.
demonstrated, once again, the importance of interfacial our ability to use simulations to screen formulations and
thermal resistance for proper understanding of heat trans- design candidate systems will increase dramatically.
fer in these materials. Even so, the mechanism of heat As already mentioned, one critical challenge in ensuring
transfer in the polymer/CNT composites is not yet fully the validity and accuracy of the continuum-level models
understood, as more recent experimental studies [48,49] (whether constitutive or FEA) is description of the heat
indicated that the increase in heat conductivity as function transfer through interfaces. Below, we briefly review recent
of the CNT volume fraction could be indeed consistent with progress on this topic.
the nanotube percolation.
Other recent applications of FEA modeling include 2.3. Interfacial thermal resistance modeling
those of polymer/platelet composites [50,51] and those
where mixed fillers are used [52]. Additionally, several Interfacial thermal resistance (often called Kapitza
authors have been working to combine FEA approach with resistance following the pioneering work by Kapitza [10])
molecular-level modeling (such as molecular dynamics) can be understood as follows. The matrix (such as poly-
to derive interfacial resistivity in a fundamental manner mer) and the filler (CNT or an inorganic particle) have
[51]. One can expect that in the next several years, FEA different spectra of quantum states for the heat carri-
simulations of composite properties, including TC, will be ers (electrons, holes, and phonons). In most polymers,
dramatically expanded and will include formulations with heat is carried only by phonons, and their density of
multiple fillers, varying degree of anisotropy, and other states is substantially lower than in highly crystalline
complexities. Furthermore, as computers become faster, CNTs or inorganic fillers. As a result, many phonons are
back-scattered at the interface between the matrix and the One important conclusion from these and other stud-
filler. The problem can be exacerbated if the interfacial cou- ies was that the surface functionalization of the filler could
pling between the filler and the matrix is weak [53–55]. be crucial to improving thermal conductivity of compos-
This results in a temperature discontinuity, T, at the ites. Within both theories of interfacial resistance (acoustic
interface; this discontinuity is proportional to the heat flux, mismatch model and diffuse mismatch model), the inter-
J (Eq. (15)) facial thermal conductivity is directly proportional to the
overlap between vibrational density of states (VDOS) spec-
J = (T ) (15)
tra of the matrix and the filler. Wang et al. [55] computed
The proportionality constant, , is the interfacial ther- VDOS of pristine or organically modified graphene sheet
mal conductivity and has units of W/(m2 K); the inverse of and compared it with those of a model paraffin (Fig. 10).
is the interfacial thermal resistance (usually labeled Rk ). The calculation shows that accurately matching the
It is often convenient to define interfacial thickness, h (Eq. VDOS of the matrix by using the right surface modi-
(16)), as follows: fier is very important. Even though the overlap integral
S increases substantially upon addition of virtually any
kB organic modifier, butyl surfactant results in the high-
h= (16)
est value, leading to the most effective reduction in
where kB is the bulk thermal conductivity in unit of the interfacial resistance for a given surface coverage
W/(m K). Typical values of h are on the order of nanometers (Fig. 11). This analysis shows that even modest surface
[54]; thus, it is often reasonable to disregard the interfa- functionalization could lead to substantial improvement
cial effects if the filler size is on the order of microns. In in the interfacial thermal transport. Recent experimental
many FEA simulations of conventional polymer–inorganic studies suggest that surface functionalization can indeed
composites discussed above, the assumption of “perfect” lead to some improvement in thermal conductivity in
interface (T = 0) was thus justified since the typical finite alumina/NylonTM -6 composites [63].
element size was much greater than h. However, if one Prediction of interfacial thermal properties is still rel-
tried to increase the composite thermal conductivity sub- atively new and rapidly developing field, and this short
stantially higher, by increasing the filler loading and/or review does not attempt to be comprehensive. As com-
reducing the filler size, the interparticle distance Dint would puters become more powerful and atomistic force fields
eventually decrease to the point where it becomes compa- more accurate, we can expect to see more advances in pre-
rable to h. Further increases in the filler volume fraction dicting interfacial thermal resistance for more and more
would then lead to no improvement in the overall TC. As diverse filler/matrix pairs. Creating such a library will help
Shenogin et al. [54] emphasized, in these nanocomposites, in developing more accurate macroscopic FEA models and
the fillers could be separated by 1 nm physically, yet from thus using modeling as part of new material design.
thermal standpoint, their separation could be as large as One critical remaining challenge for composite simula-
h ∼ 10–20 nm. This insight helps us better understand the tions so far is our limited ability to predict the morphology
results from the FEA simulations in which the interfacial and the filler dispersion and aggregation. Thus, most FEA
resistance was varied parametrically (Figs. 8 and 9) – once simulations require the a priori knowledge of the mor-
the filler volume fraction increased to the point where the phology (which could be either assumed theoretically or
interparticle distance becomes comparable to the thermal taken from experimental measurements such as electron
interfacial thickness, the overall thermal conductivity stops microscopy). Overcoming this difficulty would require bet-
increasing. ter understanding of the material preparation process,
So far, we mainly discussed continuum approaches to hydrodynamics and thermodynamics of fillers in polymer
thermal conductivity, where heat transfer through matrix, melt (for thermoplastics) or balance of flow, reaction, and
filler, and interface was described in a phenomenologi- filler motion (for thermosets).
cal manner. In order to predict the interfacial thermal
resistance, one needs to turn to molecular models. We 3. Thermal conductivity measurement techniques
already mentioned the study by Shenogin et al. [54]
who used molecular dynamics to study interfacial heat The measurement techniques of thermal conductivity
transport between a carbon nanotube and octane liq- can be divided into two groups. They are steady-state
uid. Alzina and co-workers [56] used multiscale modeling methods and transient methods. Steady-state methods are
to study heat transfer in epoxy/glass composites at very applied when the system has achieved stability, while the
low temperatures. They showed that theory and exper- transient methods are used during the process of heating
iment agreed fairly well if one assumed the so-called up or cooling down a material. Table 1 gives an overview
diffuse mismatch model (DMM), in which all phonons of main thermal conductivity measurement methods.
were assumed to diffusely scatter at the interface. Roy
et al. [57,58] demonstrated using molecular dynamics the 3.1. Steady-state methods
importance of coupling between the carbon nanotube and
the polymer matrix. Shin et al. [59] developed a multiscale Steady state condition means that temperature varies
model combining nonequilibrium molecular dynamics and across the sample but is generally time-independent (apart
finite element analysis to compute thermal conductivity in from small random fluctuations). The determination of the
SiC/epoxy composites. Similar approaches have been also thermal conductivity is based on the measurement of a
applied to various other systems [55,60–62]. heat flux and a temperature gradient. The main steady
Fig. 10. Comparison of VDOS between paraffin (blue) and graphene with various surface functionalities (red). The captions (S = 0.006, etc.) describe the
calculated VDOS overlap integrals. [55], Copyright 2015. Reproduced with permission from the American Chemical Society.
state methods include guarded hot plate method, axial flow 3.1.2. Axial flow method
method, heat flow meter method, pipe method. Axial flow method is the most widely used method for
thermal conductivity measurements at low temperatures.
3.1.1. Guarded hot plate In this method, a temperature gradient is created by locat-
Guarded hot plate method is used for testing materi- ing the test specimen between two reference specimens of
als with low thermal conductivity. The guarded hot plate known thermal conductivity to form a sample column with
device consists of one or two cold plates, a hot plate, a sys- a heater and a heat sink in the two ends [68–70].
tem of guard heaters and thermal insulation. The technique
concept is to form a temperature gradient by applying a
known heat flux on one side of the plate-like sample, and 3.1.3. Heat flow meter method
let the heat pass through the test sample to a cold plate The design of the heat flow meter method is quite sim-
on the opposite side of sample. The temperatures at each ilar to the single-specimen guarded hot plate apparatus,
side of the samples are measured when the system reaches with the difference that the main heater is replaced by
steady state [64]. Thermal conductivity can be calculated a heat flux sensor. Temperature difference was measured
based on heat flux, sample thickness and surface areas, and across a thermal resistor which contains a series of ther-
temperature difference [65–67]. mocouples [71–73].
Table 1
Comparison of different thermal conductivity test methods.
Methods Temperature Conductivity range Accuracy Materials Test standards

range (W/m K)
Guarded hot plate 80–800 K <0.8 2% Glass, polymers and ASTM C177
Steady-state method [64–67] insulation materials ISO 8302
methods EN 12667
Axial flow method 90–1300 K 0.2–200 2% Polymer, ceramics, metals ASTM E1225
[68–70]
Heat flow meter 253–523 K <10 3% Glass, polymers and ASTM C 518
method [67,71–73] insulation materials, ASTM E1530
ceramics ISO 8301
EN 12667
Pipe method [64,74] 293–2770 K 0.02–200 2% Metals, high conductivity ISO 8497
inorganics, polymer
composites
Laser flash method 373–3273 K >0.01 3–5% Glasses, polymers, ASTM E1461
Transient methods [64,75–78] ceramics, metals ISO 22007-4
ISO 18755
Transient hot wire 293–2273 K <25 1–10% Glasses, polymers, ASTM-C 1113
method [64,79–83] ceramics, most of liquid, ISO 8894-1
gas, powders ISO 8894-2
Transient plane source 20–1273 K 0.005–1800 5% Insulation materials, ISO 22007-2
method [84–87] powders, polymers,
ceramics, metals, liquids
Note: Data are cited from the references noted in method name cell.
3.2. Transient methods
The transient methods are generally used to measure

thermal diffusivity by recording temperature as function
of time following a transient or periodic heat added at the
sample surface. Thermal diffusivity can be calculated from
the specimen thickness and the time required for the tem-
perature rise to reach a percentage of its maximum value.
Thermal conductivity can be calculated through Eq. (17)
k = ˛Cp (17)
where k is the thermal conductivity, ˛ is the thermal diffu-

sivity, is the density, Cp is the specific heat. The transient
methods are much faster than steady-state methods. Typ-
ical measurement duration of one hour for a steady-state
measurement is reduced to a few minutes or to a subsec-
ond interval for a transient method. However, transient
methods usually have lower accuracy and require more
complicated data analysis.
Fig. 11. Interfacial thermal resistance (relative to that of pristine
graphene) as function of surfactant coverage for various functional- 3.2.1. Flash method
izations. [55], Copyright 2015. Reproduced with permission from the The flash method is the most frequently used method for
American Chemical Society. measuring solid thermal transport properties because of
its very wide temperature and thermal conductivity range.
The principle of the laser flash method is to heat a small disk
sample which is located in a furnace by a short laser pulse
3.1.4. Pipe method on the front side and to monitor the temperature increase
The pipe method is also referred to as the radial heat at its rear side by an infrared detector [64,75–78,88,89].
flow method. In most cases the apparatus consists of a
core heater placed at the central axis of cylindrical sam- 3.2.2. Transient hot wire method
ple and a heat sink around sample outside. Thermocouples Transient hot wire method is a modification of the
are mounted in the sample to record the temperature gra- steady-state pipe method with radial heat flux. The tran-
dient. In order to minimize the heat loss to the top and sient hot wire technique is based on a linear heat source
bottom, the ratio of sample length to radius should be (the hot wire) embedded in a test sample. The hot wire
large [64,74]. acts as both a temperature sensor and a heater. A constant
electrical current is applied to the wire. Thermal conduc- Table 2

Thermal conductivity of general thermoplastic polymers [3,95].
tivity can be calculated based on the rate of temperature
increase of the wire and power input [64,82,83]. Thermoplastic polymers Thermal
conductivity at
3.2.3. Transient plane source method room temperature
(W/(m K))
The transient plane source method, also called hot
disk method, is a transient pulse heating technique used High density polyethylene (HDPE) 0.33–0.53
Ultrahigh molecular weight polyethylene 0.41–0.51
for thermal conductivity and thermal diffusivity mea-
(UHMWPE)
surements. This technique uses a thin, plane, electrically Commercial thermotropic liquid crystalline 0.30–0.40
insulated resistive element, usually in a spiral pattern, as polymers (LCP)
both the temperature sensor and the heat source. Heating Polyoxymethylene (Homo) (POM) 0.30–0.37
element is placed between two test samples of the same Low density polyethylene (LDPE) 0.30–0.34
Poly(ethylene vinyl acetate) (EVA) 0.35
material. By recording the increase in resistance as a func- Polyphenylene sulfide (PPS) 0.30
tion of time while heating with an electrical current pulse, Poly(butylene terephthalate) (PBT) 0.25–0.29
the thermal conductivity can be deduced from one single Polytetrafluoroethylene (PTFE) 0.27
transient recording [84,85]. Polyamide-6,6 (PA66) 0.24–0.33
Polyamide-6 (PA 6) 0.22–0.33
Polyetheretherketone (PEEK) 0.25
4. Thermal conductivity of thermoplastics Polysulfone (PSU) 0.22
Polymethylmethacrylate (PMMA) 0.16–0.25
As discussed previously, polymer matrix TC plays a cru- Polycarbonate (PC) 0.19–0.21
cial role in determining the overall TC of thermoplastic Urethane base TPE (TPU) 0.19
Poly(acrylonitrile-butadiene-styrene) 0.15–0.20
composites [44,90]. It is therefore necessary to review our copolymer (ABS)
current understanding of TC of thermoplastic polymers. Polyvinyl chloride (PVC) 0.13–0.29
Molecular dynamics simulations and experimental Polyvinylidene difluoride (PVDF) 0.19
studies have shown that an individual polymer chain which Styrene/polybutadiene copolymer (SB) 0.17–0.18
Styrene-acrylonitrile copolymer (SAN) 0.15–0.17
is a low dimensional material, can in principle, have a rather
Poly(ethylene terephthalate) (PET) 0.15
high TC [91–94]. Henry and Chen [91] found that the TC Polystyrene (PS) 0.10–0.15
of single polyethylene (PE) chains with an extended con- Polyvinylidene chloride (PVDC) 0.13
formation could be as high as 350 W/(m K) when a chain Polyisobutylene (PIB) 0.12–0.20
length is longer than 100 nm. Even for PDMS with strong Polypropylene (PP) 0.11–0.17
Polyimide, thermoplastic (PI) 0.11
phonon scattering due to conformational disorder, its sin-
gle chain still has a TC of ∼7 W/(m K) [92]. However, bulk
polymers on a macro scale are usually thermal insulators. micrometer-size was recently reported to have exception-
Table 2 lists thermal conductivities of various commercial ally high TC up to 416 W/(m K) [101], which is due to its
thermoplastic polymers [3,95,96], which is in the range well-organized, highly crystalline, ordered, and oriented
of 0.1–0.5 W/(m K) as mentioned above. It is obvious that morphology, formed by a near-perfect self-assembling pro-
bulk polymers have much lower TC than corresponding cess.
single chains. For most polymers, both heat capacity Cp The above data suggest some general trends:
(primarily determined by the character of monomer repeat
unit) and phonon group velocity (strongly related to (1) Crystalline polymers (e.g. various PE) have higher TC
character/strength of backbone bonding) of bulk polymer than amorphous ones (e.g. PMMA, PS, SB, SAN, PI)
are almost the same as those of individual chains. There- and high crystalline polymers (e.g. HDPE, UHMWPE,
fore, it is the phonon mean free path (l) that causes the PBT) have higher TC than low crystalline ones (e.g.
large difference between the TC of single chains and that LDPE, EVA, PET, PEEK, PSU, PC). Exceptionally, PP is a
of bulk polymers. It has been known that phonon mean highly crystalline polymer, but has a very low TC either
free path along a chain is much larger than in a trans- because of relatively high free volume in its crystal-
verse direction since lattice vibrations associated with lites or due to the possible phonon scattering created
covalent bonds are much less anharmonic than those asso- by vibrational mode of side group methyl.
ciated with secondary bonds [97–99]. In a bulk polymer, (2) Addition of heavier atoms and presence of side chain
polymer chains are in the state of twisted random orien- or pendant group usually result in lower TC. The role
tation and intermolecular weak coupling, mainly via van of heavy atoms is illustrated by comparing, e.g. HDPE
der Waals forces. These make the average phonon mean vs. (POM, PBT, polyamide); for side group, e.g. HDPE vs.
free path for bulk polymer to be much smaller than that (EVA, PVC, PVDF, PVDC, PS, PP, PIB) (the presence of side
of a single polymer chain. Furthermore, there are inher- group can affect crystallization to some extent).
ent defects in bulk polymer, such as voids, impurities,
polymer chain ends, and entanglements [100], which fur- Below we elaborate on these general observations.
ther reduce the phonon mean free path and the TC of
bulk polymer. Clearly, to increase TC, attention should be 4.1. Effect of chain structure
paid to improving the orientation of polymer chains and
increasing the amount of ordered and oriented regions in Chain structure including molecular composition and
bulk polymers. For example, spider dragline silk with a conformation is the first level of structure that determines
Fig. 12. Thermal conductivity and structure of polymers. ␲-Conjugated polymers are plotted by solid lines with circle markers, while others are plotted
by dashed lines with triangular symbols. [102], Copyright 2014. Reproduced with permission from the American Chemical Society.
TC of various polymers. Simulation is the most widely polymers, thermal conductance path that follows the
used tool for studying the effect of chain structure on chain conformation is essentially random, which reduces
TC [102–104]. A recent large-scale molecular dynamics l and causes phonon scattering [105]. Crystallization can
simulation [102] systematically investigated the effect of increase the intrinsic order, which is likely responsible for
various molecular structures (as shown in Fig. 12) on TC higher TC in poly(ether ether ketone) [106], polytrifluo-
and showed that higher TC can be achieved in polymers rochloroethylene [105] and PE [107,108]. Polypropylene is
with rigid backbones, exemplified by ␲-conjugated poly- an interesting exception – it is highly crystalline but has
mers (such as polyacetylene, poly[diimidazo pyridinylene a low TC. This can be attributed to its low crystal density
(dihydroxy) phenylene], poly(p-phenylene benzobisthia- (i.e. low chain stacking density, ∼0.93 g/cc
1.00 g/cc of
zole), poly-(p-phenylene), polythiophene). It is because PE crystal) and/or the possible phonon scattering created
the ␲-conjugated rigid backbone can suppress segmen- by vibrational mode of side group methyl. Among thermo-
tal rotation and enable high bond strength for larger plastic polymers, PE and especially HDPE have the highest
phonon group velocity . Strong interchain interaction TC. Early study [107] indicates that the TC of HDPE can be
in polyketone, NylonTM , TeflonTM , and KevlarTM can sup- readily increased by increasing crystallinity and lamellar
press segmental rotation to some extent. However, the thickness through processing conditions.
presence of heavier atoms (e.g. oxygen), weaker back- In addition to the overall crystallinity, the crystal form
bone bonds and segments vulnerable to random rotations also affects TC of crystalline or semi-crystalline ther-
lead to lower and thus lower TC. Another simulation moplastic polymers. High-pressure and high-temperature
[104] also indicated that TC of polymer chains with mass treatment on HDPE and UHMWPE was conducted recently
disorder (other atoms or functional groups incorporated in the stable hexagonal phase [108]. The treatment
into aliphatic/aromatic pristine chains) should be about increased the crystallinity of both resins to almost the
25 times lower than TC of PE chain. It is the disorder same level, ∼68% (DSC). However, UHMWPE underwent
that creates localized vibrational modes, which impede the homogeneous and significant change from thin lamellae to
energy transport (driven by delocalized, long-wavelength extended chain crystal structure, while many thin lamellae
phonon modes). In addition, polymers with intrinsic ran- remained in HDPE. The lamellar thickness of UHMWPE was
dom segmental rotation (e.g. PTFE and PDMS [92]) have enlarged from ∼20 nm to the range of 100–150 nm. There-
low TC due to the conformational disorder-induced phonon fore, TC of UHMWPE increased from 0.42 to 0.58 W/(m K)
scattering. (37% increase), while that of HDPE only increased from
0.42 to 0.48 W/(m K) (11% increase). These results sug-
4.2. Effect of crystallinity and crystal form gest that it is the extended crystal structure and increased
lamellar thickness that are responsible for substantial TC
As seen from Table 2, crystalline polymers gener- improvement relative to a PE with randomly oriented
ally have higher TC than amorphous ones. In amorphous lamellae.
4.3. Effect of orientation
Heat is transferred more easily along the chains than

in transverse direction. Therefore, TC can be signifi-
cantly higher in the direction in which polymer chains
are preferentially oriented. As investigated by all-atom
model molecular dynamics simulation [109], mechani-
cal stretching is able to orient polymer chains that have
very high TC and thus enhance the TC of bulk polymers.
There are also numerous experimental studies on poly-
mer chain orientation for TC enhancement, encompassing
crystalline/semicrystalline, amorphous and various liquid
crystalline polymers.
4.3.1. Orientation of crystalline/semicrystalline polymers

Most work regarding the effect of chain orientation on
TC of crystalline/semicrystalline polymers was on PE. Sev-
eral methods for influencing PE chain orientation have been
developed, which include deformation by simple shear
[107], mechanical drawing [110], gel spinning [111–113]
and superdrawing [114] including superdrawing of sin-
gle crystal mats [115,116]. The general trend is that TC
Fig. 13. Thermal conductivities of various polymer fibers as functions of
along the orientation direction (TC||) rapidly increased with
their tensile moduli taken from the data sheets from the manufacturers.
increasing strain or draw ratio (), while TC perpendicular For each fiber type, the thermal conductivity is averaged over a minimum
to the draw direction (TC⊥) slightly decreased. Choy and of four fibers. [123], Copyright 2013. Reproduced with permission from
co-workers [110] investigated both TC|| and TC⊥ of ori- the American Chemical Society.
ented HDPE at relatively low . At = 25, TC|| increased
to 14 W/(m K) at 300 K and a TC anisotropy defined as investigated [123]. These fibers include PE (DyneemaTM
the ratio of TC|| to TC⊥ increased to around 60. Much SK75, UHMWPE with a crystallinity of ∼80%) and SpectraTM
higher TC|| was obtained later in high modulus PE fibers 900/2000 (HDPE) and liquid crystal polymers with rod-like
by significant increasing through gel spinning [111–113] segments (VectraTM , KevlarTM , poly(p-phenylene benzo-
and superdrawing of single crystal mats [115,116]. For bisoxazole) (PBO), polyhydroquinone diimidazopyridine
example, the TC|| of UHMWPE with a of 250 reached (PIPD, MA5S) and poly(p-phenylene benzobisthiazole)
37.5 W/(m K) [113]. Through more advanced superdraw- (PBT)). Fig. 13 shows the axial TC (i.e. TC||) vs. tensile mod-
ing of UHMWPE single crystal mats, Choy et al. [116] ulus of single fibers. PE and PBO fibers have the highest
produced PE ultradrawn fibers with up to 350 and values of ∼20 W/(m K). VectraTM has the lowest TC, which
achieved TC|| of 41.8 W/mK. With the increase in , crys- is probably attributable to the disorder in the chain seg-
talline lamellae are broken into small crystalline blocks. ments and also the relatively low crystallinity of about 20%.
The crystalline blocks are stacked and connected by taut For LCP fibers, the increase in the number of side chains in
tie molecules (intra-microfibrillar tie molecules) originat- each molecular unit is likely to reduce the axial conduc-
ing from partial chain unfolding, thus forming microfibrils tivity due to additional phonon scattering created by the
along the draw direction. The microfibrils are connected vibrational modes of the side chains.
laterally by bridging molecules, i.e. inter-microfibrillar tie Besides the polymer microfibers above prepared by
molecules. After ultra-drawing, microfibrils are further various mechanical drawing techniques, nanofibers are
deformed (accompanied by a decrease of its volume frac- of particular interest due to their more perfect structure.
tion) and the inter-microfibrillar tie molecules can be Shen and co-workers [124] recently developed a two-
significantly aligned to form long extended-chain crys- stage heating method to further scale down the fiber size
tals, even protofibril substructure (i.e. needle crystals) and fabricated ultradrawn nanoscale fibers. The obtained
[112,117,118]. Such morphology evolution leads to the ultradrawn PE nanofibers have a diameter of 50–500 nm
focus of group velocity into parallel direction and the and an estimated overall of 60–800. The obtained TC||
increase of mean free path, both of which can increase TC||. was as high as 104 W/(m K), which was about three times
It is expected that TC|| can be increased continuously with higher than previously reported values of micrometer-
the development of structure order and perfection. sized fibers. It was found that such PE nanofibers had
Apart from PE, TC of various thermoplastic polymers, a strong single-crystal nature, which confirmed that the
such as polyacetylene [119], polyphenylene sulfide [120], stretching did contribute to the nanoscale restructuring
poly(ether ether ketone) [106], other semicrystalline poly- of the polymer chains. Furthermore, fiber quality was
mers (POM, PP, polychlorotrifluoroethylene, PVDF and PET) improved with the elimination of inherently large defects
[121] and certain polymer blend [122] was reported to such as voids, impurities or large entanglement regions in
significantly increase by mechanical drawing/orientation. nanoscale size. Higher draw ratios and greater volume frac-
Recently, TC of nine commercially available high-modulus tion of chain extended crystalline region are thus more
polymer fibers (diameters 10–30 ␮m) was systematically achievable, which leads to more ‘ideal’ single-crystalline
fibers and extremely high TC. However, the TC|| of ultra-

drawn PE nanofibers cannot approach the predicted values
for a single PE chain [91] because nanofibers are composed
of multiple chains so that the inter-chain van der Waals
interactions induce phonon–phonon scattering. Even so,
based on MD simulations, it was conjectured that TC|| of
PE nanofibers can be increased to about 180 ± 65 W/(m K).
There are other simple production approaches for mak-
ing highly thermally conductive polymer nanowires or
nanofibers. For example, by an improved nanoporous tem-
plate wetting technique, HDPE nanowire arrays with a
diameter of 100 nm and 200 nm were fabricated [125]. The
estimated TC of a single HDPE nanowire is 26.5 W/(m K),
which indicates that the nanowires have a similar ori-
entation degree to ultradrawn fibers with of 50–100
[116]. NylonTM -11 nanofibers [126] were fabricated most
recently by electrospinning, a simple and versatile method
to produce polymer nanofibers. When the NylonTM -11 fiber
diameter was decreased to ∼200 nm, the TC was found
to rapidly increase because crystallites exhibited a signifi-
Fig. 14. Single-fiber thermal conductivity as a function of temperature.
cantly higher degree of preferred orientation along the fiber amorphous polythiophene (a-PT); crystalline polyethylene (c-PE) and
axis and crystal size started to grow. However, it should be crystalline polybenzobisoxazole (c-PBO). Black dashed line represents the
noted that the TC of electrospun NylonTM -11 nanofibers predicted minimum thermal conductivity for PT. [131], Copyright 2014.
(up to ∼1.6 W/(m K)) are much lower than that of the PE Reproduced with permission from the Nature Publishing Group.
nanofibers [124], which can be attributed to the intrinsic
difference in the molecular structure and the lower crys- and high TC comparable to PE fibers [132] at high temper-
tallinity of NylonTM . atures (see Fig. 14).
4.3.3. Orientation of liquid crystalline polymers

4.3.2. Orientation of amorphous polymers The effect of chain orientation on TC of rigid-rod or
Drawing of crystalline/semicrystalline polymers usually semi-rigid polymers, such as commercial nematic liquid
increases not only chain alignment but also crystallinity crystalline polymer (LCP) is quite different from the case of
[116]. Thus, studying purely amorphous polymers can sep- flexible-chain polymers. LCP usually has ordered structure
arate these factors and clarify how chain alignment affects in which molecular chains are micro-periodically oriented
TC during drawing. Early studies [127–130] showed that in the nematic liquid crystalline phase. Such orientational
TC of stretched polymethyl methacrylate and polystyrene order can be easily maintained in the LCP samples obtained
could be increased in the stretching direction. However, via a common process. TC of LCP is, therefore, twice as
the TC increase was relatively small, less than two-fold high as that of many other resins in the state in which the
in most cases. A recent study [131] showed that pure molecules are not macroscopically aligned [133]. However,
polythiophene (PT) nanofibers made via template-assisted it is not the case after mechanical drawing. In Fig. 15, the
electropolymerization can achieve a TC up to ∼4.4 W/(m K) dependence of TC|| for LCP VectraTM B950 is compared
while remaining amorphous. The TC of PT nanofibers with that of several flexible chain polymers [134]. TC|| of
increased with decreasing diameter, a trend that agrees LCP rises sharply at low , then reaches a plateau, which is
with the measured molecular chain orientation along exactly consistent with the trend of orientation degree of
the fiber axis. The PT fibers exhibited a distinct trend LCP as a function of . The semi-rigid chains in the local-
of temperature-dependent TC when compared with two ized domains of the LCP nematic melt already have parallel
commercial high-modulus and highly crystalline fibers: c- orientation. Only a small elongational stress is needed to
PE and c-PBO [123]. As shown in Fig. 14, both commercial induce the chain alignment and lead to saturation of both
fibers exhibited TC ∝ 1/T near room temperature, while TC orientation degree and TC|| at a low draw ratio. For flexible
increased modestly (TC ≈ T0.4 for d = 71 nm and TC ≈ T0.2 for crystalline/semicrystalline polymers, the degree of chain
d = 245 nm) for the amorphous PT nanofiber. The dominant orientation in amorphous regions and the number of taut
phonon-scattering mechanism in the drawn crystalline tie molecules increase steadily up to the highest achievable
fibers at room temperature is anharmonic phonon–phonon and thus TC|| increases continuously [116]. Therefore, the
scattering (not structural disorder), but the nanofibers here TC of the VectraTM at high is significantly lower than that
appear still to be dominated by interchain scattering due to of PE and POM. The simulation study [102,104] attributed
structural disorder, despite some degree of chain orienta- such phenomenon to the chain segment disorder and mass
tion. It is likely that the PT nanofibers have a strong increase disorder, which was induced by the fact that the VectraTM
in short-range order, but long-range order remains absent. backbone consists of randomly distributed phenyl groups
Compared with ultradrawn PE fiber, the PT fiber may be as well as carbon and nitrogen atoms along the chain.
more suitable for some high-temperature applications (e.g. In addition to mechanical stretching, other approaches,
thermal interface materials) considering its better stability such as injection molding [133] and high magnetic field
(0.3 W/(m K) in the other two directions). However, only

0.53 W/(m K) was achieved when the Mn doubled. In that
case, the chain-folding lamellae were randomly oriented,
even after injection molding. It was thought that molec-
ular entanglement and interlayer connectivity of lamellae
become greater with increasing Mn , which inhibits lamel-
lae sliding that is necessary for the parallel orientation.
Therefore, a low Mn is necessary to achieve high TC⊥ for
this kind of material, likely leading to inferior mechani-
cal properties. Further studies [142,143] on the composites
of this smectic LCP indicated that polymer lamellae tend
to lie parallel to the filler surface and are stacked toward
neighboring fillers. In other words, polymer chains are per-
pendicularly oriented between fillers and thus serve as
effective heat paths. The TC of the composites in both direc-
tions, in particular TC⊥, was significantly enhanced.
4.4. Polymer blends by engineered interchain

interactions
As discussed in the previous sections, molecular struc-

Fig. 15. Draw ratio dependence of the axial thermal conductivity of
VectraTM B950, HDPE, polyoxymethylene (POM), and polypropylene (PP)
ture, crystallinity and chain orientation are the important
at 300 K. [134], Copyright 1995. Reproduced with permission from John factors influencing TC. Chain orientation has been the
Wiley and Sons. most explored route to increase TC in polymers. However,
usually high TC can be obtained only in the orientation
[133,135,136] can enable LCP chain orientation as well. direction. It is thus desirable to explore polymers whose
Sumimoto’s research [133] showed that after injection TC can be increased in all directions, and yet still can be
molding TC of EkonolTM type of LCP in flow direction was processed with conventional manufacturing processes.
five times higher than that in the traverse direction. When Most recently, a breakthrough finding [144] was
a 10 T magnetic field was applied, the orientational order reported that a high TC (over 1.5 W/(m K)) can be achieved
parameter of LCPs (including EkonolTM (Type I), VectraTM in amorphous polymer blends by engineering inter-
(Type II) and X7GTM (type III)) exceeded 0.8, sometimes chain interaction. There are several requirements for the
even 0.9. Accordingly, the TC in magnetic field direction amorphous polymer blends to have high TC: (1) strong
was at least tripled, even increased by one order of magni- intermolecular bonds to replace weak interactions; (2) the
tude to higher than 2.5 W/(m K) [133]. In certain LCP film, a intermolecular bonds must connect as closely as possi-
TC over 10 W/(m K) has been achieved using a strong mag- ble to polymer backbone; (3) homogeneous distribution of
netic field [136]. However, such magnetic-field alignment these bonds at a concentration above percolation threshold
is not easy to realize in practical manufacturing processes to form a continuous thermal network; H-bond-accepting
(e.g. injection molding), especially for parts with complex and -donating polymers have strong intermolecular bonds
shapes. (hydrogen bonding). However, they are required to be mis-
For most LCPs, TC usually increases in the flow or draw- cible at the molecular level to allow polymers to intertwine
ing direction. However, Yoshihara and his co-workers [137] within the radius of gyration, as shown in Fig. 17(a). Such
recently demonstrated that some main chain LCPs can requirements are demonstrated by the purposely designed
achieve a high TC (greater than 1.0 W/(m K)) in the direc- polymer pairs, as shown in Fig. 17(b). Results showed that
tion perpendicular to the flow during injection molding only certain blends of PAP and PAA with strong H-bond
process. Fig. 16(a) shows the molecular structure of that concentration exceeding percolation threshold yielded a
LCP, which is a main-chain PB-n polyester, where n is the dense and homogeneously distributed thermal network
even number of methylene units in the spacer moiety. and reached a TC as high as 1.5 W/(m K), which is nearly an
This type of polymer is known to form a highly ordered order of magnitude higher than that of typical amorphous
smectic liquid crystalline phase with a chain-folding lamel- polymers. The rigid and short PAP chains were thought to
lar structure in both smectic and crystalline phase [138]. penetrate into the PAA radius of gyration and hold the flex-
When the PB-n is in a smectic phase, shear flow can induce ible PAA backbone in an extended state by means of strong
the orientation of the chain-folding lamellae [139–141], H-bonds. Such extended backbone is expected to improve
either in parallel (Fig. 16(b)) or perpendicular direction intrachain heat transfer, analogously to the oriented poly-
(Fig. 16(c)). It was found that injection molding of PB-n in mers discussed previously. However, for PAP:PVA and
the smectic phase selectively induced the parallel orienta- PAP:PVPh pairs, little or no TC improvement was observed
tion. Therefore, polymer chains were aligned in the normal possibly due to the weak H-bonds, the presence of heavy
direction with respect to the molding surface, thus lead- and large benzene linkers, lack of continuous network of
ing to a high TC⊥. The TC of this main chain LCP also has strong hydrogen bonds and blend homogeneity.
a strong dependence on molecular weight (Mn ). For exam- This work paves a novel way to increase TC for amor-
ple, PB-10 with Mn = 8000 g/mol had a TC⊥ of 1.20 W/(m K) phous polymers. However, it should be also noted that the
Fig. 16. (a) Chemical structure of the smectic LCP main-chain PB-n; (b) parallel and (c) perpendicular orientations of lamella by shear flow. The TC⊥ in (b)
is illustrated. [137], Copyright 2012. Reproduced with permission from John Wiley and Sons.
large increase was in the cross-plane TC in spin-cast (from improvement on the TC of final composites given the same
polymer solution) films with a nano thickness up to 70 nm. filler loading [146]. Thus, to achieve high TC in typical ther-
It is interesting to know whether this high TC is isotropic mosets, one needs to typically use high filler levels, thus
and can be maintained for bulk thick films. It is also inter- increasing the viscosity of composite resin and leading to
esting to know whether these polymer blends still have poor processability.
high TC if they are prepared by melt mixing. As discussed before, ordered structure in polymers
can promote the movement of phonons, and thus can
5. Thermal conductivity of thermosets lead to resins with higher TC. Liquid crystalline ther-
mosets, possessing long rigid rod-like “mesogen” groups,
Thermosets have also been used as the matrix for ther- can form ordered structure. When thermosets with meso-
mally conductive composites due to their easy molding gen structure are cooled from isotropic state at high
capability, high thermal resistance, and high electrical temperatures, mesogen groups can aggregate and form
insulating characteristics. Similar to most thermoplas- liquid crystal domain in matrix at a certain temperature
tics, thermosets such as epoxy resin are amorphous and range. This ordered liquid crystal domain can effectively
have a low thermal conductivity in a range from 0.17 to increase the TC of matrix. Table 3 displays the TC of vari-
0.21 W/(m K) [4,145]. As discussed before, the TC of the ous thermosets including both amorphous thermosets and
matrix has a huge effect on the TC of final composites. Even liquid crystalline thermosets. Normal non-mesogen ther-
a small increase on the TC of matrix can lead to significant mosets such as DGEBA can only form isotropic amorphous
Fig. 17. (a) Illustrations of heterogeneous (left) and homogeneous (right) distributions of thermally conductive interchain connections; (b) polymer pairs
designed to exhibit interchain H-bonding. The OH group is closely connected to the main chain for PAA and PVA, whereas it is attached to the backbone
via a benzene ring linker for PVPh. [144], Copyright 2014. Reproduced with permission from the Nature Publishing Group.
Table 3
Thermal conductivities of various thermosets.
Thermosets Monomer Hardener Thermal conductivity

W/(m K)
0.19 [148]
DGEBA
0.20 [95]
0.20 [147]
BACE – 0.29 [149]
0.26 [148]
DGETMBP
0.29 [148]
0.15–0.17 [150]
PMDA
0.44–0.45 [150]
0.44–0.46 [150]
0.30 [145,148,151]
DGEBP
0.43–0.44 [150]
DGEHAP 0.33 [152]
DGETAM 0.38 [147]
DGEHH 0.39 [153]
DGEDH 0.50 [154]
BZE 0.69 [148]
TME8 0.83–0.85 [145,148]
0.88 [148]
TME6
0.89 [145]
0.96 [145]
TME4
1.05 [148]
Fig. 20. Possible schematic image of smectic microstructure from twin-

mesogen thermosets. Dotted lines represent the lattice plane of smectic-
like structure. [145], Copyright 2003. Reproduced with permission from
John Wiley and Sons.
microscopic anisotropic liquid crystalline domains due to

␲–␲ stacking. This ordered structure leads to higher matrix
Fig. 18. Relationship between thermal conductivity and domain size and TC. However, most mono-mesogen thermosets can only
content [148].
form weakly-oriented nematic liquid crystals with either
small liquid crystal domains (size less than 400 nm) or
morphology, so the relevant TC is in a very low range of low content of liquid crystal domain (less than 25 vol.%).
0.19–0.20 W/(m K) [95,147,148]. Due to the existence of Therefore, the TC improvement is limited (about 2–3 times
ordered structure, liquid crystalline thermosets exhibit a compared to that of typical thermosets).
much higher TC. The highest value is around 5.5 times that Some mono-mesogen hardeners can also result in the
of DGEBA. formation of liquid crystal in thermosetting matrix and
higher thermal conductivities can be obtained as well. As
5.1. Effect of liquid crystal domain size and content shown in Table 3, when PMDA is cured with a normal hard-
ener, a low TC of only 0.15–0.17 W/(m K) was obtained.
In fact, TC has a strong dependence on the liquid crys- However, when it was cured with mono-mesogen harde-
tal domain size and content [155]. As shown in Fig. 18, TC ners, the TC can be increased to 0.44–0.46 W/(m K) [150].
increases with the ratio of anisotropic structure. The ratio It is worth noting that, even with both mono-mesogen
of anisotropic structure is defined as the volume percent- resin and mono-mesogen hardener, the liquid crystalline
age of anisotropic structure in the whole matrix, which also domain size and content are still limited. Therefore, overall
indicates the content of liquid crystal domain [148]. More- TC was still below 0.70 W/(m K) [148].
over, the dependence is much stronger when the size of Another kind of liquid crystalline thermosets are twin-
liquid crystal domain is larger than 400 nm and its content mesogen thermosets, which have two mesogens linked by
exceeds a threshold of 25 vol.%. a spacer. As shown in Fig. 20, this special structure can
Mono-mesogen thermosets which have one single lead to highly oriented smectic liquid crystalline morphol-
mesogen are the most common liquid crystalline ther- ogy after curing [145,156]. Their liquid crystal domain size
mosets. As shown in Fig. 19, the mesogen groups can form and content are much higher than 400 nm and 25 vol.%,
Fig. 19. Schematic representation of the strategy to afford macroscopic isotropic resin having high thermal conductivity. [145], Copyright 2003. Reproduced
with permission from John Wiley and Sons.
respectively, which results in much improved TC. TME4, spherulitic crystals increased from 10 to 80 ␮m. The rele-
TME6 and TME8 that are typical twin-mesogen thermosets, vant TC increased with the size of spherulitic crystals from
as shown in Table 3, have very high liquid crystal domain 0.33 to 1.16 W/(m K).
size (>1600 nm) and content (>40 vol.%). Their TC can reach It is also worthwhile to point out that liquid crystalline
a very high range of 0.83–1.05 W/(m K) [145,148]. Although thermosets with high TC usually have a high melting point.
twin-mesogen thermosets showed very high TC, their com- As they need to be heated to above its melting point for
plicated synthetic procedure and very high raw material mixing with fillers and hardeners and processing, this could
cost make them difficult to be commercialized. lead to short gel time and poor processability.
5.2. Effect of curing condition

5.3. Effect of orientation
Curing condition is also a key factor to the forma-
tion of liquid crystal structure which exists at a certain For liquid crystalline thermosets, the application of a
temperature range. Generally, when liquid crystalline ther- powerful external field, such as an electric field [158] or a
mosets undergo heating, they transform from crystalline magnetic field [159,160] is very effective for obtaining high
state to strongly oriented smectic liquid crystalline state, orientation of mesogen groups. In general, mesogen groups
then to weakly oriented nematic liquid crystalline state. will be aligned through the direction along the applied field.
With further temperature increase, the nematic liquid Therefore, TC will significantly increase in the orientation
crystal will eventually disappear and finally isotropic amor- direction, while the TC in the direction across the applied
phous state is reached. Some liquid crystalline thermosets field will decrease to a certain degree.
do not exhibit smectic state, but directly transition from Harada et al. [161] studied the effect of magnetic field
crystalline state to nematic state. Therefore, for liquid crys- on the TC of liquid crystalline DGETAM epoxy cured with
talline thermosets, lower curing temperature will promote 4,4 -diaminodiphenylethane (DDE). The authors cured
the formation of highly oriented structure and thus lead to DGETAM/DDE system with and without magnetic field of
a higher TC. 10 T (superconducting magnet) at 170 ◦ C. For the system
Harada et al. [147] cured diglycidyl ether cured without a magnetic field, the TC was 0.43 W/(m K).
of terephthalylidene-bis-(4-amino-3-methylphenol) For the system cured under the magnetic field (10 T), the TC
(DGETAM) at different temperatures and compared ther- in the direction along the applied magnetic field reached a
mal conductivities of resulting thermosets. When cured very high value of 0.89 W/(m K), while the TC in the direc-
at 200 ◦ C, the system formed an isotropic amorphous tion across the applied field was only 0.32 W/(m K). Wide
phase and the TC was 0.35 W/(m K). When the curing angle X-ray diffraction (WAXD) results confirmed that the
temperature decreased to 160 ◦ C, clear birefringence matrix cured under the 10 T of magnetic field had a highly
patterns indicated the formation of liquid crystalline ordered network structure along the direction of the mag-
phase. The TC increased to 0.38 W/(m K). Another example netic field as shown in Fig. 22.
is 4-(4-oxylanylbutoxy)benzoic acid-1,4 -phenylene ester Some other methods were also developed to obtain
(BZE) epoxy [148]. When cured at isotropic amorphous highly oriented liquid crystalline thermosets. Geibel et al.
state at 230 ◦ C, the TC of BZE was only 0.44 W/(m K). When [162] coated the UV-curable liquid-crystalline monomers
cured at liquid crystalline state at 180 ◦ C, its TC increased on the top of uniformly rubbed polyimide. Homogeneous
to 0.69 W/(m K). orientation of liquid crystals in the rubbing direction was
Song et al. [157] also studied the effect of curing condi- obtained. The UV-curable monomer layer held in the liq-
tion. The authors found that there were many spherulitic uid crystalline state was cured by irradiation. The TC
crystals formed in cured epoxy DGEBP, as shown in Fig. 21. reached up to 5.2 W/(m K) in the orientation direction.
With decreasing initial curing temperature, the size of However, the TC normal to the oriented direction was only
0.33–0.34 W/(m K).
Fig. 21. POM photograph of well-developed spherulite crystals. [157], Fig. 22. Model of DGETAM/DDE cured under the magnetic field. [161],
Copyright 2012. Reproduced with permission from Elsevier Ltd. Copyright 2003. Reproduced with permission from John Wiley and Sons.
Table 4
Thermal conductivities of some thermally conductive fillers.
Fillers Category TC W/(m K) Electrically conductive? Refs.
Aluminum Metal 234 Yes [167]

Copper Metal 386 Yes [168]
Silver Metal 427 Yes [169,170]
Carbon nanotube (CNT) Carbon-based 1000–4000 Yes [3,171]
Carbon fiber Carbon-based 300–1000 Yes [172,173]
Graphene Carbon-based 2000–6000 Yes [174,175]
Graphite Carbon-based 100–400 Yes [176]
␤-Silicon nitride (␤-Si3 N4 ) Ceramics 103–200 No [177]
Hexagonal boron nitride (h-BN) Ceramics 185–300 No [147,178,179]
Aluminum nitride (AlN) Ceramics 100–300 No [180,181]
Diamond Ceramics 1000 No [182]
␤-Silicon carbide (␤-SiC) Ceramics 120 [183]
␣-Alumina (␣-Al2 O3 ) Ceramics 30 No [184,185]
Beryllium oxide (BeO) Ceramics 270 No [186,187]
6. Thermally conductive fillers low TC and is not a very good candidate for high thermally
conductive composites [193]. Carbon fibers [194] and CNTs
Given that polymers have low TC (generally less than are one-dimensional (1-D) fillers, and have a high TC along
1.0 W/(m K)), making high-TC polymer-based materials the longitudinal direction. The TC of carbon fibers varies
requires introducing high-TC filler particles into polymer over a wide range depending on the organic precursor
matrix [163–165]. The TC of fillers is largely determined and processing conditions used. The two most important
by the way how heat transfers in filler particles. In gen- precursors for carbon fibers are polyacrylonitrile (PAN)
eral, fillers with purely phonon heat transfer mechanism [172] and mesophase pitch (MPP) [173]. Mesophase
(as opposed to those having both phonon and electron pitch-based carbon fibers have a much higher TC (up to
ones) have lower TC. Metal oxide such as ␣-Al2 O3 only has 1000 W/(m K)) resulting from highly crystalline graphitic
a TC around 30 W/(m K) [166]. Even for highly thermally structure and its high degree of orientation parallel to
conductive ceramic materials, such as BN and AlN, TC can the fiber axis [172]. Therefore, a large number of research
only reach to 100–300 W/(m K). In contrast, free electrons efforts have been focused on using mesophase pitch-based
are much more efficient in transporting heat compared to carbon fibers as thermally conductive reinforcements
phonons since electrons are more resistant to scattering [195,196]. Carbon nanotubes (CNTs) are an allotrope of
and move at higher speeds. Thus, metallic and carbon- carbon having a cylindrical structure and show extremely
based fillers show a much higher TC mainly due to free high TC [171]. There is a recent review on more detailed
electrons. Table 4 lists both thermal and electric properties treatment of polymer–CNT composites [3], in which the
of commonly used thermally conductive fillers. influence of atomic structure, tube size, morphology,
defects, purification, micro/nano-structure, and interfaces
6.1. Metallic fillers between CNTs and polymer matrix on the TC have been
discussed extensively. Recently, it was discovered that
Various metallic fillers, such as nickel, copper, alu- graphene has extremely high TC, which even exceeds
minum, silver, have been widely used in polymer that of CNTs [197]. Graphene is a two-dimensional carbon
composites for improving TC [188]. In general they could material with a single layer structure, for which it is easy to
be quite effective in increasing TC of composites compared form effective pathways for heat transfer. It is considered
to the pure matrix polymer. However, adding metallic par- to be a good thermally conductive additive for thermal
ticles into polymers will also cause a substantial increase in engineering application. However, the large surface area
the electrical conductivity of the composites and/or reduc- of graphene makes it difficult to be incorporated (“exfoli-
tion in the dielectric breakdown voltage. Thus, metallic ated”) in polymer at high levels and its high material cost
fillers can only be used in applications where electri- is another concern that may limit its further industrial
cal insulation and dielectric breakdown strength are not applications. Even so, compared to other nanosized carbon
important. materials, graphene is abundant, relatively cheap, and
highly thermally conductive, and has already found many
industrial applications.
6.2. Carbon based materials
Carbon-based materials, including graphite [90], 6.3. Ceramic fillers

graphene [189,190], carbon nanotube (CNTs) [3], carbon
fiber [191], etc., have been widely studied as heat conduc- Ceramic fillers have been widely studied for thermally
tive fillers because of their high intrinsic TC [192]. Even a conductive and electrically insulating composites due to
very small loading of carbon-based fillers can significantly their inherent properties. They lack free electrons, and
increase the TC of composites. Another advantage is the heat transfer is predominantly through phonons. Gen-
their light weight compared with metal or ceramic fillers. erally, except for BeO, most metal oxide fillers such as
Among carbon-based materials, carbon black has relatively alumina (Al2 O3 ) [184,185] and silica (SiO2 ) have a lower
TC. Non-oxide fillers, such as aluminum nitride (AlN) drastic improvement be observed. This implies that effi-
[198,199], boron nitride (BN), silicon nitride (Si3 N4 ) [200] cient thermally conductive pathways start to form at high
or silicon carbide (SiC) [183] have high TC as their strong filler loading levels due to filler-to-filler connections. This
interatomic bonds and crystal structure can significantly phenomenon is described well by various models [207]. It
reduce phonon scattering. In particular, BN and AlN fillers should be noted that increasing loading level is not always
are widely used for thermally conductive and electrically preferred, as a high filler loading level can cause the com-
insulating composites [201]. However, these fillers also posite to be brittle, have poor processability and result in a
have some limitations, e.g., BN has an extremely low TC high cost.
in the through-plane direction, and AlN particles are easy
to hydrolyze [179,202] (although the moisture reactivity of 7.2. Effect of filler shape
AlN could potentially be mitigated by surface modification
such as silica coating [203]). The filler shape also has a significant influence on
TC is closely related to the crystal structure and crys- the performance of polymer composites. The effects of
tallinity of fillers. The alignment of atoms and their filler shape on the TC of composite materials, especially
interactions affect the efficiency of the heat transfer when the fillers are distributed randomly, were extensively
dramatically. In fact, even two materials with identical studied [208,209]. It is important to note that thermal con-
chemical formula could have different TC, if they possess ductivity is not the only property that depends on the
different crystal structure. For example, there exist ␣ and filler shape. For processing polymer composites, lower vis-
␤ phases in Si3 N4 crystals, corresponding to low- and high- cosity is desired. However, high filler loading above the
temperature polymorphs, respectively. The TC of ␤-Si3 N4 percolation threshold is usually needed to achieve high TC.
in the direction parallel to the c-axis is much higher than High filler loading will normally cause a significant increase
that of the isotropic ␣-Si3 N4 [204]. Another example is on the viscosity of composites and make it more difficult
boron nitride, which can change from hexagonal structure for composites to be processed by conventional polymer
to cubic structure under high temperature and pressure processing techniques. The rise of viscosity mainly stems
[205]. Hexagonal BN is comprised of planar sheets of cova- from the interfacial friction between fillers, and between
lently bonded boron and nitrogen atoms, which make up fillers and polymers. Many low-cost fillers have irregular
the in-plane structure of the crystal. Different layers are shapes with many edges and corners, and can cause signif-
held together in the through-plane direction through Van icant friction during mixing or compounding. The irregular
der Waals forces. Due to its special crystal structure, it shape can cause a significant increase in the viscosity of
shows anisotropic TC with ∼300 W/(m K) in in-plane direc- composites when the filler loading is high. When deal-
tion and ∼2 W/(m K) in through-plane direction. After the ing with irregular shaped fillers with high hardness, for
transition into cubic phase, boron nitride becomes more example, irregular Al2 O3 powders, tool wearing could be an
isotropic, and its in-plane and through-plane TC become issue. Other characteristics of fillers, such as rough surface,
nearly equal and high. Similar situation exists between low flowability, aggregation, breakability, will be unfavor-
graphite and diamond [206]. Better crystallization means able for the processing of the composites, and may damage
fewer defects in crystal structure and can help improve heat TC of the composite as well. Fillers with smooth surface and
transfer by reducing the phonon scattering. Also, presence spherical morphology are favorable for lowering down the
of impurities will increase defects in the materials. Defects friction and improving processability, yet they could be less
can cause additional scattering of the electrons or phonons, efficient in increasing the composite TC. Below, we discuss
resulting in the lower overall thermal conductivity. Grain the interplay of these trends for various filler types.
boundaries between crystals (grains) in an inorganic filler
can be a significant source of phonon scattering. 7.2.1. One-dimensional fillers
Polymer composites with low-aspect-ratio fillers usu-
7. Influence of the filler morphology and properties ally exhibit a lower TC even at high filler contents. High-
on thermal conductivity aspect-ratio fillers are thus beneficial to achieve low per-
colation threshold for composites [210]. One-dimensional
Besides the intrinsic TC of fillers, the main factors affect- (1-D) fillers, such as fibers, rods, wires or tubes, are
ing the thermal properties of composites are filler volume expected to significantly improve the TC of composites
fraction (also referred to as loading level), particle shape, compared with fillers with other shapes, as it is easier for
particle size, and adhesion between the fillers and matrix high-aspect-ratio fillers to construct long heat conductive
and the thermal property of interface. pathways along the longitudinal direction of 1-D fillers in
the composites.
7.1. Effect of filler loading level For 1-D fillers with high-aspect-ratio, the TC along the
longitudinal direction is usually much higher than that in
The dependence of composite TC on filler loading level the perpendicular direction. Therefore, if fillers are ori-
was largely discussed in Section 2, so here we only recap ented in composites, the composites will show anisotropic
very briefly. In general, TC of composites increases with TC. In fact, not only TC, but also other physical properties
increase in the volume fraction of thermally conductive may be anisotropic depending on 1-D filler orientation.
filler. The increase is often non-linear [207]. At lower filler Glass fiber [211] is the most commonly used filler for
loading levels (<35 vol.%), the TC increase can be rela- strengthening polymer composites; however, its TC is not
tively small. Only at high filler loading levels can a more high. Carbon fibers [194] (and more recently CNTs [3])
are also commonly used in composites, as they can offer higher filler loading and improved processability [216,219].
mechanical strength and extra TC. Takafumi Kusunose et al. For example, spherical BN was synthesized by the reac-
[212] reported the synthesis of ␤-Si3 N4 nanowires and tion of aerosol droplets containing water-soluble boric acid
evaluated the TC of ␤-Si3 N4 nanowire/epoxy composites. (H3 BO3 ) with ammonia gas at elevated temperatures in
For comparison, commercially available irregular ␤-Si3 N4 an ammonia/nitrogen mixed gas atmosphere. The obtained
powders were employed to mix in the epoxy matrix with spherical BN was reported to have an average diameter of
the same loading level. The composite with irregular ␤- 1–2 ␮m with smooth surface, and can be a good candidate
Si3 N4 powders showed more isotropic TC. The in-plane for highly thermally conductive composites [219]. Ohashi
and through-plane TC are 2.5 and 2.6 W/(m K), respectively. et al. [217] successfully prepared spherical AlN, and the
Unlike the irregular powders, the sample with ␤-Si3 N4 spherical AlN/epoxy showed better fluidity compared with
nanowires showed much higher TC in both directions composites with commercially available irregular AlN. The
with 9.2 W/(m K) in in-plane direction and 5.7 W/(m K) polymer composite with a high filler loading (74 vol.%)
in through-plane direction. It is obvious that TC became was successfully obtained by adjusting the filler size dis-
more anisotropic. Boron nitride nanotubes (BNNTs) are tribution to the bimodal one, and achieved a high TC of
also a promising candidate for various applications due 8.2 W/(m K), which is 9 times the TC of composite with
to their superb thermal and antioxidation stability, high spherical silica fillers (0.9 W/(m K)), and twice that of the
TC and excellent mechanical properties [213]. Metal wires composite with spherical alumina fillers (4.0 W/(m K)).
or fibers are another type of 1-D fillers. The most used is Besides those conventional shapes mentioned above,
silver nanowires (AgNWs) with large-aspect-ratios, which the effects of 11 artificially designed filler shapes, such as
would lead to a lower percolation threshold, and can double Y, Y, quad Y shaped, triangle, elliptical, square, rect-
be preferred fillers for thermal interface materials (TIMs) angular, and T shapes on the TC, were investigated by Wang
[214]. However, silver nanowires are very expensive, which et al. [220]. It was found that filler shapes have great impact
makes copper nanowires (CuNWs) more competitive as on the effective TC of the composite materials. The best
1-D metal fillers. Wang et al. reported that by using a single- fillers should be tree-like, which have a long heat conduc-
crystalline CuNWs with a high-aspect-ratio as thermally tion distance and large contact areas even when they are
conductive filler, only 0.9 vol.% loading of CuNWs in poly- placed unfavorably. From this respect, the double Y shaped
acrylate can achieve a TC of 2.46 W/(m K) [168]. fillers are the best, followed by Y shaped fillers.
7.2.2. Platelet-like fillers 7.3. Size effect

Platelet-shaped fillers are also characterized by high-
aspect-ratio and low percolation thresholds. As the Filler size is also important to the TC of composites.
thickness of many platelet-shaped fillers is quite small, Composites with small filler particles have large interfa-
they can be considered two dimensional (2-D) fillers. Dur- cial area, causing phonon scattering and hindering phonon
ing processing, such as injection molding, they tend to be transport, and often have a lower TC [221]. Composites
aligned parallel to one another, and offer a high degree with large fillers have less filler/polymer interface and thus
of contact in parallel direction. Also, the fillers often have lower thermal interfacial resistance, leading to improved
a high TC in the in-plane direction. These often lead to heat conduction. Wu et al. [222] studied the TC of graphite
anisotropic thermally conductive composites with much nanoplatelets/polyetherimide (PEI) system with different
higher TC in the flow direction than that perpendicular to filler sizes from 1 ␮m to 15 ␮m. It was found that the TC
the flow direction. of composites filled with smaller particles was lower, even
It should be mentioned that plate-like fillers are not though a better filler network was observed for the smaller
limited to those materials which have platelet molecular or particle system. This fact indicated that the interfacial resis-
crystal structure, such as BN or graphene or graphite. Other tance may be more important than filler connection, at least
ceramic or metallic particles, such as Al2 O3 , TiB2 and SiC, in this particular example. Zhou et al. [223] showed that the
may also possess plate-like morphology and were used as TC of silicone rubber filled with larger Al2 O3 was superior
thermally conductive fillers for polymer composites [215]. to that of the rubber filled with the smaller particles. Li et al.
[224] enhanced the TC of polyimide films with micro- and
7.2.3. Spherical fillers nano-sized boron nitride. The micro-sized filler gave higher
In general, filler loading needs to be high to obtain high TC than nano-sized fillers.
TC. However, high filler loading will cause a significant However, there are also controversial reports on the
increase in viscosity and reduce its processability. Find- effect of particle size, in particular when particle size is
ing the optimal composition window where the viscosity in the nanoscale. Some reports claimed that the inclu-
is still relatively low while thermal conductivity is suffi- sion of nanofillers in polymers can lead to the same or
ciently high is, then, the critical challenge. Spherical fillers higher TC compared to their micro-sized counterparts
offer some promise in this regard because their viscos- [170,225–227]. TC not only depends on the particle size,
ity percolation threshold is fairly high (∼50–60 vol.%) [45]. but also depends on the surface chemistry of particles, filler
Various spherical fillers including Al2 O3 , AlN, Si3 N4 , SiC, morphology in the composites, etc. It is hard to control the
etc. have been developed and used in thermally conduc- particle size to be the only independent variable, when the
tive composites [216–218]. Even for those with platelet particle size effect was studied. In fact, in many cases, the
crystal structure, such as BN, various methods have been nano particles were prepared in a different way from micro
developed for synthesizing spherical fillers that may permit particles. The nano particles may have different properties
Fig. 23. SEM image of (a) the microcomposite and (b) the nanocomposite. While in the microcomposite the silver fillers are not connected, a network
structure is revealed in the nanocomposites. [170], Copyright 2012. Reproduced with permission from AIP Publishing LLC.
and different surface chemistry, which leads to a different effect of filler size on the elastic modulus of epoxy/alumina
dispersion state in the composites. The contribution from trihydrate also showed that modulus decreased with
these parameters sometimes dominates the contribution increasing particle size at high filler loading levels [229].
from the reduced thermal interfacial resistance in micro Smaller calcium carbonate particles provided polypropyl-
composites. This might lead to higher TC observed in the ene composites with higher strength at a given particle
composites with smaller particle size. For example, Pashayi loading. Another big advantage that nano-sized fillers could
et al. [170] reported that epoxy-silver nanocomposites offer is the optical transparency. The composites with
had up to 50-fold higher TC than composites with silver nanoparticles are often transparent, provided the particles
micro particles at a similar volume fraction of fillers, even are well-dispersed in the matrix. For example, a trans-
though they have been processed under similar conditions. parent and highly heat-conductive composite film was
SEM images showed that the silver nanoparticles formed made by incorporating nanosized boron nitride nanotubes
a conductive network in nanocomposites while the micro- (BNNT) into polyvinyl formal (PVF) [230].
particles did not, as shown in Fig. 23. It is clear that the filler
network in nanocomposites is responsible for the higher 7.4. Hybrid fillers
TC. The network was formed via low temperature sinter-
ing of nanoparticle fillers during the curing of the epoxy, Hybrid fillers are commonly used in the thermally
as the silver nano particles have a lower melting point. conductive composites, as a better thermally conductive
The details will be discussed in Section 8.3.3. Kemaloglu network can be formed by using mixed fillers with differ-
et al. [226] found that the boron nitride/SEBS/EVA ternary ent sizes or types. Hybrid filler system can help to form
blends with smaller BN particles showed higher TC than large thermally conductive network by building bridges
that of the composites with larger particles at the same between fillers and maximizing the filler packing den-
filler content. These small particles are really the mixture sity. Another advantage of a hybrid filler system is that
of spherical BN particles and plate-like BN particles. There- it may help to significantly reduce the overall filler load-
fore, the higher TC cannot be totally ascribed to the size ing, thus reducing the system viscosity. A small amount
effect. Fu et al. [227] revealed that epoxy adhesives filled of CNT (1 vol.%) in AlN/epoxy system with 25 vol.% AlN
with nanosized Al2 O3 had higher TC than those filled with loading showed comparable TC to that of epoxy compos-
microsized ones. The authors acknowledged the nanopar- ite containing 50 vol.% AlN [231]. Hybrid filler system not
ticles are somewhat more polydisperse and better able to only can increase the packing density and lower the sys-
form network, thus the differences in TC cannot be totally tem viscosity, but also can improve through-plane TC for
attributed to the difference in particle size. some platelet filler composite systems, e.g. adding spheri-
Although nanoparticles have high interfacial area, and cal or irregular fillers to BN/polymer composites to disrupt
thus increased thermal interfacial resistance, they could the alignment of platelet fillers. Table 5 lists various hybrid
enhance TC significantly when combined with micro- filler systems reported in the literature.
particles. This has been demonstrated in several systems
[224,227]. This is because the nano-sized fillers acted as 7.4.1. Mixture of fillers with different sizes
bridges between micro-sized fillers to enhance thermal Mixing fillers with the same shape but different sizes
contact, while the micro-sized fillers could form the pri- was employed for the optimization of packing density of
mary thermally conductive path in composites. fillers. The mixture of large BN (0.4 ␮m) and small BN
It should be noted that nano-fillers often provide some (0.2 ␮m) gave a higher TC than those with a single parti-
other advantages over micro-fillers. The nanocomposites cle size at the same filler volume fraction [238]. A binary
often have superior mechanical properties. For example, mixture of AlN with small particles around 2 ␮m and
the silica nanocomposites had better mechanical proper- large AlN particles around 30 ␮m, was filled into epoxy
ties (higher modulus and strength) than composites with composite, and showed improved flowability and TC com-
their micron-sized counterparts [228]. The study on the pared to the composite with a single sized filler. When the
Table 5 polygonal AlN and platelet-shaped BN were added into

Various hybrid filler systems reported in the literature.
epoxy matrix to get the optimal TC [201]. It was found
Filler 1 Filler 2 Refs. that the packing efficiency is closely related to the diame-
h-BN Different sized h-BN [224,232]
ter ratio of the two filler particles. The optimal shape and
Irregular shaped Si3 N4 Irregular shaped SiO2 [233] composition of the AlN/BN composite systems were dis-
h-BN Carbon fibers [232] cussed (Fig. 25) and the maximum TC up to 8.0 W/(m K)
h-BN CNTs [232,234] was achieved in the AlN/BN hybrid system with a similar
Hollow glass Smaller-sized nitride [235]
particle size and 1/1 volume ratio of the two particles.
microspheres particles
AlN CNTs [231] A novel idea about hybrid filler system was to take
Nano sized SiC CNTs [236] advantage of 1-D fillers to create a 3-D interconnected ther-
Graphite Carbon fiber [237] mally conductive pathway throughout the polymer matrix.
Graphene BN [164] There are positive synergistic effects when combining 1D
Platelet h-BN ZnO whiskers [209]
Graphite nanoplatelets SiC microparticles [165]
and platelet-shaped fillers, as platelet-shaped fillers tend to
orientate in one direction, and most of them can only con-
tact with each other basically in the parallel direction. 1-D
fillers are more flexible and can serve as bridges to connect
platelet-shaped fillers in different layers to construct a 3-
D filler network. It was found that a significant synergistic
improvement on TC of the BN/polyphenylene sulfide (PPS)
composite has been achieved by adding a small amount of
CNT to form a hybrid filler system, as shown in Fig. 26 [234].
7.5. Surface treatment
As inorganic fillers are dispersed in organic polymer

matrix, numerous filler/polymer interfaces are created.
The interfacial thermal resistance between polymer matrix
and fillers is one of the main reasons for limited ther-
mal conductivity improvement in heat conductive polymer
composites (see also Section 2 for discussion on funda-
Fig. 24. The concept of using particles with different sizes resulted in
mentals and computer simulations of interfacial thermal
more efficient packing of fillers. [240], Copyright 2013. Reproduced with
permission from Elsevier Ltd.
resistance). The interfacial thermal resistance can be clas-
sified in two ways: (1) by thermal contact resistance and
(2) by thermal boundary resistance. Thermal contact resis-
fraction of small particles in the binary mixture of the AlN tance originates from mismatches in the surface conditions
was around 0.2–0.3, the maximum improvement on the TC of two solid bodies. Thermal boundary resistance is due
was observed [239] (Fig. 24). to the mismatch in the vibrational harmonics of energy
carriers (i.e. phonon spectra) and/or the flow of electronic
7.4.2. Mixture of fillers with different shapes energy carriers between the materials [241]. Thermal con-
Fillers with same type, but with different shape and size tact resistance depends on the contact area and how well
have also been blended to improve TC. Larger platelet fillers the filler and polymer matrix are bonded together. Sur-
were combined with smaller spherical fillers to get positive face treatment of fillers is widely used to reduce the
synergistic effect. Spherical BN agglomerates (BN-PTX60) thermal interfacial resistance and improve the thermal
and platelet BN (BN-PT110) were added to PPS in a melt- conductivity. However, it must be mentioned that surface
blending process [232]. The smaller PTX60 fit between functionalization may not always lead to improved thermal
larger PT110 platelets to create interconnectivity through- conductivity, if it also leads to the formation of defects. For
out the polymer matrix, and increased the effective TC of example, carbon nanotubes were treated with strong acid
the composites. to introduce carboxylic group onto the surface of carbon
nanotubes. The functionalization increased the interface
7.4.3. Mixture of different types of fillers coupling, but also led to the formation of defects, which
Different types of fillers have also been combined to impede the acoustic phonon transport in the carbon nano-
get desired TC, processability and other properties. The tubes [242].
hybrid particles usually have different shapes and sizes. A Various surface modifiers have been used to mod-
combination of irregular shaped Al2 O3 and AlN with dif- ifying filler surface to decrease thermal interfacial
ferent sizes was added into polymer matrix. Two cases resistance, including surfactants [243], coupling agents,
of large-sized AlN combined with small-sized Al2 O3 , and e.g. organo-silanes and titanates [63,236,238,244–259],
large-sized Al2 O3 combined with small-sized AlN were functional polymers [260–263], inorganic coatings
studied. At total filler loading of 58.4 vol.% and the volume [264–277]. Organo-silanes with a general formula of
ratio of large particles/small particles was 7/3, a maximum (RO)3 –Si–(CH2 )n –Y (where Y is a polymer reactive group)
TC of 3.402 W/(m K) and 2.842 W/(m K) has been achieved are the most popular and effective surface modifiers.
for the two systems, respectively [240]. The mixture of The trialkoxysilane group can hydrolyze resulting in
Fig. 25. Schematic of the prepared composite systems with the morphologies of the different filler sizes, and the bimodal distribution characteristics with
relative compositions of 2:1, 1:1, and 1:2. [201], Copyright 2012. Reproduced with permission from Elsevier Ltd.
conductivity. Above the critical concentration, however,

covalent bonding results in less effective improvement of
thermal conductivity in the composites. The critical filler
concentration is interpreted as the percolation threshold
of fillers in polymer matrix. The covalent bonding between
the particles and the matrix is important to increase the
thermal conductivity of the composites with filler loading
lower than the critical concentration. Direct contacts
among neighboring particles, however, give higher ther-
mal conductivity enhancement in the composites where
particle loading exceeds the critical concentration [244].
It should be mentioned that silane treatment also
helped to improve the dispersion of filler in polymer and
reduce viscosity during processing. To improve the disper-
Fig. 26. Schematic diagram of the BN/CNTs filler network in the polymer
sion of AlN in epoxy, Lee et al. used a two-layer surfactant
matrix. [234], Copyright 2012. Reproduced with permission from Elsevier
Ltd. approach with silane as the first layer and phosphate ester
as the second layer dispersant [278]. Hong et al. [279]
the formation of silanol groups (Si–OH) and then con- compared the TC of poly(dimethyl siloxane) (PDMS)/CNTs
dense with filler surface hydroxyls to form permanent composites prepared with raw CNT and surface modified
attachments, while polymer reactive group can form a CNT. Better dispersion and 10% increase on TC of compos-
chemical link to the polymer matrix. Thus, the adhesion ites were observed after modifications. It was elucidated
between polymers and fillers is improved and interfacial that the thermal conductivity enhancement originating
scattering of phonons is reduced. The thermal conductivity from the interface improvement could be significantly off-
improvement depends on the filler surface chemistry, set by the defects (e.g., voids and pores). Voids and pores
polymer type, silane chemistry, and amount of silane. can be reduced by a certain surface treatment technique,
Research results suggested that there was an optimum i.e., functionalization of AlN by mercapto-terminated silane
amount of silane to be used; too little or too much of for epoxy/AlN composites [244].
silane would not lower the interfacial resistance [253]. Silane treatment not only can improve TC, but also
The thermal conductivity enhancement of the composites can protect some moisture sensitive fillers, such as AlN.
due to the improved interface also depends on the filler Metal oxide has numerous –OH group on the surface and
loading. Below the critical concentration the covalent can be easily modified. It is much easier to be incor-
bonding tends to result in higher enhancement of thermal porated in polymer matrix even without modification,
Fig. 27. Schematic diagram illustrating process of as-grown BNNTs transformation to POSS fuctionalized BNNTs. [280], Copyright 2013. Reproduced with
permission from John Wiley and Sons.
comparing with nitride or carbide ceramic fillers. Among more effective in improving the interfacial adhesion than
those ceramic fillers, hexagonal BN is most difficult to the more conventional silane treatment [260]. Glycidyl
be surface-functionalized because of its layered hexago- methacrylate-butyl acrylate-maleic anhydride (GBM) ter-
nal crystal structure. The platelet 001 plane is molecularly polymers with different molecular weights were used
smooth and there are no surface functional groups avail- to treat aluminum nitride (AlN). It was shown that the
able for chemical modifications. The only spot that can be AlN treated with GBM gave higher thermal conductiv-
used for grafting function groups is at the edge planes of the ity in cyanate ester composites than those modified with
platelets. The functional groups can be covalently bonded silane [261]. Polyimide (PI)-modified AlN fillers referred
to the boron atoms, and help h-BN to be well dispersed in as AlN@PI were reported and used for electronic encap-
polymer matrix and improve the interface adhesion. sulation in high performance polymer composites [181].
Polymers have also been used to modify fillers, as dis- These AlN@PI/epoxy composites exhibited better thermal
cussed by many researchers. These modified fillers have and dielectric properties than the untreated AlN composite.
a core–shell structure that exhibit enhanced physical and At 40 wt.% of filler loading, the TC of AlN@PI/epoxy compos-
chemical properties that are associated with shell surface ite reached 2.03 W/(m K). As shown in Fig. 27, polyhedral
and encapsulated core materials. The core is usually sur- oligosilsesquioxane (POSS) encapsulated BN nanotubes
rounded by a homogeneous spherical shell containing a were successfully achieved and were demonstrated to be
thermally different material. These core–shell structured very effective nanofillers for making composites with high
fillers are advantageous due to the improvements on ther- TC [280].
mal and/or mechanical properties, modifications to the Not only polymers can be coated on the surface of
surface properties of the fillers, and improvements in inorganic filler, but also inorganic material can be coated
the dispersion of the fillers in the binder matrix. Admi- to form a core–shell structure. Core/shell structure of
cellar polymerization was used to coat polystyrene and metal/metal oxide, metal/carbon, metal/ceramic have been
polymethyl methacrylate on the BN surface to improve reported. For many applications, thermally conductive
the interfacial adhesion in the BN-filled epoxy compos- composites need to be electrically insulating. Metal parti-
ite. This admicellar surface treatment was found to be cle have high TC. However, they are electrically conductive,
The shell materials do not necessarily have to be elec-

trically insulating. Yu et al. [270] demonstrated a facile
synthesis of carbon fiber–copper core–shell structured
filler. Copper shells were coated on carbon fibers through
electroplating method and post-treated via rapid thermal
annealing technique, as shown in Fig. 29. The epoxy/Cu-
plated carbon fiber composites with Cu shell of 12.0 vol.%>
prepared via compression molding, had 18 times larger in-
plane TC (47.2 W/(m K)) and 6 times larger through-plane
TC (3.9 W/(m K)) than epoxy/carbon fiber composite. This
can be attributed to the Cu shell with a high degree of
crystal perfection providing continuous heat conduction
pathway.
Mesoporous materials can also be employed as shell
to construct a highly thermally conductive pathway.
Fig. 28. Schematic illustration depicting the formation process of
Core/shell fillers consisting of FeCr metal core and meso-
the Al@Al2 O3 @SiO2 nanoparticles and multilayer core–shell (AlF-
Al)@Al2 O3 @SiO2 fibers. [266], Copyright 2012. Reproduced with permis- porous Al2 O3 shell particles were fabricated [271]. The
sion from AIP Publishing LLC. composite with the core–shell structured fillers showed
a highly enhanced TC compared with that with typical
untreated FeCr particles. Because Al2 O3 had a lower TC than
and cannot be used in such applications. Metallic fillers metal core, the improved TC was considered to be the result
can be made electrically insulating after being coated of better thermal conductance through the shell layer from
with an insulating layer. With this strategy, heat conduc- the core particle to the continuous polymer medium, as
tive and electrically insulating polymer composites can shown in Fig. 30.
be prepared. Silica-coated silver nanowires (AgNWs@SiO2 ) Beside metal oxides, nitride materials can also be used
were intentionally designed and synthesized by a sol–gel as a shell. Various methods have been developed to coat BN
method and then incorporated into epoxy matrix [264]. on the surface of carbon fibers, CNTs, SiO2 , GaN nanorods
The silica outer layer not only helps to disperse AgNWs [272]. For example, H3 BO3 –urea mixture was used as pre-
in epoxy with enhanced interfacial interaction, but also cursors. Upon heating process under N2 protection, H3 BO3
improves the TC of the composites. Experimental data decomposed to B2 O3 , which reacts with NH3 produced by
showed that TC of an epoxy/AgNWs@SiO2 composite with the decomposition of urea to form highly thermally con-
4 vol.% filler loading was increased to 1.03 W/(m K) com- ductive BN out layer [273,274].
pared to 0.57 W/(m K) of an epoxy/AgNWs composite with
identical nanowire loading. In the meanwhile, the insulat-
ing silica nanolayer effectively avoided the formation of an 8. Microstructure control of heat conductive
electrically conductive network of AgNWs in epoxy, lead- polymer composites
ing to a highly electrical insulating composite. Al2 O3 coated
spherical Al particles [265] and SiO2 coated Al particles or 8.1. Filler orientation
fibers [266] have also been proposed as thermally conduc-
tive fillers, as shown in Fig. 28. With similar concept, SiC Many heat conductive fillers have anisotropic TC. These
particles are thermally coated with a nano layer of SiO2 fillers have non-spherical shape, such as plate-like boron
by oxidation at high temperatures [267]. Silane was used nitride, fibers, carbon nanotubes, graphene sheets, graphite
to further treat the surface of silica coated SiC particles to flakes, wires. These non-spherical fillers can be oriented
improve the interaction between filler and polymer matrix. during processing, which usually leads to high TC in the
Choi et al. reported a synthesis of SiO2 coated graphite by a orientation direction, but low TC in the direction perpen-
polyvinylpyrrolidone (PVP)-assisted sol–gel reaction [268]. dicular to the orientation direction. Anisotropic TC due to
They found that both in-plane and through-plane TC values the orientation of fillers has been observed in the com-
of silica-coated graphite/TPEE composites were improved posites with boron nitride [163,281–284], carbon nanotube
comparing with those of raw graphite/TPEE composite. [285–288], carbon fibers [237,289–293], Al2 O3 fiber [294],
Zhao et al. [269] reported a synthesis of SiO2 coated carbon stainless-steel fiber [295], Hemp fiber [296], glass fibers
nanotubes (SiO2@CNTs) with different coating thickness [293], silicon nitride nanowire [212], graphite [297,298],
by a sol–gel method. SiO2 @CNTs was compounded with graphene [299], Ni platelets [300], and diamond [301].
polyurethane (PU) to prepare PU/SiO2 @CNT composites. Some applications require composites with isotropic TC,
The effect of SiO2 @MWCNTs on the TC and electrical insu- while in other applications high TC in a specific direction is
lation properties of PU was investigated and the results more desired. However, in many cases, the preferred direc-
showed that silica layer improved the interfacial interaction to have high TC differs from the direction in which the
tion between MWCNTs and PU, enhanced the dispersion of composites have a high TC. For example, in many cases,
MWCNTs in the PU matrix, and improved the TC of com- the heat conductive fillers have been oriented in the flow
posite. Due to the electrical insulation of the SiO2 outer direction, which led to a high TC in in-plane direction,
layer, PU/SiO2 @CNT composites had almost similar electri- while a high TC in through-plane direction is more desired.
cal resistivity to pure PU. Therefore, randomizing filler orientation or using specific
Fig. 29. (a) Schematic of fabrication process of epoxy/Cu-plated CF composite and cross-view SEM micrographs of (b) pristine CF, (c) Cu-plated CF, (d)
epoxy/Cu-plated CF composite and side-view SEM micrograph (e) epoxy/Cu-plated CF composite. Arrows in part d indicate the epoxy matrix, Cu shell, and
carbon fiber core. [270], Copyright 2014. Reproduced with permission from the American Chemical Society.
orientation approaches to orient fillers into the desired

direction is needed.
8.1.1. Orientation by shear or extensional flow during

processing
Orientation of fillers can result from the shear or stretch-
ing force during various methods of processing, such as
injection molding [237,294,295], extrusion [288], com-
pression molding [212], tape casting [163,298], foaming
[283], solid state shear milling [297], multistage stretching
[282], electrospinning [281,285].
8.1.2. Orientation by external fields

Orientation of fillers can be also obtained by an external
field, such as magnetic and electrical field. Solis and Martin
[300] used magnetic fields to both orient and agglomerate
Ni spheres and platelets into structures that can efficiently
Fig. 30. The improvement of thermal conductivity of the composite con- conduct heat. Field structuring increased the TC substan-
sisting of core–shell particles and untreated uni-modal structure. [271], tially along the applied field without a significant reduction
Copyright 2015. Reproduced with permission from Elsevier Ltd. in the transverse direction to the field. The most important
advantage of field structuring is that without increasing the

viscosity of filled resin, one could increase the composite
TC almost by a factor of three.
Yan et al. [299] used magnetic field to align graphene
nanosheets (GNSs) in epoxy composites. To serve
this purpose, the nanosheet surface was modified by
Fe3 O4 through a co-precipitation method. The resulting
epoxy/GNS–Fe3 O4 composites possessed high TC in a
parallel magnetic-alignment direction at low GNS–Fe3 O4
loadings, which greatly outperformed the composites with
randomly dispersed bare GNSs. Fig. 31. Sketch of boron nitride agglomerates.
Electric field has also been used to orient thermally con-
ductive fillers. Linear assemblies of BN nanosheets (LABNs)
were fabricated in polysiloxane/BN nanosheet composite structure of the individual MWCNTs are the dominating
film under a high DC electric field. The stretched structure factors that control the overall TC of the composites.
induced effective heat conduction through the out-of-
plane direction of the composite film [302]. This strategy 8.2. Filler agglomerates
was further advanced to assemble BN nanosheets into a
LABN structure with extended length and denser popula- Plate-like fillers, such as boron nitride and graphite,
tion in polymer based composites using microscopic molds have anisotropic TC. The TC in the crystal plane is different
[284]. The relocation of diamond fillers was performed from the through-plane TC. For example, the in-plane TC
in a prepolymer mixture of polysiloxane under different of hexagonal born nitride has been estimated to be higher
electric fields AC, DC, and switched DC, before the mix- than 300 W/(m K), while the through-plane TC is only about
ture became cross-linked. In the cured composite film, the 3 W/(m K). When BN powders are used as fillers, the com-
self-assemblies of linearly aligned diamond fillers (LADFs) posite materials often have anisotropic properties, largely
connect the film planes as bridges. Composites exhibited determined by the orientation of the BN platelet in the final
enhanced TC and electrical insulation [301]. part. To overcome this problem, boron nitride agglomer-
ates have been developed with/without binders (as shown
in Fig. 31). In such agglomerates, plate-like BN crystals are
8.1.3. Orientation of carbon nanotube by chemical vapor held together to form a larger particle to randomize their
deposition orientation. Using BN agglomerates, more isotropic TC has
Vertically aligned multiwalled carbon nanotube (VA- been obtained. This has been demonstrated in HDPE [305],
MWCNT) arrays have generated significant interest in silicone [216], polyimide [306], polycarbonate, NylonTM 6
recent years. Aligned multi-walled carbon nanotubes and NylonTM 66 [307].
(MWCNTs) were made by chemical vapor deposition. By These agglomerates were produced via secondary
placing CNTs vertically and spanning the system com- processing of BN platelets. A process of spray drying an
ponents, one can reduce the number of CNT/polymer aqueous slurry comprising non-spherical shaped boron
interfaces in the through-thickness direction, thus mini- nitride particles, water, an organic binder and a base for
mizing the effective thermal interfacial resistance in the maintaining the pH of the slurry basic was developed
direction of heat flow. to make the agglomerates [308]. The agglomeration of
A TC of 3.8 W/(m K) in a VA MWCNT–polymer composite the boron nitride primary particles can be also made by
along the alignment direction with 2% volume fraction of compacting the plate shaped, hexagonal boron nitride
MWCNTs was reported by Borca-Tasciuc et al. [303]. Ivanov between two rotating opposite rollers. The rollers were
et al. [304] reported a TC of 5.5 ± 0.7 W/(m K) for an epoxy- pressed against each other with a defined pressure. The
infiltrated VA-MWCNT array (8 ± 1 vol.%, 2 mm long). Kaul compaction step was carried out many times. A thermal
et al. [286] estimated the TC of their CNT epoxy com- treatment was carried out as an additional step preferably
posites to be about 5.8 W/(m K); TC also exhibited slight at 1600–1950 ◦ C, after the compaction step [309]. High-
increase with temperature in the range of 240–300 K. It density agglomerated particles of hexagonal boron nitride
was suggested that through-thickness thermal conductiv- can be also formed by a cold forming process carried out
ities as high as 25 W/(m K) can be potentially fabricated, at ambient temperature comprising the following steps:
as the TC of a VA-MWCNT array can be increased by an crushing agglomerated chunks of relatively large size high
order of magnitude by using a standard high-temperature purity hexagonal boron nitride into boron nitride particles,
post-annealing step. Sihn et al. [287] incorporated aligned cold pressing the crushed particles into a compacted form,
multi-walled carbon nanotubes (MWCNTs) in the adhe- granulating the compacted form into a granulated powder.
sive layer to enhance the through-thickness TC in the The steps of cold pressing and crushing can be repeated
adhesively bonded joints. The measured value of the several times until desired density has been achieved [310].
through-thickness TC of the modified adhesive joint with One of the biggest challenges of using BN agglomer-
the MWNCT was over 250 W/(m K), which was greater than ates is to preserve the structure through all the processing
the TC of neat adhesive joint by several orders of magni- steps. BN agglomerates are relatively weak and are likely
tudes. These studies suggest that the thermal conductance to break down into their component platelet crystals if
at CNT-capping layer interfaces and the morphological sheared aggressively during processing. In particular for
Fig. 32. The PS composites with Cu shell networks revealed 60 times larger thermal conductivity than conventional simple melt-mixed Cu beads/PS
composite. [313], Copyright 2013. Reproduced with permission from the American Chemical Society.
thermoplastic composites, the BN agglomerates can break have the highest TC, while the melt mixing gave the lowest
down during the extrusion step and/or the molding step. TC. It is obvious that the TC of a composite was significantly
Both the screw configuration during extrusion and the affected by the dispersion state in the composite. Powder
flow field during molding determine the extent of shear mixing can lead to a more connected filler network, while
exerted on BN agglomerates. These processing steps should in melt mixing, fillers were dispersed in isolated or passi-
be monitored and controlled to preserve the agglomerates’ vated phase with thermally insulating polymer matrix. A
beneficial structure. similar approach was used to make polymer composites
with Cu thin film network [313]. The polymer compos-
8.3. Formation of continuous filler network ites were created by the Cu metallization of PS beads and
then hot press molding of Cu-plated PS beads. The unique
As discussed earlier, the formation of the continuous three-dimensional Cu shell-networks in the PS matrix were
percolation structure of the fillers with large TC is the key to responsible for isotropic and ideal conductive performance
achieve high TC for the composites. The high filler content at even extremely low Cu contents, as shown in Fig. 32.
of above 60–70 vol.% could lead to a continuous heat con-
duction path of fillers in the polymer composite. However,
8.3.2. In situ polymerization in porous heat conductive
high filler loading makes processing difficult and increases
filler network
the cost. Various approaches have been used to form heat
The heat conductive network can also be formed by infil-
conductive path at a lower filler loading level.
tration of monomer into a porous heat conductive filler
body with a percolated pore structure. In situ polymeriza-
8.3.1. Molding of the mixture of filler and polymer tion occurs in the porous ceramic upon heating up, and
powders the final morphology of the composite consists of an inter-
One way to achieve heat conductive path is mixing filler penetrating polymer network which fills in an AlN ceramic
particles and polymer particles together and then compres- skeleton. This special interpenetrating network gave rela-
sion molding the mixture to get a special filler dispersion tively high TC [199].
state in which the polymer particles are surrounded by
heat conductive filler particles. The segregated fillers in
the boundary between polymer particles can form heat 8.3.3. Self-assembly process
conductive paths at low filler content. This strategy [311] Self-assembly process can also be used to construct
has been demonstrated in the Si3 N4 particle filled epoxy heat conductive metallic network. The self-constructed
composite. Using particulate epoxy resin and molding pro- networks in the polymer matrix are formed via low tem-
cess, a continuous network was formed by the filler in the perature sintering of nanoparticle fillers during the curing
composites. Heat conduction is mainly performed by the of the polymer. Nanoparticle fillers are desirable since they
filler, resulting in a high overall TC. This strategy has also can be sintered together at lower temperatures than those
been used to prepare AlN filled PS [198]. PS and AlN were for microparticles. High heat conductive epoxy–silver com-
mixed at room temperature, then compression molded. posites have been obtained through this process, whose
It was found that the TC of composites was higher for a heat conductivity is up to ∼50 times higher than that
PS particle size of 2 mm than that for a particle size of of silver microcomposites at a similar volume fraction of
0.15 mm. Four mixing methods, powder mixing, solution fillers [170,314]. The networks form through a three-step
mixing, roll-milled mixing and melt mixing, were used to process comprising of self-assembly by diffusion limited
make PE and graphite composites [312]. It was found that aggregation of polyvinylpyrrolidone (PVP) coated silver
the powder mixing from which the specimens were cast nanoparticles, removal of PVP coating from the surface,
Fig. 34. (a) Thermal conductivity of hybrid films: (䊉) PI blends and ()
Homo-PIs. The lines are the values calculated for an ideal PI blend model
(rigid line) and a Homo-PI model (broken line) based on the Bruggeman
theory. (b) Schematic representations for ideal model structures for PI
blends with VDP structures (left) and for homo-PIs with homogeneously
dispersed ZnO-NPs (right). [321], Copyright 2011. Reproduced with per-
mission from the Society of Chemistry.
blends. The TC of the blends were significantly improved

Fig. 33. The mechanism for the formation of the self-constructed metallic by selectively localizing SiC nanoparticles [318] or hybrid
network in the nanocomposite and its synergy with its flow and thermal fillers of multi-walled carbon nanotube (MWCNTs) and sil-
properties: (a) high k nanoparticles functionalized with PVP and dispersed icon carbide (SiC) nano particles in the PVDF phase [319].
in uncured epoxy form a high solids volume fraction slurry able to flow Double percolation approach has been used to make
and fill the roughness of two mating surfaces; (b) the molecular coating is
removed by low temperature processing and the metallic nanoparticles
composites with high heat dissipation, sharp and large
start to agglomerate and sinter into high k networks of nanorods; (c) the amplitude PTC effect and independent adjustability of elec-
high network at the end of the process with the cured polymer matrix pro- trical conductivity and TC. It is a co-continuous structure
viding compliant mechanical support. [170], Copyright 2012. Reproduced associating a thermally conductive polymer phase (syn-
with permission from AIP Publishing LLC.
diotactic poly(styrene) (sPS) filled with aluminum oxide
(Al2 O3 ) or boron nitride (BN)) with an electrically con-
and sintering of silver nanoparticles in high aspect ratio ductive polymer phase (high-density polyethylene)(HDPE)
networked structures, as shown in Fig. 33. filled with carbon nanoparticles (CNP) in appropriate pro-
portions [320].
8.3.4. Double percolation The “vertical double percolation (VDP)” morphology
Double percolation approach has been used to lower the were spontaneously formed in the PI blend films composed
percolation threshold of electrical conductivity in electrical of a sulfur- and a fluorine-containing PI and ZnO nano par-
conductive polymer composites [315–317]. However, only ticles [321], as shown in Fig. 34. The two phases were
recently this approach has been used to enhance TC. For separately aligned along the out-of-plane direction, and
this technique, two-phase immiscible polymer blends, in ZnO-NPs were preferentially precipitated in the fluorine-
which at least one phase is continuous, can be used as a containing PI phase. At 27 vol.% of ZnO nano particles, the
polymer matrix, and conducting fillers are selectively and blend film exhibited 410% enhancement of TC, whereas
preferentially localized and percolated in the continuous only 90% enhancement was observed in the monophase PI
phase. In this way, the TC can be increased more at the film containing homogeneously dispersed ZnO-NPs.
same filler loading level or less filler is needed to achieve The double percolation concept has also been used to
the same TC. increase the TC of epoxy molding compound (EMC). An
A double percolation structure has been successfully island-sea structure has been designed by filling epoxy
constructed to effectively reduce the volume fraction of resin with heat conductive Si3 N4 particles and precured
fillers in poly(vinylidene fluoride) (PVDF)/polystyrene(PS) epoxy-SiO2 composite particles obtained by crushing the
cured epoxy-SiO2 composites [233]. As Si3 N4 particles can- processing. Moreover, aluminum based heat sink can add
not enter the precured epoxy composite particles, they will substantially to the weight of a fixture or device. This might
be confined in the epoxy-sea phase. The composite gave have negative effects on both new designs and retrofit
higher TC than the composites with homogeneously dis- installations. The trend in heat sink design is to develop
tributed Si3 N4 . materials with less weight, lower cost and adequate ther-
mal dissipation performance. Heat conductive polymer
9. Applications of thermally conductive polymer composites could be an alternative solution to this. Poly-
composites mer composites are steadily gaining more market share
for heat sink application due to their good processability,
Nowadays, with the increasing cooling demand in lightweight, design flexibility, corrosion resistance, lower
emerging industries, new thermally conductive materi- processing cost, etc.
als are in high demand. Compared with other thermally Placing chip directly on the heat sink is a new direction
conductive materials, such as metals, ceramics, carbon for LED thermal management system development. In this
materials, polymer composites have following features: case, electrical insulation is required for heat sink materials.
Ceramic filled polymer composites are the main selection.
(a) Electrical insulation and electrical conduction can be In general, a TC higher than 10 W/(m K) is needed for this
controlled via selecting appropriate fillers. application. To achieve high TC, a sufficiently high filler
(b) Easy processing for integrated parts or complex geom- loading (>40 vol.%) is needed. At this high filler loading,
etry. the polymer composites tend to have poor processability
(c) Light weight. and higher overall cost. How to reduce filler loading and
(d) Corrosion resistant. keep the same TC is one significant challenge for electri-
(e) Compliance to the geometry of the adjacent rough sur- cally insulating polymer composites. Besides, the geometry
faces, if flexible polymer is used. of heat sink could also influence the heat dissipation effi-
(f) Vibration damping due to resilience of polymer com- ciency at natural convection condition [327–329]. A final
posites. heat management solution for LED heat sink could be
an appropriate thermally conductive polymer composites
This section lists some emerging application areas of combined with suitable design.
thermally conductive polymer composites: LED devices,
electronic assembly and packaging, battery and solar. 9.2. Electronic packaging
9.1. LED devices Higher power density and miniaturization have made
thermal management of consumer electronic devices one
Compared with conventional lighting, light-emitting of the major challenges for the semiconductor industry
diode (LED) lighting has the advantages of a longer lifetime [330]. The objective of thermal management in electron-
and higher energy efficiency (30%). However, 70% energy in ics is to ensure that temperatures of all components in the
LED devices is transformed into heat. Such large quantity of system are maintained within their functional temperature
heat can significantly influence the LED’s performance, in range. Thus, there should be effective thermally conduc-
terms of luminous efficiency and lifetime. At constant oper- tive path for heat dissipation and eliminating hot spots
ation current, the luminous efficiency decreases by about in electronic parts [1]. The discussion on the applications
5% and the lifetime of LED decreased by half for every 10 ◦ C of thermally conductive polymer composites in electronic
rise in temperature. Therefore, a good thermal manage- assembly and packaging will be focused on three areas with
ment system for dissipating the heat from inside of the LED pressing thermal management needs: single chip packages,
package to the surroundings is very important. 3D chip stack packages and automotive electronic control
A conventional LED thermal management system is units (ECU).
composed of a heat spreader, thermal interface material
(TIM) and heat sink. Copper and aluminum alloys are typ- 9.2.1. Single chip packages
ically used for heat spreader. Silicon based grease, gels, The International Technology Roadmap for Semi-
phase change materials (PCM) have been used as TIM. conductors (ITRS) projected power density and junction-to
There are continuous needs on increasing the TC of TIM or ambient thermal resistance for high-performance chips
decreasing the thermal interfacial resistance of TIM. Some at the 14 nm technology generation to be >100 W/cm2
recent progress has been made on using a soft matrix mixed and <0.2 ◦ C/W, respectively [330]. The main bottlenecks in
with CNTs and/or nanofibers for TIM application [322]. In reducing the junction-to-ambient thermal resistance are
order for TIM to achieve higher TC, the promising approach the thermal resistances of the thermal interface material
is to align CNT perpendicularly with the contact surface (TIM) and the heat sink. There is a need for TIM that pro-
[323–326]. vides the highest possible TC during chip operation, good
Aluminum alloys with high TC between 150 and adhesion, and conformability to fill the gaps between two
300 W/(m K) are typically used as heat sink material. rough surfaces.
Although metal based heat sink has excellent TC and rel- The TC of TIM is relatively low, about 3–5 W/(m K).
atively low raw material cost, the processing cost is high Increasing TC of electrically insulating polymer composites
and electrical conductivity is certain. Cost analysis indi- for TIM application is one challenge of single chip packag-
cated that the majority of the heat sink cost (>80%) is from ing. As discussed in the earlier section, carbon-nanotubes
(CNT) or carbon nano fibers (CNF) [331–336] are potential the central problems in their growing use. The heat gen-
emerging materials used in TIMs that showed significant eration rate can exceed heat dissipation rate when cells
improvement in heat conductivity. An insulating layer like are used at high discharge/charge rate or at high ambient
mica could be used to provide electrical insulation when temperature. The cell may experience thermal runaway
CNT is employed as TIM [337]. accompanied by internal pressure buildup, which could
lead to bursting of the cell. Because of this safety issue,
9.2.2. 3D chip stack assembly and packaging highly thermally conductive materials are needed. Car-
3D chip stack technology with through silicon via bon nanotube/fiber filled PVDF [343] with higher TC is
(TSV) interconnect can be used to enhance communica- a good option to replace conventional electrode material
tion between ICs. However, the cooling of a stack of logic which contain PVDF and carbon-black in Li-ion cells [344].
chips becomes more complex as compared to the single TC of polyacrylonitrile (PAN) nanofiber separators could
chip case. In a single chip, heat can be readily accessed and increase from 3.5 W/(m K) to 8.5 W/(m K) by only adding
removed from the back side (bulk silicon side) of the die. 8 wt.% graphene nanoflakes [345].
In a 3D chip stack, there is no direct access to the back Electric vehicles (EVs) and hybrid electric vehicles
side of the chips in the stack or a method of spreading (HEVs) have many batteries which are combined together
the heat laterally from the stack. Because of the increased in a pack. Uneven temperature distribution in a pack could
power density and various low-TC materials in the stack, break electrical balance and reduce power output. For this
challenge of cooling becomes magnified. Increasing the TC reason, it is important to reduce the temperature variations
of the chip-to-chip interfaces within the stack would be among cells. A thermal management system is needed to
beneficial in 3D thermal management. High thermally con- maintain a battery pack at an optimum average temper-
ductive underfill material is critical to the heat dissipation ature. In this case, phase changing material and coolants
of 3D chip stack packaging [185,338]. For this application, can be used for heat dissipation of battery pack [346,347].
underfill not only needs high TC, but also needs to have low However, they are not very realistic to be applied to real
viscosity, good wetting to multiple surfaces, low shrink- systems due to the lack of cost effectiveness and complexity
age during cure, and low CTE (10–14 ppm/◦ C) [339]. TC can of structures. Thermally conductive polymer composites
always be increased by increasing filler loading. However, may open a new window for thermal system design of
this causes underfill materials to have poor flowability, battery pack.
which is critical to the capillary underfill application. How
to increase TC, but maintain good processability is still a
major challenge. 9.4. Solar
9.2.3. Automotive electronic control units (ECUs) Solar as a renewable and sustainable energy is growing
The ECU plays the crucial role of controlling and rapidly in recent years. Because of the low conversion effi-
integrating different complex actions such as mixture for- ciency of photovoltaic (PV) cells, most of the absorbed solar
mation, combustion and exhaust gas treatment. In order to energy converts into thermal energy within the cell [348].
reduce the pollutant emissions and achieve the require- It is well known that the solar conversion efficiency of the
ments of Euro 6 standards, it is expected that ECU will PV cells decreased with the rise of its operating tempera-
have a greater functionality, a reduced size and weight, ture. Increasing the temperature of PV cells by 1 ◦ C leads
and an increased power density. Thus, the challenge of to about 0.4–0.5% reduction of the relative conversion effi-
thermal management within the chip and surrounding ciency for the crystalline silicon based cells [349,350] and
packaging is greater than ever. The die-level power dis- around 0.25% for the amorphous silicon cells [351]. In order
sipation density is projected to exceed 100 W/cm2 in the to increase conversion efficiency of PVs, it is desired to
next 6 years [340]. Another concern is the impact of harsh remove the accumulated heat from the PV surface, espe-
under-hood automotive ambient temperatures on elec- cially for concentrated PV system [352]. How to conduct
tronics used in ECU assembly. The performance of thermal heat efficiently from cell to outside is a key challenge.
grease, thermal pads and phase change materials (PCMs) EVA as the typical cell encapsulating materials has very
may degrade when exposed to elevated temperature over low TC (0.23 W/(m K)). Besides, a typical backsheet compo-
a period of time [341,342]. Silicone polymer composites nent material DuPont TedlarTM (poly(vinylfluoride) film)
are good option for TIM in ECU because of low modulus of also has a low TC (0.36 W/(m K)). These two low thermally
elasticity, good wetting characteristics, good thermal sta- conductive layers are the bottlenecks for heat dissipation
bility and excellent vibration damping capability [335,337]. from cell. Using electrical insulating fillers to enhance the
However, the thermal resistance of gels is in the range of TC of EVA [353] and backsheet could be an effective way to
0.4–0.8 (K cm2 )/W which is still considered high for this improve the heat dissipation from cells. However, the cost
application. Increasing TC and maintaining other proper- and impact of fillers on other properties, such as adhesive,
ties in harsh environment is the main challenge for polymer barrier properties, should also be considered.
composites in ECU application.
9.3. Batteries 10. Conclusions and future prospects
As batteries become more powerful and utilized in more In this review, we discussed the research progress
diverse applications, thermal management becomes one of on thermally conductive polymer composites and their
potential applications. Based on these studies, we can for- Many other techniques, such as self-assembly process,
mulate several key messages. are being developed to form heat conductive pathways
at a lower filler loading level.
(1) Thermal conductivity of polymer composites is a func- (9) The orientation of fillers with anisotropic TC can lead
tion of the thermal conductivities of polymer matrix to the composites with anisotropic TC. Filler orien-
and fillers, the morphology of polymer composites and tation should be avoided or enhanced according to
the interaction of polymers and fillers. We need to take specific application requirements. Orientation of fillers
a holistic approach when designing highly thermally can result from processing flow, or can be obtained by
conductive polymer composites. external fields, such as magnetic and electrical field.
(2) Theory and modeling play an important role in the
screening potential compositions and selecting the Despite all the successes, there are still many challenges
more promising ones. Modeling can help determine for heat conductive polymer composites. More research is
(qualitatively or quantitatively) the dependence of expected in the following areas:
composite thermal conductivity (TC) on TC of the
matrix, TC of the filler(s), filler volume fraction and (1) To achieve high TC, significant amount of fillers is
filler shape and morphology. Combining modeling and needed in polymer matrix. High filler loading not only
experimental inputs can provide not only explanation can cause the final composites to be brittle, but also
of trends, but also quantitative designs for subsequent creates more challenges to the processing of polymer
experiments. composites. Moreover, as the fillers are more expen-
(3) Thermal conductivity of polymers is particularly sive than polymer matrix, the cost of final composites
important at low filler loadings. At low filler contents, is high. Thus, further research is needed to obtain high
polymer matrix acts as the thermal barrier and TC at low filler loadings.
becomes rate-limiting in the thermal conduction chan- (2) Thermal conductivity of polymers is important to the
nel. Increasing the TC of polymer matrix can help TC of composites. Some liquid crystalline thermosets
increase the TC of composites. Polymer ordering (crys- can give high TC. However, they are usually expensive
talline and liquid crystalline) and orientation could be due to the complicated synthesis procedure or com-
possible ways to achieve higher TC of the matrix. mercial non-availability of starting materials, which
(4) The addition of thermally conductive inorganic fillers is prevent it from being used commercially. More impor-
indispensable to achieve high TC. The filler type, loading tantly, most of them have very high melting point.
level, filler size, and filler shape have strong influence As they need to be heated to above melting tem-
on the TC of polymer composites. Unlike the electri- perature for curing, the gel time will be extremely
cal conductive behavior, there does not seem to be a short at high temperatures, which makes it impossible
clear percolation threshold for thermally conductive to process them with conventional processing tech-
composites. High TC can only be achieved at high filler niques. Thermosets with high TC, good processability
loadings, when fillers contact with each other and form and reasonable cost are still needed for practical appli-
thermally conductive pathways. cations. Liquid crystalline thermoplastics with smectic
(5) Filler size is also important to the TC of composites. structure can have a high TC. However, they usu-
Composites with large fillers have less filler/polymer ally have low molecular weight and give very poor
interface, lower thermal interfacial resistance, and thus mechanical properties. Further improvement of their
higher TC. However, when the filler size is in the mechanical properties is needed. Polymer blends with
nanoscale, filler properties could also change; for exam- highly packed structure and high density of hydrogen
ple, the melting point of nano-metallic particles could bonds can give high TC. How to introduce this concept
be significantly lower, thus leading to the sintering of into conventional polymers needs further exploration.
nanoparticles and formation of a heat conductive net- Highly oriented polymeric fibers can give high TC. It is
work during polymer curing process. believed that they will find more applications on ther-
(6) Hybrid fillers, such as mixture of fillers of different type, mally conductive composites in the future.
size, and shape, can help achieve higher TC at a lower (3) Future research is needed to develop novel heat con-
filler loading level. They can also help to lower viscosity, ductive fillers. Fillers with high aspect ratio can giver
thus improving processability. Hybrid fillers can be also higher TC at low filler content. Heat conductive, but
used to control TC anisotropy. More isotropic TC can be electrical insulating inorganic fibers are particularly
obtained with using appropriate hybrid fillers. useful for electrical insulating heat conductive com-
(7) Filler surface treatment can help improve filler disper- posites. However, those fibers are rarely commercially
sion, while also reducing the overall viscosity. More available. The spherical fillers, such as spherical Al2 O3 ,
importantly, it decreases thermal interfacial resistance; can give lower viscosity than fillers with irregular shape
therefore, higher TC can be obtained. or fillers with high aspect ratio. However, Al2 O3 has rel-
(8) Phase separated polymer blends with island-sea or atively low TC. Spherical fillers with high TC, such as
co-continuous structure can be used to design double- spherical boron nitride particles could be very useful
percolated structures. Higher TC can be achieved at a for many applications.
lower filler loading level, if the fillers are concentrated (4) Controlling the orientation and location of fillers can
in the continuous phase in the island-sea structure or help to achieve high TC at lower filler content. Novel
located in the interface in the co-continuous structure. processing techniques that can assist fillers to form
heat conductive pathways will continue to emerge in [25] Ramani K, Vaidyanathan A. Finite element analysis of effective ther-
mal conductivity of filled polymeric composites. J Compos Mater
the future, for example, the recent development of 3D
1995;29:1725–40.
printing could help to locate filler into a specific loca- [26] Lewis RW, Nithiarasu P, Seetharamu K. Fundamentals of the finite
tion in the composites. element method for heat and fluid flow. West Sussex: John Wiley
& Sons; 2004, 329 pp.
[27] Rao SS. The finite element method in engineering. 4th ed. Burling-
Acknowledgements ton & Oxford: Elsevier Butterworth-Heinemann; 2005, 688 pp.
[28] Dhatt G, Lefrançois E, Touzot G. Finite element method. In: Bre-
itkopf P, editor. Numerical methods serial. 1st ed. Hoboken &
The authors are grateful to Kevin Howard, Michael Mills, London: John Wiley & Sons & ISTE Ltd; 2012, 597 pp.
and Jeff Zawisza at The Dow Chemical Company for many [29] Hughes TJR. The finite element method: linear static and dynamic
fruitful discussions during the preparation of this work. finite element analysis. 1st ed. North Chelmsford: Courier Corpo-
ration; 2000, 682 pp.
[30] Onsager L. Reciprocal relations in irreversible processes. I. Phys Rev
1931;37:405–26.
References [31] Onsager L. Reciprocal relations in irreversible processes. II. Phys Rev
1931;38:2265–79.
[1] Moore AL, Shi L. Emerging challenges and materials for thermal [32] Kumlutas D, Tavman IH, Coban MT. Thermal conductivity of par-
management of electronics. Mater Today 2014;17:163–74. ticle filled polyethylene composite materials. Compos Sci Technol
[2] Tong XC. Advanced materials for thermal management of electronic 2003;63:113–7.
packaging. New York: Springer; 2011, 616 pp. [33] Peterson R, Anderson A. The Kapitza thermal boundary resistance.
[3] Han ZD, Fina A. Thermal conductivity of carbon nanotubes J Low Temp Phys 1973;11:639–65.
and their polymer nanocomposites: a review. Prog Polym Sci [34] Singh IV, Tanaka M, Endo M. Effect of interface on the ther-
2011;36:914–44. mal conductivity of carbon nanotube composites. Int J Therm Sci
[4] Huang XY, Jiang PK, Tanaka T. A review of dielectric polymer 2007;46:842–7.
composites with high thermal conductivity. IEEE Electr Insul Mag [35] Annapragada SR, Sun D, Garimella SV. Prediction of effective
2011;27:8–16. thermo-mechanical properties of particulate composites. Comput
[5] Bigg DM. Thermal conductivity of heterophase polymer composi- Mater Sci 2007;40:255–66.
tions. Adv Polym Sci 1995;119:1–30. [36] Liang JZ, Li FH. Heat transfer in polymer composites filled with
[6] Garnett JCM. Colours in metal glasses and in metallic films. Proc R inorganic hollow micro-spheres: a theoretical model. Polym Test
Soc Lond Ser A 1904;73:443–5. 2007;26:1025–30.
[7] Bruggeman DAG. Berechnung Verschiedener Physikalischer Kon- [37] Liang JZ, Li FH. Simulation of heat transfer in hollow-glass-bead-
stanten von Heterogenen Substanzen, I. Dielektrizitatskonstanten filled polypropylene composites by finite element method. Polym
und Leitfahigkeiten der Mischkorper aus Isotropen Substanzen. Test 2007;26:419–24.
Annal Phys Leipz 1935;24:636–79. [38] Nayak R, Dora PT, Satapathy A. A computational and experimental
[8] Mohaddespour A, Abolghasemi H, Mostaedi MT, Habibzadeh S. A investigation on thermal conductivity of particle reinforced epoxy
new model for estimation of the thermal conductivity of poly- composites. Comput Mater Sci 2010;48:576–81.
mer/clay nanocomposites. J Appl Polym Sci 2010;118:1042–50. [39] Moniruzzaman M, Winey KI. Polymer nanocomposites containing
[9] Fricke H. Mathematical treatment of the electric conductivity and carbon nanotubes. Macromolecules 2006;39:5194–205.
capacity of the dispersed systems. Phys Rev 1924;24:575–87. [40] Wernik JM, Meguid SA. Recent developments in multifunc-
[10] Kapitza PL. The study of heat transfer in helium II. J Phys tional nanocomposites using carbon nanotubes. Appl Mech Rev
1941;4:181–5. 2010;63:050801/1–50801.
[11] Nan CW, Birringer R, Clarke DR, Gleiter H. Effective thermal conduc- [41] Rahmat M, Hubert P. Carbon nanotube–polymer interactions
tivity of particulate composites with interfacial thermal resistance. in nanocomposites: a review. Compos Sci Technol 2011;72:
J Appl Phys 1997;81:6692–9. 72–84.
[12] Tavman IH. Thermal and mechanical properties of aluminum [42] Spitalsky Z, Tasis D, Papagelis K, Galiotis C. Carbon nanotube–
powder-filled high-density polyethylene composites. J Appl Polym polymer composites: chemistry, processing, mechanical and elec-
Sci 1996;62:2161–7. trical properties. Prog Polym Sci 2010;35:357–401.
[13] Gharagozloo PE, Goodson KE. Aggregate fractal dimensions [43] Du F, Scogna RC, Zhou W, Brand S, Fischer JE, Winey KI. Nanotube
and thermal conduction in nanofluids. J Appl Phys 2010;108: networks in polymer nanocomposites: rheology and electrical con-
074309/1–74309. ductivity. Macromolecules 2004;37:9048–55.
[14] Lin W. Modeling of thermal conductivity of polymer nanocom- [44] Haggenmueller R, Guthy C, Lukes JR, Fischer JE, Winey KI. Single
posites. In: Mittal V, editor. Modeling and prediction of polymer wall carbon nanotube/polyethylene nanocomposites: thermal and
nanocomposite properties. Weinheim, Germany: Wiley-VCH Ver- electrical conductivity. Macromolecules 2007;40:2417–21.
lag GmbH & Co. KGaA; 2013. p. 169–200. [45] Bicerano J, Douglas JF, Brune DA. Model for the viscosity of
[15] Hashin Z, Shtrikman S. A variational approach to the theory particle dispersions. J Macromol Sci Rev Macromol Chem Phys
of the effective magnetic permeability materials. J Appl Phys 1999;C39:561–642.
1962;33:3125–31. [46] Shenogina N, Shenogin S, Xue L, Keblinski P. On the lack of ther-
[16] Hashin Z. Assessment of the self consistent scheme approxi- mal percolation in carbon nanotube composites. Appl Phys Lett
mation: conductivity of particulate composites. J Compos Mater 2005;87:133106/1–133106.
1968;2:284–300. [47] Li XT, Fan XY, Zhu YD, Li J, Adams JM, Shen S, Li HZ. Computational
[17] Hamilton R, Crosser O. Thermal conductivity of heterogeneous two- modeling and evaluation of the thermal behavior of randomly
component systems. Ind Eng Chem Fundam 1962;1:187–91. distributed single-walled carbon nanotube/polymer composites.
[18] Hamilton RL [Dissertation] Thermal conductivity of two-phase Comput Mater Sci 2012;63:207–13.
materials. Norman, OK: University of Oklahoma; 1960, 125 pp. [48] Kim BW, Park SH, Kapadia RS, Bandaru PR. Evidence of perco-
[19] Hatta H, Taya M. Effective thermal conductivity of a misoriented lation related power law behavior in the thermal conductivity
short fiber composite. J Appl Phys 1985;58:2478–86. of nanotube/polymer composites. Appl Phys Lett 2013;102:
[20] Agari Y, Ueda A, Nagai S. Thermal conductivity of a polymer com- 243105/1–243105.
posite. J Appl Polym Sci 1993;49:1625–34. [49] Kwon SY, Kwon IM, Kim YG, Lee S, Seo YS. A large increase in the
[21] Cheng S, Vachon R. A technique for predicting the thermal conduc- thermal conductivity of carbon nanotube/polymer composites pro-
tivity of suspensions, emulsions and porous materials. Int J Heat duced by percolation phenomena. Carbon 2013;55:285–90.
Mass Transfer 1970;13:537–43. [50] Hill RF, Strader JL. Rudimentary finite element thermal modeling
[22] Nielsen L. Generalized equation for the elastic moduli of composite of platelet-filled polymer-ceramic composites. IEEE Trans Compon
materials. J Appl Phys 1970;41:4626–7. Packag Technol 2007;30:235–41.
[23] Nielsen L. Thermal conductivity of particulate-filled polymers. J [51] Mortazavi B, Benzerara O, Meyer H, Bardon J, Ahzi S. Com-
Appl Polym Sci 1973;17:3819–20. bined molecular dynamics-finite element multiscale modeling of
[24] Lewis T, Nielsen L. Dynamic mechanical properties of particulate- thermal conduction in graphene epoxy nanocomposites. Carbon
filled polymers. J Appl Polym Sci 1970;14:1449–71. 2013;60:356–65.
[52] Sanada K, Tada Y, Shindo Y. Thermal conductivity of polymer [74] Anonymous. Thermal insulation – determination of steady-state
composites with close-packed structure of nano and micro fillers. thermal transmission properties of thermal insulation for circular
Composites A 2009;40:724–30. pipes. International Standard ISO 8497:1994; 2013, 16 pp.
[53] Keblinski P, Eastman JA, Cahill DG. Nanofluids for thermal transport. [75] Anonymous. Standard test method for thermal diffusivity by the
Mater Today 2005;8:36–44. flash method, ASTM E1461-13. American Society for Testing and
[54] Shenogin S, Xue L, Ozisik R, Keblinski P, Cahill DG. Role of thermal Materials; 2013, 11 pp.
boundary resistance on the heat flow in carbon-nanotube compos- [76] Anonymous. Plastics – determination of thermal conductivity and
ites. J Appl Phys 2004;95:8136–44. thermal diffusivity – Part 4; Laser flash method. International
[55] Wang Y, Zhan HF, Xiang Y, Yang C, Wang CM, Zhang YY. Standard ISO 22007-4:2008; 2011, 12 pp.
Effect of covalent functionalization on thermal transport across [77] Anonymous. Fine ceramics (advanced ceramics, advanced techni-
graphene–polymer interfaces. J Phys Chem C 2015;119:12731–8. cal ceramics) – determination of thermal diffusivity of monolithic
[56] Alzina A, Toussaint E, Béakou A, Skoczen B. Multiscale modelling of ceramics by laser flash method. International Standard ISO
thermal conductivity in composite materials for cryogenic struc- 18755:2005; 2014, 31 pp.
tures. Compos Struct 2006;74:175–85. [78] Anonymous. An overview of the various thermal conductivi-
[57] Roy AK, Farmer BL, Varshney V, Sihn S, Lee J, Ganguli S. Importance ties depending on the used method. Netzsch Company; 2015.
of interfaces in governing thermal transport in composite mate- https://www.netzsch-thermal-analysis.com/us/landing-pages/
rials: modeling and experimental perspectives. ACS Appl Mater which-method-is-the-best-suited-to-my-particular-sample/2015
Interfaces 2012;4:545–63. [accessed October 2015].
[58] Varshney V, Patnaik SS, Roy AK, Froudakis G, Farmer BL. Modeling [79] Anonymous. Standard test method for thermal conductivity of
of thermal transport in pillared-graphene architectures. ACS Nano refractories by hot wire (platinum resistance thermometer tech-
2010;4:1153–61. nique), ASTM C1113/C1113M – 09(2013). American Society for
[59] Shin H, Yang S, Chang S, Yu S, Cho M. Multiscale homog- Testing and Materials; 2013, 6 pp.
enization modeling for thermal transport properties of poly- [80] Anonymous. Refractory materials – determination of thermal con-
mer nanocomposites with Kapitza thermal resistance. Polymer ductivity – Part 1; Hot-wire methods (cross-array and resistance
2013;54:1543–54. thermometer). International Standard ISO 8894-1:2010; 2015, 19
[60] Gharib-Zahedi MR, Tafazzoli M, Böhm MC, Alaghemandi M. pp.
Interfacial thermal transport and structural preferences in car- [81] Anonymous. Refractory materials – determination for thermal
bon nanotube–polyamide-6,6 nanocomposites: how important conductivity – Part 2; Hot-wire method (parallel). International
are chemical functionalization effects? Phys Chem Chem Phys Standard ISO 8894-2:2007; 2011, 13 pp.
2015;17:14502–12. [82] Vozar L. A computer-controlled apparatus for thermal conductiv-
[61] Tanaka K, Ogata S, Kobayashi R, Tamura T, Kouno T. A molecular ity measurement by the transient hot wire method. J Therm Anal
dynamics study on thermal conductivity of thin epoxy polymer 1996;46:495–505.
sandwiched between alumina fillers in heat-dissipation composite [83] Kwon SY, Lee S. Precise measurement of thermal conductiv-
material. Int J Heat Mass Transfer 2015;89:714–23. ity of liquid over a wide temperature range using a transient
[62] Konatham D, Striolo A. Thermal boundary resistance at the hot-wire technique by uncertainty analysis. Thermochim Acta
graphene-oil interface. Appl Phys Lett 2009;95:163105/1–163105. 2012;542:18–23.
[63] Sato K, Ijuin A, Hotta Y. Thermal conductivity enhancement of alu- [84] Solorzano E, Rodriguez-Perez MA, de Saja JA. Thermal conductivity
mina/polyamide composites via interfacial modification. Ceram Int of cellular metals measured by the transient plane source method.
2015;41:10314–8. Adv Eng Mater 2008;10:596–602.
[64] Buck W, Rudtsch S. Thermal properties. In: Czichos H, Saito T, Smith [85] Gustafsson SE. Transient plane source techniques for thermal-
L, editors. Springer handbook of materials measurement methods. conductivity and thermal-diffusivity measurements of solid
Berlin, Heidelberg: Springer; 2006. p. 399–429. materials. Rev Sci Instrum 1991;62:797–804.
[65] Anonymous. Standard test method for steady state heat flux mea- [86] Anonymous. Instruments comparison chart (hot disk AB); 2015.
surements and thermal transmission properties by mean of the http://www.hotdiskinstruments.com/products/instruments-for-
guarded hot plate apparatus. ASTM C177-13: American Society for thermal-conductivity-measurements.html#instruments-
Testing and Materials; 2013, 23 pp. comparison-chart [accessed October 2015].
[66] Anonymous. Determination of steady-state thermal resistance and [87] Anonymous. Plastics – determination of thermal conductivity and
related properties – guarded hot plate apparatus. International thermal diffusivity – Part 2; Transient plane heat source (hot disk)
Standard ISO 8302:1991; 2013, 47 pp. method. International Standard, ISO 22007-2:2015; 2015, 20 pp.
[67] Anonymous. Thermal performance of building materials and prod- [88] Min S, Blumm J, Lindemann A. A new laser flash system for
ucts – determination of thermal resistance by means of guarded hot measurement of the thermophysical properties. Thermochim Acta
plate and heat flow meter methods – products of high and medium 2007;455:46–9.
thermal resistance. EN 12667: 2001. European Standard; 2001, 52 [89] dos Santos WN, Mummery P, Wallwork A. Thermal diffusivity
pp. of polymers by the laser flash technique. Polym Test 2005;24:
[68] Corsan JM. Axial heat flow methods of thermal conductivity 628–34.
measurement for good conducting materials. In: Maglic KD, Cezair- [90] Krupa I, Novak I, Chodak I. Electrically and thermally conductive
liyan A, Peletsky VE, editors. Compendium of thermophysical polyethylene/graphite composites and their mechanical proper-
property measurement methods. Recommended measurement ties. Synth Met 2004;145:245–52.
techniques and practices, vol. 2. New York: Plenum Press; 1992. [91] Henry A, Chen G. High thermal conductivity of single polyethyl-
p. 3–31. ene chains using molecular dynamics simulations. Phys Rev Lett
[69] Maglic KD, Cezairliyan A, Peletsky VE. Compendium of ther- 2008;101:235502/1–235502.
mophysical property measurement methods, vol. 1: Survey [92] Luo TF, Esfarjani K, Shiomi J, Henry A, Chen G. Molecular dynam-
of measurement techniques. New York: Plenum Press; 1984. ics simulation of thermal energy transport in polydimethylsiloxane
p. 11–40. (PDMS). J Appl Phys 2011;109:074321/1–74321.
[70] Anonymous. Standard test method for thermal conductivity of [93] Wang RY, Segalman RA, Majumdar A. Room temperature ther-
solids using the guarded-comparative-longitudinal heat flow tech- mal conductance of alkanedithiol self-assembled monolayers. Appl
nique, ASTM E1225-13. American Society for Testing and Materials; Phys Lett 2006;89:173113/1–173113.
2013, 10 pp. [94] Wang ZH, Carter JA, Lagutchev A, Koh YK, Seong NH, Cahill DG,
[71] Anonymous. Standard test method for steady-state thermal trans- Dlott DD. Ultrafast flash thermal conductance of molecular chains.
mission properties by means of the heat flow meter apparatus, Science 2007;317:787–90.
ASTM C518-10. American Society for Testing and Materials; 2010, [95] Yang Y. Thermal conductivity. In: Mark JE, editor. Physical prop-
16 pp. erties of polymers handbook. New York: Springer-Verlag; 2007. p.
[72] Anonymous. Standard test method for evaluating the resistance of 155–63.
thermal transmission of materials by the guarded heat flow meter [96] Anonymous. Thermal properties of polymers; 2015. http://www.
technique, ASTM E1530-11. American Society for Testing and Mate- netzsch-thermal-analysis.com/uploads/tx nxnetzschmedia/files/
rials; 2011, 9 pp. TPoP web mit Hervorhebung.pdf [accessed January 2015].
[73] Anonymous. Thermal insulation – determination of steady-state [97] Yamanaka A, Takao T. Thermal conductivity of high strength poly-
thermal resistance and related properties – heat flow meter appa- ethylene fiber and its application for cryogenic use. Sen-I Gakkaishi
ratus. International Standard ISO 8301:1991; 2014, 38 pp. 2012;68:111–7.
[98] Henry A, Chen G, Plimpton SJ, Thompson A. 1D-to-3D transition of [126] Zhong ZX, Wingert MC, Strzalka J, Wang HH, Sun T, Wang J, Chen RK,
phonon heat conduction in polyethylene using molecular dynamics Jiang Z. Structure-induced enhancement of thermal conductivities
simulations. Phys Rev B 2010;82:144308/1–144308. in electrospun polymer nanofibers. Nanoscale 2014;6:8283–91.
[99] Yorifuji D, Ando S. Molecular structure dependence of out-of- [127] Algaer EA, Alaghemandi M, Bohm MC, Muller-Plathe F. Anisotropy
plane thermal diffusivities in polyimide films: a key parameter of the thermal conductivity of stretched amorphous polystyrene
for estimating thermal conductivity of polymers. Macromolecules in supercritical carbon dioxide studied by reverse nonequi-
2010;43:7583–93. librium molecular dynamics simulations. J Phys Chem B
[100] Chae HG, Kumar S. Materials science – making strong fibers. Science 2009;113:14596–603.
2008;319:908–9. [128] Choy CL. Thermal-conductivity of polymers. Polymer 1977;18:
[101] Huang XP, Liu GQ, Wang XW. New secrets of spider silk: excep- 984–1004.
tionally high thermal conductivity and its abnormal change under [129] Eiermann K, Hellwege KH. Thermal conductivity of high polymers
stretching. Adv Mater 2012;24:1482–6. from −180 ◦ C to 90 ◦ C. J Polym Sci 1962;57:99–106.
[102] Zhang T, Wu XF, Luo TF. Polymer nanofibers with outstanding [130] Hennig J. Anisotropy and structure in uniaxially stretched amor-
thermal conductivity and thermal stability: fundamental linkage phous high polymers. J Polym Sci C Polym Symp 1967;16:2751–8.
between molecular characteristics and macroscopic thermal prop- [131] Singh V, Bougher TL, Weathers A, Cai Y, Bi KD, Pettes MT, McMe-
erties. J Phys Chem C 2014;118:21148–59. namin SA, Lv W, Resler DP, Gattuso TR, Altman DH, Sandhage KH,
[103] Zhang T, Luo TF. Morphology-influenced thermal conductivity Shi L, Henry A, Cola BA. High thermal conductivity of chain-oriented
of polyethylene single chains and crystalline fibers. J Appl Phys amorphous polythiophene. Nat Nanotechnol 2014;9:384–90.
2012;112:094304/1–94304. [132] Zhang T, Luo TF. High-contrast, reversible thermal conductivity reg-
[104] Liu J, Yang RG. Length-dependent thermal conductivity of single ulation utilizing the phase transition of polyethylene nanofibers.
extended polymer chains. Phys Rev B 2012;86:104307/1–104307. ACS Nano 2013;7:7592–600.
[105] Anderson DR. Thermal conductivity of polymers. Chem Rev [133] Okamoto S, Matsumi Y, Saito S, Miyakoshi R, Kondo T. Newly devel-
1966;66:677–90. oped innovative LCP resins applied into LED usages. Sumitomo
[106] Choy CL, Kwok KW, Leung WP, Lau FP. Thermal-conductivity of Kagaku 2011;2011(I):1–14.
poly(ether ether ketone) and its short-fiber composites. J Polym [134] Choy CL, Wong YW, Lau KWE, Yang GW, Yee AF. Thermal-
Sci B Polym Phys 1994;32:1389–97. conductivity and thermal expansivity of thermotropic liquid-
[107] Hansen D, Bernier GA. Thermal-conductivity of polyethylene crystalline polymers. J Polym Sci B Polym Phys 1995;33:2055–64.
– effects of crystal size, density and orientation on thermal- [135] Hasegawa M, Suyama N, Shimoyama N, Aoki H, Nunokawa T,
conductivity. Polym Eng Sci 1972;12:204–8. Kimura T. Enhanced thermal conductivity of semi-aliphatic liquid
[108] Yu JC, Sundqvist B, Tonpheng B, Andersson O. Thermal conductivity crystalline polybenzoxazoles using magnetic orientation. Polym Int
of highly crystallized polyethylene. Polymer 2014;55:195–200. 2011;60:1240–7.
[109] Liu J, Yang RG. Tuning the thermal conductivity of polymers with [136] Kimura T, Shimoyama N, Aoki H. High thermal conductive polymer
mechanical strains. Phys Rev B 2010;81:174122/1–174122. prepared in a magnetic field. Soc Polym Sci Jpn 2010;59:77–80.
[110] Choy CL, Luk WH, Chen FC. Thermal-conductivity of highly oriented [137] Yoshihara S, Ezaki T, Nakamura M, Watanabe J, Matsumoto K.
polyethylene. Polymer 1978;19:155–62. Enhanced thermal conductivity of thermoplastics by lamellar
[111] Anandakumaran K, Roy SK, Manley RS. Drawing-induced changes crystal alignment of polymer matrices. Macromol Chem Phys
in the properties of polyethylene fibers prepared by gelation crys- 2012;213:2213–9.
tallization. Macromolecules 1988;21:1746–51. [138] Tokita M, Osada K, Yamada M, Watanabe J. Chain-folded lamel-
[112] Choy CL, Fei Y, Xi TG. Thermal-conductivity of gel-spun polyethyl- lar structure in the smectic H phase of a main-chain polyester.
ene fibers. J Polym Sci B Polym Phys 1993;31:365–70. Macromolecules 1998;31:8590–4.
[113] Mergenthaler DB, Pietralla M, Roy S, Kilian HG. Thermal- [139] Tokita M, Tokunaga K, Funaoka S, Osada K, Watanabe J. Par-
conductivity in ultraoriented polyethylene. Macromolecules allel and perpendicular orientations observed in shear aligned
1992;25:3500–2. S-CA liquid crystal of main-chain polyester. Macromolecules
[114] Choy CL, Leung WP. Thermal-conductivity of ultradrawn polyeth- 2004;37:2527–31.
ylene. J Polym Sci Polym Phys Ed 1983;21:1243–6. [140] Osada K, Koike M, Tagawa H, Hunaoka S, Tokita M, Watanabe J.
[115] Kanamoto T, Tsuruta A, Tanaka K, Takeda M, Porter RS. Super- Two distinct types of orientation process observed in uniaxially
drawing of ultrahigh molecular-weight polyethylene. 1. Effect of elongated smectic LC melt. Macromolecules 2005;38:7337–42.
techniques on drawing of single-crystal mats. Macromolecules [141] Fernandez-Blazquez JP, Bello A, Perez E. Parallel and perpendicular
1988;21:470–7. orientation in a thermotropic main-chain liquid-crystalline poly-
[116] Choy CL, Wong YW, Yang GW, Kanamoto T. Elastic modulus and mer. Macromolecules 2007;40:703–9.
thermal conductivity of ultradrawn polyethylene. J Polym Sci B [142] Yoshihara S, Sakaguchi M, Matsumoto K, Tokita M, Watanabe J.
Polym Phys 1999;37:3359–67. Influence of molecular orientation direction on the in-plane ther-
[117] Brady JM, Thomas EL. Conversion of single-crystal mats to ultrahigh mal conductivity of polymer/hexagonal boron nitride composites.
modulus polyethylene – the formation of a continuous crystalline J Appl Polym Sci 2014;131:39768/1–39768.
phase. Polymer 1989;30:1615–22. [143] Yoshihara S, Tokita M, Ezaki T, Nakamura M, Sakaguchi M, Mat-
[118] Vanaerle NAJM, Braam AWM. A structural study on solid-state sumoto K, Watanabe J. Main-chain smectic liquid crystalline
drawing of solution-crystallized ultrahigh molecular-weight poly- polymer exhibiting unusually high thermal conductivity in an
ethylene. J Mater Sci 1988;23:4429–36. isotropic composite. J Appl Polym Sci 2014;131:39896/1–39896.
[119] Piraux L, Kinanyalaoui M, Issi JP, Begin D, Billaud D. Thermal- [144] Kim GH, Lee D, Shanker A, Shao L, Kwon MS, Gidley D, Kim J, Pipe
conductivity of an oriented polyacetylene film. Solid State Commun KP. High thermal conductivity in amorphous polymer blends by
1989;70:427–9. engineered interchain interactions. Nat Mater 2015;14:295–300.
[120] Langer L, Billaud D, Issi JP. Thermal conductivity of stretched and [145] Akatsuka M, Takezawa Y. Study of high thermal conductive epoxy
annealed poly(p-phenylene sulfide) films. Solid State Commun resins containing controlled high-order structures. J Appl Polym Sci
2003;126:353–7. 2003;89:2464–7.
[121] Choy CL, Chen FC, Luk WH. Thermal-conductivity of oriented crys- [146] Giang T, Park J, Cho I, Ko Y, Kim J. Effect of backbone moiety
talline polymers. J Polym Sci Polym Phys Ed 1980;18:1187–207. in epoxies on thermal conductivity of epoxy/alumina composite.
[122] Choy CL, Lau KWE, Wong YW, Ma HM, Yee AF. Thermal conduc- Polym Compos 2013;34:468–76.
tivity and thermal expansivity of in situ composites of a liquid [147] Harada M, Hamaura N, Ochi M, Agari Y. Thermal conductivity of liq-
crystalline polymer and polycarbonate. Polym Eng Sci 1996;36: uid crystalline epoxy/BN filler composites having ordered network
827–34. structure. Composites B 2013;55:306–13.
[123] Wang XJ, Ho V, Segalman RA, Cahill DG. Thermal conductivity of [148] Masaki A, Yoshitaka T, Katsuo S. Thermosetting resin compounds.
high-modulus polymer fibers. Macromolecules 2013;46:4937–43. US20060276568A1. Saitama & Hitachinake & Hitachi Ltd.; 2006.
[124] Shen S, Henry A, Tong J, Zheng RT, Chen G. Polyethylene nanofi- [149] Zhao CB, Xu SC, Qin YF, Su L, Yang XJ. Thermal conductivity cyanate
bres with very high thermal conductivities. Nat Nanotechnol ester resin composites filled with boron nitride. Adv Mater Res
2010;5:251–5. 2014;893:259–62.
[125] Cao BY, Li YW, Kong J, Chen H, Xu Y, Yung KL, Cai A. High ther- [150] Masaki A, Yoshitaka T, Yuzo I. Insulating composition. Tokyo, Japan:
mal conductivity of polyethylene nanowire arrays fabricated by Hitachi Ltd.; 2001. US6261481B1.
an improved nanoporous template wetting technique. Polymer [151] Fukushima K, Takahashi H, Takezawal Y, Hattori M, Itoh
2011;52:1711–5. M, Yonekura M. High thermal conductive epoxy resins with
controlled high-order structure [electrical insulation applications]. approach for the composites containing the geometrized fillers.
In: CEIDP’04: annual report conference on electrical insulation and Composites A 2015;69:219–25.
dielectric phenomena, October 17th, 2004. Colorado City, USA: [176] Ha SM, Lee HL, Lee SG, Kim BG, Kim YS, Won JC, Choi WJ, Lee DC, Kim
IEEE; 2004. p. 340–3. J, Yoo Y. Thermal conductivity of graphite filled liquid crystal poly-
[152] Uehara T, Oshimi K, Kawai K, Inagaki S, Sunaga T. Epoxy resin com- mer composites and theoretical predictions. Compos Sci Technol
position and cured material of the same. Nippon Kayaku Co., Ltd.; 2013;88:113–9.
2010. JP2010163540A. [177] Zhou WY, Wang CF, Ai T, Wu K, Zhao FJ, Gu HZ. A novel fiber-
[153] Uehara T, Oshimi K, Kawai K, Inagaki S, Sunaga T. Epoxy resin, epoxy reinforced polyethylene composite with added silicon nitride
resin composition, and cured product thereof. JP2010174134A. particles for enhanced thermal conductivity. Composites A
Nippon Kayaku Co., Ltd.; 2010. 2009;40:830–6.
[154] Masayuki T, Toru K, Tsukasa I, Naoyuki S, Hisashi A, Mitsukazu O, [178] Sato K, Horibe H, Shirai T, Hotta Y, Nakano H, Nagai H, Mitsuishi
Miyuki H. Thermally-conductive epoxy resin molded article and K, Watari K. Thermally conductive composite films of hexagonal
method of producing the same. US20040224163A1. Polymatech boron nitride and polyimide with affinity-enhanced interfaces. J
Co., Ltd.; 2004. Mater Chem 2010;20:2749–52.
[155] Fukushima K, Takahashi H, Takezawa Y, Kawahira T, Itoh M, [179] Zhou WY, Qi SH, An QL, Zhao HZ, Liu NL. Thermal conductivity of
Kanai J. High thermal conductive resin composites with con- boron nitride reinforced polyethylene composites. Mater Res Bull
trolled nanostructures for electric devices. IEEJ Trans Fundam 2007;42:1863–73.
Mater 2006;126:1167–72. [180] Kume S, Yamada I, Watari K, Harada I, Mitsuishi K. High-thermal-
[156] Shiota A, Ober CK. Synthesis and curing of novel LC twin epoxy conductivity AlN filler for polymer/ceramics composites. J Am
monomers for liquid crystal thermosets. J Polym Sci A Polym Chem Ceram Soc 2009;92:153–6.
1996;34:1291–303. [181] Zhou YC, Yao YG, Chen CY, Moon K, Wang H, Wong CP. The use
[157] Song SH, Katagi H, Takezawa Y. Study on high thermal conductiv- of polyimide-modified aluminum nitride fillers in AlN@PI/epoxy
ity of mesogenic epoxy resin with spherulite structure. Polymer composites with enhanced thermal conductivity for electronic
2012;53:4489–92. encapsulation. Sci Rep 2014;4:1–6.
[158] Shiota A, Ober CK. Orientation of liquid crystalline epoxides under [182] Yu Y, Wu LZ, Zhi JF. Diamond nanowires: fabrication, struc-
ac electric fields. Macromolecules 1997;30:4278–87. ture, properties, and applications. Angew Chem Int Ed
[159] Barclay GG, Mcnamee SG, Ober CK, Papathomas KI, Wang DW. 2014;53:14326–51.
The mechanical and magnetic alignment of liquid-crystalline epoxy [183] Zhou TL, Wang X, Mingyuan GU, Liu XH. Study of the thermal
thermosets. J Polym Sci A Polym Chem 1992;30:1845–53. conduction mechanism of nano-SiC/DGEBA/EMI-2,4 composites.
[160] Benicewicz BC, Smith ME, Earls JD, Priester RD, Setz SM, Duran RS, Polymer 2008;49:4666–72.
Douglas EP. Magnetic field orientation of liquid crystalline epoxy [184] Hill RF, Supancic PH. Determination of the thermal resistance of the
thermosets. Macromolecules 1998;31:4730–8. polymer-ceramic interface of alumina-filled polymer composites.
[161] Harada M, Ochi M, Tobita M, Kimura T, Ishigaki T, Shimoyama J Am Ceram Soc 2004;87:1831–5.
N, Aoki H. Thermal-conductivity properties of liquid-crystalline [185] Lee WS, Han IY, Yu J, Kim SJ, Byun KY. Thermal characteriza-
epoxy resin cured under a magnetic field. J Polym Sci B Polym Phys tion of thermally conductive underfill for a flip-chip package
2003;41:1739–43. using novel temperature sensing technique. Thermochim Acta
[162] Geibel K, Hammerschmidt A, Strohmer F. In-situ photopolymer- 2007;455:148–55.
ized, oriented liquid-crystalline diacrylates with high thermal- [186] Wang XF, Wang RC, Peng CQ, Li TT, Liu B. Synthesis and sin-
conductivities. Adv Mater 1993;5:107–9. tering of beryllium oxide nanoparticles. Prog Nat Sci Mater Int
[163] Xie BH, Huang X, Zhang GJ. High thermal conductive polyvinyl 2010;20:81–6.
alcohol composites with hexagonal boron nitride microplatelets [187] Akishin GP, Turnaev SK, Vaispapir VY, Gorbunova MA, Makurin YN,
as fillers. Compos Sci Technol 2013;85:98–103. Kiiko VS, Ivanovskii AL. Thermal conductivity of beryllium oxide
[164] Liem H, Choy HS. Superior thermal conductivity of polymer ceramic. Refract Ind Ceram 2009;50:465–8.
nanocomposites by using graphene and boron nitride as fillers. [188] Mamunya YP, Davydenko VV, Pissis P, Lebedev E. Electrical and
Solid State Commun 2013;163:41–5. thermal conductivity of polymers filled with metal powders. Eur
[165] Zhou T, Wang X, Cheng P, Wang T, Xiong D, Wang X. Improving the Polym J 2002;38:1887–97.
thermal conductivity of epoxy resin by the addition of a mixture of [189] Shahil KMF, Balandin AA. Graphene-multilayer graphene
graphite nanoplatelets and silicon carbide microparticles. Express nanocomposites as highly efficient thermal interface materials.
Polym Lett 2013;7:585–94. Nano Lett 2012;12:861–7.
[166] Wypych G. Handbook of fillers: physical properties of fillers and [190] Zhang L, Wu JT, Jiang L. Graphene and its polymer nanocomposites.
filled materials. 2nd ed. Toronto: ChemTec Publishing; 2000, 831 Prog Chem 2014;26:560–71.
pp. [191] Wang M, Kang QJ, Pan N. Thermal conductivity enhancement of
[167] Chung DDL. Materials for thermal conduction. Appl Therm Eng carbon fiber composites. Appl Therm Eng 2009;29:418–21.
2001;21:1593–605. [192] Gaxiola DL, Keith JM, Mo N, King JA, Johnson BA. Predicting the
[168] Wang SL, Cheng Y, Wang RR, Sun J, Gao L. Highly thermal conductive thermal conductivity of multiple carbon fillers in polypropylene-
copper nanowire composites with ultralow loading: toward appli- based resins. J Compos Mater 2011;45:1271–84.
cations as thermal interface materials. ACS Appl Mater Interfaces [193] Wolff S, Wang MJ. Carbon black reinforcement of elastomers. In:
2014;6:6481–6. Donnet JB, Bansal RC, Wang MJ, editors. Carbon black science and
[169] Bjorneklett A, Halbo L, Kristiansen H. Thermal-conductivity of technology. 2nd ed. New York: Marcel Dekker; 1993. p. 289–355.
epoxy adhesives filled with silver particles. Int J Adhes Adhes [194] Chen YM, Ting JM. Ultra high thermal conductivity polymer com-
1992;12:99–104. posites. Carbon 2002;40:359–62.
[170] Pashayi K, Fard HR, Lai FY, Iruvanti S, Plawsky J, Borca-Tasciuc [195] Adams PM, Katzman HA, Rellick GS, Stupian GW. Characterization
T. High thermal conductivity epoxy-silver composites based on of high thermal conductivity carbon fibers and a self-reinforced
self-constructed nanostructured metallic networks. J Appl Phys graphite panel. Carbon 1998;36:233–45.
2012;111:104310/1–104310. [196] Manocha LM, Warrier A, Manocha S, Sathiyamoorthy D, Banerjee S.
[171] Zhang WB, Xu XF, Yang JH, Huang T, Zhang N, Wang Y, Zhou Thermophysical properties of densified pitch based carbon/carbon
ZW. High thermal conductivity of poly (vinylidene fluoride)/carbon materials – I. Unidirectional composites. Carbon 2006;44:480–7.
nanotubes nanocomposites achieved by adding polyvinylpyrroli- [197] Pizza A, Metz R, Hassanzadeh M, Bantignies JL. Life cycle assess-
done. Compos Sci Technol 2015;106:1–8. ment of nanocomposites made of thermally conductive graphite
[172] Minus ML, Kumar S. The processing, properties, and structure of nanoplatelets. Int J Life Cycle Assess 2014;19:1226–37.
carbon fibers. JOM 2005;57:52–8. [198] Yu SZ, Hing P, Hu X. Thermal conductivity of polystyrene-aluminum
[173] Endo M. Structure of mesophase pitch-based carbon-fibers. J Mater nitride composite. Composites A 2002;33:289–92.
Sci 1988;23:598–605. [199] Pezzotti G, Kamada I, Miki S. Thermal conductivity of
[174] Stankovich S, Dikin DA, Dommett GHB, Kohlhaas KM, Zimney EJ, AlN/polystyrene interpenetrating networks. J Eur Ceram Soc
Stach EA, Piner RD, Nguyen ST, Ruoff RS. Graphene-based composite 2000;20:1197–203.
materials. Nature 2006;442:282–6. [200] Morelli DT, Heremans JP. Thermal conductivity of germanium, sil-
[175] Kim SY, Noh YJ, Yu J. Thermal conductivity of graphene icon, and carbon nitrides. Appl Phys Lett 2002;81:5126–8.
nanoplatelets filled composites fabricated by solvent-free [201] Hong JP, Yoon SW, Hwang T, Oh JS, Hong SC, Lee Y, Nam JD. High
processing for the excellent filler dispersion and a theoretical thermal conductivity epoxy composites with bimodal distribution
of aluminum nitride and boron nitride fillers. Thermochim Acta [227] Fu JF, Shi LY, Zhang DS, Zhong QD, Chen Y. Effect of nanoparticles on
2012;537:70–5. the performance of thermally conductive epoxy adhesives. Polym
[202] Liu Z, Wu BT, Gu MY. Effect of hydrolysis of AlN particulates on cor- Eng Sci 2010;50:1809–19.
rosion behavior of Al/AlNp composite in neutral chloride solution. [228] Sumita M, Shizuma T, Miyasaka K, Ishikawa K. Effect of reducible
Composites A 2007;38:94–9. properties of temperature, rate of strain, and filler content on the
[203] Edwards KM, Howard KE. Evaluation of a thermally enhanced mold tensile yield stress of nylon 6 composites filled with ultrafine par-
compound, containing SCAN TM silica coated aluminum nitride ticles. J Macromol Sci B Phys 1983;22:601–18.
filler, for SO-8 packaged power. In: Proc IEEE 4th Int Symp AdvPackg [229] Fu SY, Feng XQ, Lauke B, Mai YW. Effects of particle size,
Mater. 1998. p. 215–22. particle/matrix interface adhesion and particle loading on mechan-
[204] Yokota H, Ibukiyama M. Microstructure tailoring for high ther- ical properties of particulate-polymer composites. Composites B
mal conductivity of beta-Si3 N4 ceramics. J Am Ceram Soc 2008;39:933–61.
2003;86:197–9. [230] Terao T, Bando Y, Mitome M, Zhi CY, Tang CC, Golberg D. Thermal
[205] Xia ZP, Li ZQ. Structural evolution of hexagonal BN and cubic BN conductivity improvement of polymer films by catechin-modified
during ball milling. J Alloys Compd 2007;436:170–3. boron nitride nanotubes. J Phys Chem C 2009;113:13605–9.
[206] Zhang K, Zhang L. The model of structure place transformation for [231] Teng CC, Ma CCM, Chiou KC, Lee TM. Synergetic effect of thermal
phase change from graphite to diamond through explosion. Proc conductive properties of epoxy composites containing function-
Eng 2012;27:1364–72. alized multi-walled carbon nanotubes and aluminum nitride.
[207] Bigg DM. Electrical properties of metal-filled polymer composites. Composites B 2012;43:265–71.
In: Bhattacharya SK, editor. Metal-filled polymers. New York: Mar- [232] Leung SN, Khan MO, Chan E, Naguib HE, Dawson F, Adinkrah
cel Dekker; 1986. p. 165–226. V, Lakatos-Hayward L. Synergistic effects of hybrid fillers on the
[208] Wang HJ, Yi XS. Effects of interfacial thermal barrier resistance and development of thermally conductive polyphenylene sulfide com-
particle shape and size on the thermal conductivity of AlN/PI composites. J Appl Polym Sci 2013;127:3293–301.
posites. Compos Sci Technol 2004;64:1623–8. [233] Zeng J, Fu RL, Shen Y, He H, Song XF. High thermal conductive
[209] Yuan FY, Zhang HB, Li XF, Li XZ, Yu ZZ. Synergistic effect of boron epoxy molding compound with thermal conductive pathway. J Appl
nitride flakes and tetrapod-shaped ZnO whiskers on the thermal Polym Sci 2009;113:2117–25.
conductivity of electrically insulating phenol formaldehyde com- [234] Pak SY, Kim HM, Kim SY, Youn JR. Synergistic improvement of
posites. Composites A 2013;53:137–44. thermal conductivity of thermoplastic composites with mixed
[210] Tao Y, Yang ZG, Lu XL, Tao GL, Xia YP, Wu HP. Influence of filler boron nitride and multi-walled carbon nanotube fillers. Carbon
morphology on percolation threshold of isotropical conductive 2012;50:4830–8.
adhesives (ICA). Sci China Technol Sci 2012;55:28–33. [235] Zhu BL, Zheng H, Wang J, Ma J, Wu J, Wu R. Tailoring of ther-
[211] Tavman IH, Akinci H. Transverse thermal conductivity of fiber mal and dielectric properties of LDPE-matrix composites by the
reinforced polymer composites. Int Commun Heat Mass Transfer volume fraction, density, and surface modification of hollow glass
2000;27:253–61. microsphere filler. Composites B 2014;58:91–102.
[212] Kusunose T, Yagi T, Firoz SH, Sekino T. Fabrication of epoxy/silicon [236] Yang K, Gu MY. Enhanced thermal conductivity of epoxy nanocom-
nitride nanowire composites and evaluation of their thermal con- posites filled with hybrid filler system of triethylenetetramine-
ductivity. J Mater Chem A 2013;1:3440–5. functionalized multi-walled carbon nanotube/silane-modified
[213] Yu J, Chen Y, Wuhrer R, Liu ZW, Ringer SP. In situ forma- nano-sized silicon carbide. Composites A 2010;41:215–21.
tion of BN nanotubes during nitriding reactions. Chem Mater [237] Yoo Y, Lee HL, Ha SM, Jeon BK, Won JC, Lee SG. Effect of graphite
2005;17:5172–6. and carbon fiber contents on themorphology and properties of
[214] Xu J, Munari A, Dalton E, Mathewson A, Razeeb KM. Silver nanowire thermally conductive composites based on polyamide 6. Polym Int
array-polymer composite as thermal interface material. J Appl Phys 2014;63:151–7.
2009;106:124310/1–124310. [238] Yung KC, Liem H. Enhanced thermal conductivity of boron nitride
[215] Hill RF, Supancic PH. Thermal conductivity of platelet-filled poly- epoxy-matrix composite through multi-modal particle size mixing.
mer composites. J Am Ceram Soc 2002;85:851–7. J Appl Polym Sci 2007;106:3587–91.
[216] Gowda A, Paisner SN, Tonapi S, Meneghetti P, Hans P, Strosaker G, [239] Bae JW, Kim W, Cho SH, Lee SH. The properties of AlN-filled epoxy
Acharya A, Nagarkar K, Srihari K. Spherical boron nitride fillers for molding compounds by the effects of filler size distribution. J Mater
high performance thermal greases. In: Proc 7th Electr Pack Technol Sci 2000;35:5907–13.
Conf, vol. 2. 2005. p. 8–16. [240] Choi S, Kim J. Thermal conductivity of epoxy composites with a
[217] Ohashi M, Kawakami S, Yokogawa Y, Lai GC. Spherical aluminum binary-particle system of aluminum oxide and aluminum nitride
nitride fillers for heat-conducting plastic packages. J Am Ceram Soc fillers. Composites B 2013;51:140–7.
2005;88:2615–8. [241] Warzoha RJ, Fleischer AS. Heat flow at nanoparticle interfaces. Nano
[218] Leung SN, Khan MO, Chan E, Naguib H, Dawson F, Adinkrah V, Energy 2014;6:137–58.
Lakatos-Hayward L. Analytical modeling and characterization of [242] Gulotty R, Castellino M, Jagdale P, Tagliaferro A, Balandin AA.
heat transfer in thermally conductive polymer composites filled Effects of functionalization on thermal properties of single-wall and
with spherical particulates. Composites B 2013;45:43–9. multi-wall carbon nanotube-polymer nanocomposites. ACS Nano
[219] Pruss EA, Wood GL, Kroenke WJ, Paine RT. Aerosol assisted 2013;7:5114–21.
vapor synthesis of spherical boron nitride powders. Chem Mater [243] Wattanakul K, Manuspiya H, Yanumet N. The adsorption of cationic
2000;12:19–21. surfactants on BN surface: its effects on the thermal conductivity
[220] Wang XJ, Zhang LZ, Pei LX. Thermal conductivity augmentation and mechanical properties of BN-epoxy composite. Colloids Surf A
of composite polymer materials with artificially controlled filler 2010;369:203–10.
shapes. J Appl Polym Sci 2014;131:39550/1–39550. [244] Huang XY, Iizuka T, Jiang PK, Ohki Y, Tanaka T. Role of interface
[221] Tsutsumi N, Takeuchi N, Kiyotsukuri T. Measurement of thermal- on the thermal conductivity of highly filled dielectric epoxy/AlN
diffusivity of filler-polyimide composites by flash radiometry. J composites. J Phys Chem C 2012;116:13629–39.
Polym Sci B Polym Phys 1991;29:1085–93. [245] Guo JX, Saha P, Liang JF, Saha M, Grady BP. Multi-walled
[222] Wu H, Drzal LT. High thermally conductive graphite nanoplatelet/ carbon nanotubes coated by multi-layer silica for improving ther-
polyetherimide composite by precoating: effect of percolation and mal conductivity of polymer composites. J Therm Anal Calorim
particle size. Polym Compos 2013;34:2148–53. 2013;113:467–74.
[223] Zhou WY, Qi SH, Tu CC, Zhao HZ, Wang CF, Kou JL. Effect of the [246] Lee B, Dai G. Influence of interfacial modification on the thermal
particle size of Al2 O3 on the properties of filled heat-conductive conductivity of polymer composites. J Mater Sci 2009;44:4848–55.
silicone rubber. J Appl Polym Sci 2007;104:1312–8. [247] Zhou WY. Effect of coupling agents on the thermal conductiv-
[224] Li TL, Hsu SLC. Enhanced thermal conductivity of polyimide films ity of aluminum particle/epoxy resin composites. J Mater Sci
via a hybrid of micro- and nano-sized boron nitride. J Phys Chem B 2011;46:3883–9.
2010;114:6825–9. [248] Hong JP, Yoon SW, Hwang TS, Lee YK, Won SH, Nam JD. Inter-
[225] Kochetov R [Ph.D. dissertation] Thermal and electrical properties of phase control of boron nitride/epoxy composites for high thermal
nanocomposites, including material processing. Delft Netherlands: conductivity. Korea–Australia Rheo J 2010;22:259–64.
Delft University of Technology; 2012, 183 pp. [249] Wang ZH, Lu YL, Liu J, Dang ZM, Zhang LQ, Wang WM. Preparation
[226] Kemaloglu S, Ozkoc G, Aytac A. Thermally conductive boron of nanoalumina/EPDM composites with good performance in ther-
nitride/SEBS/EVA ternary composites: “processing and characteri- mal conductivity and mechanical properties. Polym Adv Technol
zation”. Polym Compos 2010;31:1398–408. 2011;22:2302–10.
[250] Li SS, Qi SH, Liu NL, Cao P. Study on thermal conductive BN/novolac [273] Gomathi A, Harika MR, Rao CNR. Urea route to coat inorganic
resin composites. Thermochim Acta 2011;523:111–5. nanowires, carbon fibers and nanotubes by boron nitride. Mater
[251] Lee GW, Park M, Kim J, Lee JI, Yoon HG. Enhanced thermal conduc- Sci Eng A 2008;476:29–33.
tivity of polymer composites filled with hybrid filler. Composites A [274] Zheng Y, Wang SB. Synthesis of boron nitride coatings on quartz
2006;37:727–34. fibers: thickness control and mechanism research. Appl Surf Sci
[252] Yu JH, Duan JK, Peng WY, Wang LC, Peng P, Jiang PK. Influence 2011;257:10752–7.
of nano-AlN particles on thermal conductivity, thermal stability [275] Hwang Y, Kim M, Kim J. Effect of Al2 O3 coverage on SiC particles
and cure behavior of cycloaliphatic epoxy/trimethacrylate system. for electrically insulated polymer composites with high thermal
Express Polym Lett 2011;5:132–41. conductivity. RSC Adv 2014;4:17015–21.
[253] Hsieh CY, Chung SL. High thermal conductivity epoxy molding com- [276] Im H, Kim J. Effect of homogeneous Al(OH)3 covered MWCNT
pound filled with a combustion synthesized AlN powder. J Appl addition on the thermal conductivity of Al2 O3 /epoxy-terminated
Polym Sci 2006;102:4734–40. poly(dimethylsiloxane) composites. J Mater Sci 2012;47:
[254] Peng WY, Huang XY, Yu JH, Jiang PK, Liu WH. Electrical and 6025–33.
thermophysical properties of epoxy/aluminum nitride nanocom- [277] Chiu HT, Sukachonmakul T, Kuo MT, Wang YH, Wattanakul K.
posites: effects of nanoparticle surface modification. Composites A Surface modification of aluminum nitride by polysilazane and its
2010;41:1201–9. polymer-derived amorphous silicon oxycarbide ceramic for the
[255] Ganguli S, Roy AK, Anderson DP. Improved thermal conductivity for enhancement of thermal conductivity in silicone rubber composite.
chemically functionalized exfoliated graphite/epoxy composites. Appl Surf Sci 2014;292:928–36.
Carbon 2008;46:806–17. [278] Lee ES, Lee SM, Cannon WR, Shanefield DJ. Improved dispersion of
[256] Hirano H. Treatment of inorganic filler surface by silane-coupling aluminum nitride particles in epoxy resin by adsorption of two-
agents: investigation of treatment condition and analysis of bond- layer surfactants. Colloids Surf A 2008;316:95–103.
ing state of reacted agent. Int J Chem Biol Eng 2012;6:29–33. [279] Hong J, Lee J, Hong CK, Shim SE. Effect of dispersion state of carbon
[257] Tan FT, Qiao XL, Chen JG, Wang HS. Effects of coupling agents on nanotube on the thermal conductivity of poly(dimethyl siloxane)
the properties of epoxy-based electrically conductive adhesives. Int composites. Curr Appl Phys 2010;10:359–63.
J Adhes Adhes 2006;26:406–13. [280] Huang XY, Zhi CY, Jiang PK, Golberg D, Bando Y, Tanaka T. Polyhedral
[258] Xu YS, Chung DDL. Increasing the thermal conductivity of boron oligosilsesquioxane-modified boron nitride nanotube based epoxy
nitride and aluminum nitride particle epoxy-matrix compos- nanocomposites: an ideal dielectric material with high thermal
ites by particle surface treatments. Compos Interfaces 2000;7: conductivity. Adv Funct Mater 2013;23:1824–31.
243–56. [281] Terao T, Zhi CY, Bando Y, Mitome M, Tang CC, Golberg D. Align-
[259] Yu G, Zhang MQ, Zeng HM. Carbon-black-filled polyolefine as ment of boron nitride nanotubes in polymeric composite films
a positive temperature coefficient material: effect of composi- for thermal conductivity improvement. J Phys Chem C 2010;114:
tion, processing, and filler treatment. J Appl Polym Sci 1998;70: 4340–4.
559–66. [282] Zhang XL, Shen LY, Wu H, Guo SY. Enhanced thermally conductivity
[260] Teng CC, Ma CCM, Chiou KC, Lee TM, Shih YF. Synergetic effect and mechanical properties of polyethylene (PE)/boron nitride (BN)
of hybrid boron nitride and multi-walled carbon nanotubes on composites through multistage stretching extrusion. Compos Sci
the thermal conductivity of epoxy composites. Mater Chem Phys Technol 2013;89:24–8.
2011;126:722–8. [283] Chan E, Leung SN, Khan MO, Naguib HE, Dawson F, Adinkrah
[261] Wattanakul K, Manuspiya H, Yanumet N. Effective surface treat- V. Novel thermally conductive thermoplastic/ceramic composite
ments for enhancing the thermal conductivity of BN-filled epoxy foams. Macromol Mater Eng 2012;297:1014–20.
composite. J Appl Polym Sci 2011;119:3234–43. [284] Fujihara T, Cho HB, Nakayama T, Suzuki T, Jiang WH, Suematsu H,
[262] Zhang MC, Gu AJ, Liang GZ, Yuan L. Preparation of high ther- Kim HD, Niihara K. Field-induced orientation of hexagonal boron
mal conductive aluminum nitride/cyanate ester nanocomposite nitride nanosheets using microscopic mold for thermal interface
using a new macromolecular coupling agent. Polym Adv Technol materials. J Am Ceram Soc 2012;95:369–73.
2012;23:1503–10. [285] Datsyuk V, Trotsenko S, Reich S. Carbon-nanotube-polymer
[263] Yu JH, Huang XY, Wu C, Wu XF, Wang GL, Jiang PK. Interfacial mod- nanofibers with high thermal conductivity. Carbon 2013;52:605–8.
ification of boron nitride nanoplatelets for epoxy composites with [286] Kaul PB, Bifano MFP, Prakash V. Multifunctional carbon nanotube-
improved thermal properties. Polymer 2012;53:471–80. epoxy composites for thermal energy management. J Compos
[264] Chen C, Tang Y, Ye YS, Xue Z, Xue Y, Xie X, Mai YW. High- Mater 2013;47:77–95.
performance epoxy/silica coated silver nanowire composites as [287] Sihn S, Ganguli S, Roy AK, Qu LT, Dai LM. Enhancement of through-
underfill material for electronic packaging. Compos Sci Technol thickness thermal conductivity in adhesively bonded joints using
2014;105:80–5. aligned carbon nanotubes. Compos Sci Technol 2008;68:658–65.
[265] Badi N, Mekala R. Heat-sink solution through artificial nanodi- [288] Ghose S, Working DC, Connell JW, Smith JG, Watson KA, Delozier
electrics for LED lighting application. In: Proc 2012 COMSOL Conf. DM, Sun YP, Lin Y. Thermal conductivity of UltemTM /carbon
2012. p. 1–5. nanofiller blends. High Perform Polym 2006;18:961–77.
[266] Zhou YC, Bai YY, Yu K, Kang Y, Wang H. Excellent thermal conduc- [289] Fu SY, Mai YW. Thermal conductivity of misaligned short-
tivity and dielectric properties of polyimide composites filled with fiber-reinforced polymer composites. J Appl Polym Sci 2003;88:
silica coated self-passivated aluminum fibers and nanoparticles. 1497–505.
Appl Phys Lett 2013;102:252903/1–252903. [290] Han S, Lin JT, Yamada Y, Chung DDL. Enhancing the thermal con-
[267] Liang QZ, Moon KS, Zhang YL, Wong CP. Low stress and high thermal ductivity and compressive modulus of carbon fiber polymer-matrix
conductive underfill for Cu/low-k application. In: ECTC 2008: 58th composites in the through-thickness direction by nanostruc-
electronic components and technology conference, May 27–30, turing the interlaminar interface with carbon black. Carbon
2008. Lake Buena Vista, FL: IEEE; 2008. p. 1958–62. 2008;46:1060–71.
[268] Choi S, Kim K, Nam J, Shim SE. Synthesis of silica-coated graphite by [291] Agarwal S, Khan MMK, Gupta RK. Thermal conductivity of poly-
enolization of polyvinylpyrrolidone and its thermal and electrical mer nanocomposites made with carbon nanofibers. Polym Eng Sci
conductivity in polymer composites. Carbon 2013;60:254–65. 2008;48:2474–81.
[269] Zhao JC, Du FP, Cui W, Zhu P, Zhou XP, Xie XL. Effect of silica coat- [292] Chen CH, Wang YC. Effective thermal conductivity of mis-
ing thickness on the thermal conductivity of polyurethane/SiO2 oriented short-fiber reinforced thermoplastics. Mech Mater
coated multiwalled carbon nanotube composites. Composites A 1996;23:217–28.
2014;58:1–6. [293] Choy CL, Leung WP, Kowk KW, Lau FP. Elastic-moduli and
[270] Yu S, Park BI, Park C, Hong SM, Han TH, Koo CM. RTA-treated carbon thermal-conductivity of injection-molded short-fiber reinforced
fiber/copper core/shell hybrid for thermally conductive compos- thermoplastics. Polym Compos 1992;13:69–80.
ites. ACS Appl Mater Interfaces 2014;6:7498–503. [294] Zhang SA, Ke YC, Cao XY, Ma YM, Wang FS. Effect of Al2 O3 fibers on
[271] Kim SW, Choi HS, Lee KS. Thermal conductivity of thermally the thermal conductivity and mechanical properties of high density
conductive composites consisting of core–shell particles with polyethylene with the absence and presence of compatibilizer. J
nanostructured shell layers. Mater Res Bull 2014;60:843–8. Appl Polym Sci 2012;124:4874–81.
[272] Yan W, Zhang Y, Sun HW, Liu SW, Chi ZG, Chen XD, Xu JR. Polyimide [295] Feller JF, Roth S, Bourmaud A. Conductive polymer composites:
nanocomposites with boron nitride-coated multi-walled carbon electrical, thermal, and rheological study of injected isotactic
nanotubes for enhanced thermal conductivity and electrical insu- poly(propylene)/long stainless-steel fibers for electromagnetic
lation. J Mater Chem A 2014;2:20958–65. interferences shielding. J Appl Polym Sci 2006;100:3280–7.
[296] Behzad T, Sain M. Measurement and prediction of thermal con- morphology and mechanical property study. Polymer 2002;43:
ductivity for hemp fiber reinforced composites. Polym Eng Sci 3717–25.
2007;47:977–83. [318] Cao JP, Zhao J, Zhao XD, You F, Yu HZ, Hu GH, Dang ZM. High thermal
[297] Wu HJ, Sun XW, Zhang W, Zhang XX, Lu CH. Effect of solid- conductivity and high electrical resistivity of poly(vinylidene flu-
state shear milling on the physicochemical properties of thermally oride)/polystyrene blends by controlling the localization of hybrid
conductive low-temperature expandable graphite/low-density fillers. Compos Sci Technol 2013;89:142–8.
polyethylene composites. Composites A 2013;55:27–34. [319] Cao JP, Zhao X, Zhao J, Zha JW, Hu GH, Dang ZM. Improved thermal
[298] Zhou SX, Zhu Y, Du HD, Li BH, Kang FY. Preparation of oriented conductivity and flame retardancy in polystyrene/poly(vinylidene
graphite/polymer composite sheets with high thermal conductivi- fluoride) blends by controlling selective localization and sur-
ties by tape casting. New Carbon Mater 2012;27:241–9. face modification of SiC nanoparticles. ACS Appl Mater Interfaces
[299] Yan HY, Tang YX, Long W, Li YF. Enhanced thermal conductivity in 2013;5:6915–24.
polymer composites with aligned graphene nanosheets. J Mater Sci [320] Droval G, Feller JF, Salagnac P, Glouannec P. Conductive poly-
2014;49:5256–64. mer composites with double percolated architecture of carbon
[300] Solis KJ, Martin JE. Field-structured magnetic platelets as a nanoparticles and ceramic microparticles for high heat dissi-
route to improved thermal interface materials. J Appl Phys pation and sharp PTC switching. Smart Mater Struct 2008;17:
2012;111:073507/1–73507. 025011/1–25011.
[301] Cho HB, Konno A, Fujihara T, Suzuki T, Tanaka S, Jiang WH, [321] Yorifuji D, Ando S. Enhanced thermal conductivity over percolation
Suematsu H, Niihara K, Nakayama T. Self-assemblies of lin- threshold in polyimide blend films containing ZnO nano-pyramidal
early aligned diamond fillers in polysiloxane/diamond composite particles: advantage of vertical double percolation structure. J
films with enhanced thermal conductivity. Compos Sci Technol Mater Chem 2011;21:4402–7.
2011;72:112–8. [322] Hu X, Jiang L, Goodson KE. Thermal conductance enhancement of
[302] Cho HB, Nakayama T, Suzuki T, Tanaka S, Jiang WH, Suematsu particle-filled thermal interface materials using carbon nanotube
H, Niihara K. Linear assembles of BN nanosheets, fabricated in inclusions. In: ITHERM’04. The ninth intersociety conference on
polymer/BN nanosheet composite film. J Nanomater 2011;2011: thermal and thermomechanical phenomena in electronic systems,
1–7. June 1–4, 2004, IEEE; December 7–9, Singapore: IEEE. 2004. p. 63–9.
[303] Borca-Tasciuc T, Mazumder M, Son Y, Pal SK, Schadler LS, Ajayan [323] Cola BA, Amama PB, Xu XF, Fisher TS. Effects of growth tempera-
PM. Anisotropic thermal diffusivity characterization of aligned ture on carbon nanotube array thermal interfaces. J Heat Transfer
carbon nanotube-polymer composites. J Nanosci Nanotechnol 2008;130:114503/1–114503.
2007;7:1581–8. [324] Shaikh S, Li L, Lafdi K, Huie J. Thermal conductivity of an aligned
[304] Ivanov I, Puretzky A, Eres G, Wang H, Pan ZW, Cui HT, Jin RY, carbon nanotube array. Carbon 2007;45:2608–13.
Howe J, Geohegan DB. Fast and highly anisotropic thermal trans- [325] Cola BA, Xu J, Cheng CR, Xu XF, Fisher TS, Hu HP. Photoacoustic char-
port through vertically aligned carbon nanotube arrays. Appl Phys acterization of carbon nanotube array thermal interfaces. J Appl
Lett 2006;89:223110/1–223110. Phys 2007;101:054313/1–54313.
[305] Raman C. Boron nitride in thermoplastics: effect of loading, particle [326] Cao AY, Dickrell PL, Sawyer WG, Ghasemi-Nejhad MN, Ajayan
morphology and processing conditions. In: Proc NATAS 36th Ann PM. Super-compressible foamlike carbon nanotube films. Science
Conf Therm Anal Appl. 2008. p. 1–60. 2005;310:1307–10.
[306] Tanimoto M, Yamagata T, Miyata K, Ando S. Hexagonal boron [327] Costa VAF, Lopes AMG. Improved radial heat sink for led lamp cool-
nitride filled polyimide films: effects of filler particle size and ori- ing. Appl Therm Eng 2014;70:131–8.
entation on composite thermal conductivity. In: The 32nd Japan [328] Jang D, Park SJ, Yook SJ, Lee KS. The orientation effect for cylindri-
symposium on thermophysical properties, November 21–23, 2011, cal heat sinks with application to LED light bulbs. Int J Heat Mass
Yokohama, Japan. 2011. p. 1–3. Transfer 2014;71:496–502.
[307] Raman C, Murugaiah A, Xiang B, Roden R. Thermally conductive but [329] Naik S, Probert SD, Wood CI. Natural-convection characteristics of
electrically insulating plastics for thermal management applica- a horizontally-based vertical rectangular fin-array in the presence
tions; 2015. http://scholar.google.com.hk/scholar?q=THERMALLY+ of a shroud. Appl Energy 1987;28:295–319.
CONDUCTIVE+BUT+ELECTRICALLY+INSULATING+PLASTICS+FOR+ [330] Anonymous. International technology roadmap for semi-
THERMAL+MANAGEMENT+APPLICATIONS&btnG=&hl=zh-CN&as conductors. 2011 edition: Assembly and packaging; 2014. http://
sdt=0%2C5&as vis=1 [accessed January 2015]. www.itrs.net/Links/2011ITRS/2011Chapters/2011AP.pdf
[308] David L, Tereshko JW, Ajit S, Thomas F, Paulo M. Boron nitride [accessed December 2014].
particles of spherical geometry and process for making thereof. [331] Xu J, Fisher TS. Enhancement of thermal interface materials with
US7976941B2. Albany, NY: Momentive Performance Materials Inc.; carbon nanotube arrays. Int J Heat Mass Transfer 2006;49:1658–66.
2011. [332] Ngo Q, Cruden BA, Cassell AM, Sims G, Meyyappan M, Li J, Yang CY.
[309] Martin E, Krishna U, Jens E. Boron nitride agglomerates, method Thermal interface properties of Cu-filled vertically aligned carbon
of production thereof and use thereof. US20120114905A1. ESK nanofiber arrays. Nano Lett 2004;4:2403–7.
Ceramics, Gmbh Co. KG; 2012. [333] Chuang HF, Cooper SM, Meyyappan M, Cruden BA. Improvement of
[310] Shaffer GW, Frank HR. High density agglomerated boron nitride thermal contact resistance by carbon nanotubes and nanofibers. J
particles. US5898009A. Lakewood, OH: Advanced Ceramics Corpo- Nanosci Nanotechnol 2004;4:964–7.
ration; 1999. [334] Zhu L, Hess DW, Wong CP. Assembling carbon nanotube films as
[311] He H, Fu R, Han Y, Shen Y, Wang D. High thermal conductive Si3 N4 thermal interface materials. In: Proc IEEE 57th Electr Components
particle filled epoxy composites with a novel structure. J Electron Technology Conf. 2007. p. 2006–10.
Packag 2007;129:469–72. [335] Sarvar F, Whalley DC, Conway PP. Thermal interface materials –
[312] Agari Y, Ueda A, Nagai S. Thermal-conductivities of compos- a review of the state of the art. In: Electronics system integration
ites in several types of dispersion-systems. J Appl Polym Sci technology conference. 1st (vol. 2), September 5–7, 2006. Dresden:
1991;42:1665–9. IEEE; 2006. p. 1292–302.
[313] Yu S, Lee JW, Han TH, Park C, Kwon Y, Hong SM, Koo CM. Copper shell [336] Lin W, Wong CP. Vertically aligned carbon nanotubes for thermal
networks in polymer composites for efficient thermal conduction. interface materials: quality control, alignment improvement and
ACS Appl Mater Interfaces 2013;5:11618–22. laser flash measurement. In: Proc IEEE 60th Electr Components
[314] Pashayi K, Fard HR, Lai FY, Iruvanti S, Plawsky J, Borca-Tasciuc T. Technol Conf. 2010. p. 967–72.
Self-constructed tree-shape high thermal conductivity nanosilver [337] Bischak G, Vogdes C. Thermal management design criteria and solu-
networks in epoxy. Nanoscale 2014;6:4292–6. tions. In: Proc IEEE Northcon/98 Conf. 1998. p. 128–33.
[315] Sumita M, Sakata K, Hayakawa Y, Asai S, Miyasaka K, Tane- [338] Liang QZ, Moon KS, Jiang HJ, Wong CP. Thermal conductivity
mura M. Double percolation effect on the electrical-conductivity enhancement of epoxy composites by interfacial covalent bonding
of conductive particles filled polymer blends. Colloid Polym Sci for underfill and thermal interfacial materials in Cu/low-K applica-
1992;270:134–9. tion. IEEE Trans Compon Packag Manuf Technol 2012;2:1571–9.
[316] Matsumura T, Ochi M, Nagata K. Thermomechanical prop- [339] Anonymous. International technology roadmap for semi-
erties, phase structure, and conductivity of organic/inorganic conductors. 2011 edition: Emerging research materials; 2014.
hybrid material filled with a conductive filler. J Appl Polym Sci http://www.itrs.net/Links/2011ITRS/2011Chapters/2011ERM.pdf
2003;90:1980–4. [accessed December 2014].
[317] Thongruang W, Spontak RJ, Balik CM. Bridged double percola- [340] Cola BA. Carbon nanotubes as high performance thermal interface
tion in conductive polymer composites: an electrical conductivity, materials. Electron Cooling Mag 2010;16:10–5.
[341] Narumanchi S, Mihalic M, Kelly K, Eesley G. Thermal interface [347] Haimovich N, Dekel DR, Brandon S. A simulator for system-level
materials for power electronics applications. In: ITHERM 2008: the analysis of heat transfer and phase change in thermal batteries. J
11th intersociety conference on thermal and thermomechanical Electrochem Soc 2009;156:442–53.
phenomena in electronic systems. May 28–31, 2008. Orlando, FL: [348] Royne A, Dey CJ, Mills DR. Cooling of photovoltaic cells under con-
IEEE; 2008. p. 395–404. centrated illumination: a critical review. Sol Energy Mater Sol Cells
[342] Prasher R. Thermal interface materials: historical perspective, sta- 2005;86:451–83.
tus, and future directions. Proc IEEE 2006;94:1571–86. [349] Brinkworth BJ, Cross BM, Marshall RH, Yang HX. Thermal regulation
[343] Koo B, Goli P, Sumant AV, Claro PCD, Rajh T, Johnson CS, of photovoltaic cladding. Sol Energy 1997;61:169–78.
Balandin AA, Shevchenko EV. Toward lithium ion batter- [350] Krauter S, Araujo RG, Schroer S, Hanitsch R, Salhi MJ, Triebel
ies with enhanced thermal conductivity. ACS Nano 2014;8: C, Lemoine R. Combined photovoltaic and solar thermal sys-
7202–7. tems for facade integration and building insulation. Sol Energy
[344] Maleki H, Selman JR, Dinwiddie RB, Wang H. High thermal conduc- 1999;67:239–48.
tivity negative electrode material for lithium-ion batteries. J Power [351] Kalogirou SA, Tripanagnostopoulos Y. Hybrid PV/T solar systems
Sources 2001;94:26–35. for domestic hot water and electricity production. Energy Convers
[345] Khan WS, Asmatulu R, Rodriguez V, Ceylan M. Enhancing thermal Manag 2006;47:3368–82.
and ionic conductivities of electrospun PAN and PMMA nanofibers [352] Kinsey GS, Edmondson KM. Spectral response and energy out-
by graphene nanoflake additions for battery-separator applica- put of concentrator multijunction solar cells. Prog Photovolta
tions. Int J Energy Res 2014;38:2044–51. 2009;17:279–88.
[346] Al Hallaj S, Selman JR. A novel thermal management system for elec- [353] Lee B, Liu JZ, Sun B, Shen CY, Dai GC. Thermally conductive and
tric vehicle batteries using phase-change material. J Electrochem electrically insulating EVA composite encapsulants for solar pho-
Soc 2000;147:3231–6. tovoltaic (PV) cell. Express Polym Lett 2008;2:357–63.

1 s2.0 S0079670016000216 Main

Uploaded by

Copyright:

Available Formats

1 s2.0 S0079670016000216 Main

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1 s2.0 S0079670016000216 Main

Uploaded by

Copyright:

Available Formats

Progress in Polymer Science 59 (2016) 41–85

Contents lists available at ScienceDirect

Progress in Polymer Science

Thermal conductivity of polymer-based composites:

∗ Corresponding author. Tel.: +86 21 38513128; fax: +86 21 58954173.

3. Thermal conductivity measurement techniques. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .51

PE polyethylene Thermal dissipation is critical to the performance,

interfacial resistance; however, they are often time-

2.1. Constitutive (micromechanical) modeling

The simplest (and crudest) approach – that often works

kc = (1 − )kp + kf (4) where ı = kf /kp . Dependence of the composite TC on the

Fig. 3. Bruggeman model prediction of composite TC as a function of par-

Fig. 4. Schematic of well-dispersed aggregates in composite materials. A

that provide a general description of the method and its

∇ ((r)∇ T (r)) = 0 (12)

subject to the boundary condition,

Fig. 6. Comparison between experimental data (line) and FEA simulations

conduction, making polymer/CNT nanocomposites reason-

Methods Temperature Conductivity range Accuracy Materials Test standards

3.2. Transient methods

The transient methods are generally used to measure

where k is the thermal conductivity, ˛ is the thermal diffu-

electrical current is applied to the wire. Thermal conduc- Table 2

4.3. Effect of orientation

Heat is transferred more easily along the chains than

4.3.1. Orientation of crystalline/semicrystalline polymers

ﬁbers and extremely high TC. However, the TC|| of ultra-

4.3.3. Orientation of liquid crystalline polymers

(0.3 W/(m K) in the other two directions). However, only

4.4. Polymer blends by engineered interchain

As discussed in the previous sections, molecular struc-

Thermosets Monomer Hardener Thermal conductivity

BACE – 0.29 [149]

DGEHAP 0.33 [152]

DGETAM 0.38 [147]

DGEHH 0.39 [153]

DGEDH 0.50 [154]

BZE 0.69 [148]

TME8 0.83–0.85 [145,148]

Fig. 20. Possible schematic image of smectic microstructure from twin-

microscopic anisotropic liquid crystalline domains due to

5.2. Effect of curing condition

Fillers Category TC W/(m K) Electrically conductive? Refs.

Aluminum Metal 234 Yes [167]

Carbon-based materials, including graphite [90], 6.3. Ceramic ﬁllers

7.2.2. Platelet-like ﬁllers 7.3. Size effect

Table 5 polygonal AlN and platelet-shaped BN were added into

7.5. Surface treatment

As inorganic ﬁllers are dispersed in organic polymer

conductivity. Above the critical concentration, however,

The shell materials do not necessarily have to be elec-

orientation approaches to orient ﬁllers into the desired

8.1.1. Orientation by shear or extensional ﬂow during

8.1.2. Orientation by external ﬁelds

advantage of ﬁeld structuring is that without increasing the

blends. The TC of the blends were signiﬁcantly improved

9.3. Batteries 10. Conclusions and future prospects

You might also like