You are on page 1of 42

Materials Science & Engineering R 142 (2020) 100577

Contents lists available at ScienceDirect

Materials Science & Engineering R


journal homepage: www.elsevier.com/locate/mser

Thermal conductivity of graphene-based polymer nanocomposites T


a, b a c d a,
Xingyi Huang *, Chunyi Zhi , Ying Lin , Hua Bao , Guangning Wu , Pingkai Jiang *,
Yiu-Wing Maie,*
a
Shanghai Key Laboratory of Electrical Insulation and Thermal Ageing, State Key Laboratory of Metal Matrix Composites, Shanghai Jiao Tong University, Shanghai
200240 China
b
Department of Materials Science and Engineering, City University of Hong Kong, Tat Chee Avenue, Kowloon, Hong Kong, China
c
University of Michigan - Shanghai Jiao Tong University Joint Institute, Shanghai Jiao Tong University, Shanghai 200240, China
d
School of Electrical Engineering, Southwest Jiaotong University, Chengdu, China
e
Centre for Advanced Materials Technology (CAMT), School of Aerospace, Mechanical and Mechatronic Engineering J07, The University of Sydney, Sydney, NSW 2006,
Australia

ARTICLE INFO ABSTRACT

Keywords: As a material possessing extremely high thermal conductivity, graphene has been considered as the ultimate
Graphene filler for fabrication of highly thermally conductive polymer composites. In the past decade, graphene and its
Polymer composites derivatives were demonstrated in many studies to be very effective in enhancing the thermal conductivity of
Thermal conductivity various polymers. This paper reviews current progress in the development of graphene/polymer composites with
Surface modification
high thermal conductivity. We began with the effects of isotopes, defects/doping, edges and substrate, poly­
crystallinity, functionalization, size and layer number, and folding/twisting on the thermal conductivity of
graphene. We then modelled the thermal conductivity of graphene/polymer composites and, through molecular
dynamics (MD) simulations, demonstrated its dependence on interfacial thermal conductance as well as size,
dispersion and volume fraction of graphene. After a critique of recent studies on thermally conductive graphene/
polymer composites and their potential applications, we identified several outstanding issues, new challenges
and opportunities for future endeavours.

1. Introduction new findings of heat-flow physics which potentially lead to novel


thermal management applications. Graphene with a perfect structure
Thermally conductive polymer composites are attracting consider­ has almost the highest thermal conductivity (TC) among all known
able attention especially in recent years because increasingly more materials (> 5000 W/m·K that is on par with diamond materials) [7,8].
powerful electronics are being developed. This class of materials can be Also, its 2D morphology and super-high aspect ratio make it suitable as
comprehensively used as electronic packaging materials, thermal in­ a filler for polymers to achieve superbly high TC [9–12].
terface materials (TIMs) and base materials for printed circuit broads, However, when applying graphene as a filler in polymer matrices,
etc. [1]. These thermally conductive polymer composites consist of a many scientific and technological issues appear and the advantages of
polymer matrix and thermally conductive fillers. Their final thermal graphene as a highly thermally conductive filler cannot be realized.
conductivities are determined by the processing techniques and ther­ Indeed, in some situations, it is found that thick graphite sheets or
mally conductive fillers which can be metal powder, ceramic powder platelets can deliver even better TC results than graphene [13].
(mainly for thermally conductive but electrically insulating composites) There are two issues of highly thermally conductive graphene/
and carbon materials [2,3]. Among these, carbon materials are the most polymer composites. One issue is due to the graphene itself. Super-high
used and intensively studied fillers with excellent thermal con­ TC can only be achieved if there is a perfect graphene structure; the
ductivities, different morphologies, low densities and inexpensive costs. presence of any edges, defects, doping and modifications gives dramatic
These carbon-based fillers incorporated in polymers include graphite drops in the intrinsic TC [7,14]. The other issue arises from the fabri­
platelets, carbon nanotubes (CNTs) and graphene [4–6]. cation of polymer composites, which includes poor dispersion of gra­
The unusual thermal properties of graphene originate from its un­ phene due to its 2D morphology and high aspect ratio, interfacial and
ique two-dimensional (2D) nature, giving a solid platform for different contact thermal resistances, difficulties of constructing 3D structures,


Corresponding authors.
E-mail addresses: xyhuang@sjtu.edu.cn (X. Huang), pkjiang@sjtu.edu.cn (P. Jiang), yiu-wing.mai@sydney.edu.au (Y.-W. Mai).

https://doi.org/10.1016/j.mser.2020.100577
Received 18 May 2020; Received in revised form 29 June 2020; Accepted 3 July 2020
Available online 22 September 2020
0927-796X/ © 2020 Elsevier B.V. All rights reserved.
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

Nomenclature NMP N-methyl-2-pyrrolidone


NR natural rubber
1D one-dimensional PA polyamide
2D two-dimensional PA6 polyamide 6
3D three-dimensional PAN polyacrylonitrile
ABS acrylonitrile-butadiene-styrene PBT polybutylene terephthalate
ATRP atom transfer radical polymerization PC polycarbonate
BN boron nitride PCC phase change composite
CB carbon black PCF pitch-based carbon fiber
CNT carbon nanotube PCM phase change material
CNF carbon nanofiber PDMS polydimethylsiloxane
CPU central processing unit PE polyethylene
CVD chemical vapour deposition PEG polyethylene-glycol
DMF N, N-dimethylformamide PEI polyether imide
DP degree of polymerization phr per hundred parts of resin
E-51 bisphenol-A epoxy PGMA poly(glycidyl methacrylate)
GRMs graphene-materials PI polyimide
EG expanded graphite PMMA polymethyl methacrylate
EMD equilibrium molecular dynamics POSS polyhedral oligomeric silsesquioxane
EP epoxy resin PP polypropylene
EVA ethylene-vinyl acetate copolymer PPS polyphenylene sulphide
f-GF functionalized graphene flake PS polystyrene
f-GO functionalized graphene oxide PTFE polytetrafluoroethylene
FLG few-layered graphene PU polyurethane
f-RGO functionalized reduced graphene oxide Py pyrene
GF graphene flake PVA poly(vinyl alcohol)
GMP graphite microparticle PVDF poly(vinylidene fluoride)
GNP, GnP graphene nanoplatelet RGO, rGO reduced graphene oxide
GNR graphene nanoribbon SCA silane coupling agents
GO graphene oxide SCF supercritical fluid
GrO graphite oxide SLG single-layered graphene
GS graphene sheet SWCNT single-walled carbon nanotube
HDPE high density polyethylene TC thermal conductivity
KH-560 γ-(2,3-epoxypropoxy)propytrimethoxysilane TCA titanate coupling agents
LLDPE linear low-density polyethylene TCE thermal conductivity enhancement
MD molecular dynamics THF tetrahydrofuran
MLG multi-layered graphene TIM thermal interface material
MSA methanesulfonic acid VDOS vibrational density of states
MWCNT multi-walled carbon nanotube VPS vibration power spectrum
NEMD non-equilibrium molecular dynamics XRD X-ray diffraction
NFG natural flake graphite

etc. as illustrated in Fig. 1. experimental results and advances obtained on the TC of polymer
This paper is a critical review on thermally conductive polymer composites filled with graphene materials and graphene-based hybrids.
composites with graphene materials as fillers. Herein, graphene mate­ Here, graphene-based hybrids are defined as one graphene material
rials (GRMs) are defined as 2D, flake- or sheet-like carbon forms, which plus another filler, and in some cases, graphene-based hybrids also
include few-layer graphene (FLG), graphene flake (GF), graphene na­ mean two different graphene materials. Issues on the fabrication tech­
noflake (GNF), graphene nanoplatelet (GNP, GnP), graphene oxide niques are highlighted. While the practical applications of these highly
(GO), graphene nanoribbon (GNR), multi-layered graphene (MLG), re­ thermally conductive graphene/polymer composites are not mature
duced graphene oxide (rGO, RGO) and single-layered graphene (SLG) enough yet, Section 6 shows three potential applications as graphene-
[15] as well as three-dimensional (3D) forms (e.g., aerogels and foams) enabled TIMs, phase change materials (PCMs) and separators for ef­
of these 2D carbon materials. Other 2D carbon materials such as ex­ fective thermal management of electronic devices and lithium batteries.
panded graphite (EG), exfoliated graphite (ExG), graphite micro-par­ Finally, in Section 7, we summarize the current progress made in the
ticle (GMP) and natural flake graphite (NFG) are not considered as field and identify some unsolved problems, pointing to future research
graphene materials. However, their composites will be reviewed, as directions.
appropriate, for comparative studies. All potential factors which affect
the TC of fabricated polymer composites, and their current and emer­
2. TC of graphene
ging applications will be discussed (Fig. 1). In Section 2, we first
summarize recent progress on the intrinsic TC of graphene materials
Graphene has an extremely high in-plane TC, probably the highest
and the main factors which reduce their TC. Section 3 is focused on
in comparison with any currently known material. Fig. 2 shows a brief
theoretical studies and molecular dynamics (MD) simulations on TC of
summary of experimental TCs of some typical carbon structures and
graphene/polymer composites. The effects of interfacial thermal re­
metals. Although the values reported in the literature are quite di­
sistance due to graphene functionalization, size, volume fraction and
versified, the figure clearly shows extremely high TCs of carbon-based
dispersion are included. Then, in Sections 4 and 5, we discuss the
materials. Experimentally, at room temperature, a free-standing

2
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

Fig. 1. Factors which affect the fabrication of thermally conductive graphene/polymer composites to maximize their thermal conductivity (TC) and TC enhancement
(TCE) and their prospective applications. The flowchart also shows the framework of this review article.

graphene sheet exhibits an in-plane TC of up to 1500−5400 W/m·K


[16–24], which is even higher than that of diamond materials (∼
2000 W/m·K) [25,26]. It should be noted that the highest in-plane TC of
graphene claimed is achieved in isotopically purified samples with large
grains; whereas defects, small grains and edges will dramatically reduce
the TC values [16,17].
Graphene is both thermally and electrically conductive. Thus, both
electrons and phonons contribute to its high TC. However, instead of
electrons, phonon behaviours dominate the thermal transport of gra­
phene [28,29]. The mechanism of the ultrahigh TC value of graphene
has been discussed in detail in the published literature. In short, its high
TC is mainly attributed to two mechanisms. First, the sp2 orbital hy­
bridization and light carbon atoms provide a very high phonon group
velocity and extremely harmonic lattice. Second, the 2D lattice allows
through-plane flexural (Z) phonon vibrations and these modes have
small scattering rates owing to the reflectional symmetry of the gra­
Fig. 2. Summary of room-temperature experimental TC values of diamond phene lattice [30,31]. These two features yield an extremely high TC of
[25,26], graphite (in-plane) [16,17,23,27], CNT [5], suspended graphene graphene, which is also very sensitive to any phonon scattering.
[16,19] and some metal materials. The very high TC of graphene makes it very suitable for fabrication
of thermally conductive polymer composites. Except interface issues
which are commonly considered for polymer composites, factors

Fig. 3. TC of graphene as a function of its iso­


topic composition. (a) TC of graphene with 13C
isotope concentrations of 0.01 %, 1.1 % (nat­
ural abundance), 50 % and 99.2 % measured by
the micro-Raman method. (b) Average value of
measured TC as a function of 13C concentration
at ∼380 K. MD simulation results for TC are
shown as yellow squares for comparison
(Reproduced with permission from Ref. [17]).
(For interpretation of the references to colour in
this figure legend, the reader is referred to the
web version of this article.)

3
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

affecting the TC of graphene itself are also important. External per­ of defect types (substitutional N and Si, structurally different pyridinic
turbations are known to have strong effects on the intrinsic TC of low- N, pure structural single vacancy and Stone–Wales defects as shown in
dimensional materials [32]. Below, we will concisely examine and Fig. 5) on TC of 2D graphene [48]. The structural defects give com­
discuss major results of previous studies on the TC of graphene-based parable reductions in TC because their scattering effects are sig­
materials and their applications in polymer composites. Interested nificantly weakened by the 2D feature of graphene. By contrast, the
readers may refer to published reviews on this topic [7,33]. substitutional Si defect yields a larger reduction of TC. Unlike the effect
of substitutional N defect, pyridinic N leads to a higher reduction of TC
2.1. Isotope effects compared to single vacancy defects.
The effect of single and multiple defects/doping (single vacancy,
Carbon has many isotopes with 12C and 13C as most typical. CH3 functional group and N-doping) on the TC of MLG was investigated
Different mass of carbon isotopes results in different phonon feature through equilibrium MD (EMD) simulation [49] at 300, 700 and 1500
and scattering process, which affect the TC of graphene as it is domi­ K. When only one single type of defect/doping is present, the single
nated by lattice vibrations instead of electrons as mentioned earlier. vacancy defect gives the largest effect on TC reduction at 300, 700 and
Most studies on isotope effects regarding TC of graphene were 1500 K; however, the effect of N-doping is the least. When the graphene
performed by theoretical calculations. In a typical MD simulation study, sheets contain three types of defects/doping, the single vacancy defect
it was found that the TC of graphene with 5 % 13C is 12 % lower than produces a most significant effect on TC reduction at 300 and 700 K,
that of pure 12C graphene caused by the so-called phonon localization while the CH3 functional group dominates TC reduction at 1500 K.
[34]. In another MD simulation work [35], it is confirmed that isotope Besides, compared to the effect of a single type of defects/doping, the
mixing can reduce TC of graphene and super-lattice distribution gives presence of multiple defects/doping can further reduce the lattice TC at
higher reduction than random distribution. 300 and 700 K.
Chen et al. conducted experiments to evaluate the isotope effects on
TC of graphene [17]. Graphene with different percentage of 13C was
2.3. Effects of edges and substrate
fabricated on a copper foil by chemical vapour deposition (CVD) and
their thermal conductivities were determined by the optothermal
GNRs with smooth edges possess higher TC than those with rough
Raman technique. It was revealed that isotopically pure 12C (0.01 %
13 edges. Rough edges can induce more diffusive phonon boundary scat­
C) graphene has a TC value up to > 4000 W/m·K at ∼320 K, which is
tering and hence hinder more effectively phonon transport [50–52].
more than two times that in graphene comprising a 50:50 mixture of
12 Because graphene has small intrinsic phonon scattering, the edge
C and 13C (see Fig. 3a). These results agree with those obtained by MD
roughness can effectively lower the TC of graphene, especially for GNRs
simulations (see Fig. 3b). Natural graphite has 12C:13C of roughly 99:1.
with small widths [28,50,53]. As shown in Fig. 6, a rough edge can
According to this study [17], natural graphene has a TC ∼2700 W/m·K
reduce TC of GNR by one order of magnitude. This is because for GNRs
at 300 K, which is high enough for fabrication of thermally conductive
with rough edges, only the phonon modes with wavelengths compar­
polymer composites.
able to the ribbon width can contribute to heat transport, and a ma­
jority of oscillatory eigen-modes are localized and do not contribute to
2.2. Effect of defects/doping
the TC. Concerning the two typical edges (i.e., zigzag and armchair),
MD simulations show that TC of graphene depends on the edge chirality
All theoretical studies have revealed that defects and doping can
and GNRs with zigzag edges possess 20–50 % higher TC than that of
yield remarkable reductions of TC of graphene, which is attributed to
armchair edges induced by different phonon scattering rate. Though the
phonon scattering that greatly shortens the phonon mean free path
edge effect is relatively strong in extremely narrow GNRs where phonon
[36–39]. Yeo et al. [40] confirmed the presence of Stone-Wales defects
boundary scattering dominates, this effect is expected to be much less
(pairs of opposing 5-membered or 7-membered rings formed by just
prominent for wider GNRs.
rotating a pair of carbon atoms) or double vacancy defects would de­
The interaction between graphene and substrate may markedly af­
crease TC of graphene by over 80 % as the defect densities were in­
fect its TC. Yu and Zhang reported that if a graphene is suspended
creased to 10 % coverage for both zigzag- and armchair-structured
between two metal blocks (substrates), the heat flux can only be 40 %
GNRs. In zigzag-structured GNRs, double vacancies in a nanoribbon led
that of ideally suspended graphene (no contact to anything, i.e., no
to more significant reductions in TC than Stone-Wales defects. These
substrate) [54]. Guo et al. revealed that a SiC substrate has strong
trends are obvious in the temperature range 100−600 K. Other studies
also reported similar results [28,41–43]. For example, Ng et al. [44]
showed that Stone-Wales defects could decrease TC of graphene by over
50 %, with much higher reductions in zigzag than armchair GNRs for all
defect densities. Malekpour et al. [45] also found that TC of graphene
decreased [∼(1.8 ± 0.2)×103 to ∼(4.0 ± 0.2)×102 W/m·K] with in­
creasing defect density near ambient temperature at relatively low
densities (2.0 × 1010 to 1.8 × 1011/cm2). However, with higher defect
densities, the TC of graphene showed a weak dependence on defect
density at a relatively high value of ∼400 W/m·K.
Similar to defects, doping can greatly enhance phonon scattering,
resulting in a lower TC. N-doping has been intensively studied for
graphene by theory. Substituting only 1 % of N atoms in the graphene
lattice may lead to more than 50 % reduction of TC [46] (Fig. 4). It is
interesting that N-doped graphene can decrease dramatically the TC
dependence on chirality due to the scattering of phonons caused by the
high loading of N atoms concealing the chirality effect. In another
study, it has been found that N-doped GNRs exhibit higher TC than
GNRs with defects alone; but the TC of both are much lower than the Fig. 4. Effect of nitrogen atom (N) concentration on the normalized TC of SLG
pristine defect-free graphene [47]. along the armchair and zigzag chirality directions. NB: This figure is redrawn
Using non-equilibrium (NE) MD simulation, Lee compared the role based on the original data. (Reproduced with permission from Ref. [46]).

4
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

Fig. 5. (a) Defect-free graphene andgraphene sheets with (b) quaternary N, (c) pyridinic N, (d) single vacancy (SV) and (e) Stone-Wales (SW) defects. (NB: This figure
is redrawn based on the original [48]). TC values of MLG with single kind of defect including (f) 1-4 % N atoms, (g) 2-8 % methyl group, and (h) 1-4 % single vacancy
at 300, 700 and 1500 K. (Reproduced with permission from Ref [49].).

interaction with SLG during thermal transport, which greatly reduces definitive conclusion from existing data published in different papers.
the TC. However, high TC can be obtained for the second layer of a For example, the TC of graphene on SiO2 substrate was measured to be
suspended double-layered graphene due to the weak van der Waals ∼ 1800−2200 W/m·K [57], but experiments showed that suspended
force between the two layers [55]. Hence, weak interactions between graphene had a TC of only ∼ 100−1000 W/m·K [58]. It is believed that
graphene and substrate is critical to maintaining the high TC of gra­ the quality of graphene, measurement method and associated large
phene. Also, the interaction of graphene with substrate effectively re­ errors may have overwhelmed the substrate/edge effects although they
duces the length dependence of TC, that is, the TC of graphene on are theoretically predicted to be significant.
substrate is no longer sensitive to its length [56].
No systematic experimental studies on substrate/edge effects on TC
of graphene have been reported to date. It is impossible to draw a

Fig. 6. (A) Examples of hydrogen-terminated


zigzag nanoribbons with rough edges for the
density of edge layers where: (a) d = 0, (b) d
= 0.1, (c) d = 0.5, (d) d = 0.9 and (e) d = 1.
Here, “d” is defined as the ratio of the number
of atoms in the edge layers to four times the
dimensionless length of the nanoribbon
(number of longitudinal cells). (B) Dependence
of TC of a finite nanoribbon with rough edges
on density d. TC is calculated from temporal
evolution during 3 ns. (Reproduced with per­
mission from Ref. [53]).

5
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

2.4. Effect of polycrystallinity reduce the TC of graphene. For example, randomly distributed methyl
and phenyl groups are found to decrease TC of graphene of up to 50 %
CVD is by far the most powerful technique to prepare industry-scale with a functional group coverage regime of as few as 1.25 % [68] (see
graphene sheets, which have a polycrystalline rather than mono­ Fig. 8). Detailed analyses show the rapid TC drop depends closely on
crystalline structure [59]. Polycrystalline graphene sheets contain in­ the angular momentum of the functional groups. That is, the high an­
evitably grain boundaries, grain misorientation and other structures, gular momentum rotates the unsupported sp3 bonds and thus the
which can cause severe phonon scattering than that of single crystalline phonon mean free paths are reduced.
graphene sheets [60–62]. Many studies have illustrated that the TC of The effect of alkyl functional groups on the TC of GO sheets was
polycrystalline graphene depends primarily on its grain size. Lee et al. investigated by Aref et al. [69] by using a reverse NEMD simulation
reported the in-plane thermal conductivities of polycrystalline gra­ method. The results show that, since functionalization of GO has in­
phene on Si3N4 substrate at 320∼510 K as 2600∼1230, 1890∼1020 tensified phonon scattering [69], its TC decreases dramatically after
and 680∼340 W/m·K for average grain size of 4.1, 2.2 and 0.5 μm, alkylation and deteriorates with increasing chain length of the func­
respectively, which are much lower than the best fitted values tional groups. In another NEMD simulation study on fluorinated gra­
(5500∼1980 W/m·K obtained by the simple linear chain model in the phene [70], Huang et al. found that the TC was dependent on the
same temperature range) of single crystalline graphene [63]. Fig. 7a structure (zigzag or armchair) of graphene and the distribution of
shows that, as grain size decreases from infinity to 0.5 μm, significant fluorine atoms. Hence, TC decreases rapidly with increasing fluorine
TC reductions are observed. This result is mainly caused by the in­ coverage from 0 % to 20 %, stabilizes from 20 % to 70 %, and then
creased grain boundaries and higher defect concentration as the grain increases quickly when the coverage approaches 100 %. These results
size is decreased (see Fig. 7b). agree with the fact that the functionalization induced TC reduction is
Lee et al. [64] studied the effect of grain misorientation angle on in- caused by phonon scattering at the interface of fluorinated and non-
plane TC of polycrystalline graphene. Their results showed that thermal fluorinated graphene regions, and when fluorine coverage increases to
conduction in graphene also depends mainly on grain size. A slight 100 %, there are no interfaces. However, it must be noted that the fully
misorientation (< 4°) between two adjacent grains would cause a large fluorinated graphene possesses only 40 % the TC of pristine graphene.
decrease in the TC of bi-crystalline graphene. Since phonon scattering Apart from the mono-layer graphene, it has also been confirmed that
in polycrystalline graphene is determined by the grain boundaries and functional groups can also reduce the TC of bi-layer graphene (BLG)
corresponding defects, the TC of graphene can be tailored by control­ sheets [71].
ling the nucleation densities by adjusting the CVD process parameters,
such as operating temperature and partial pressure.
2.6. Effects of size and layer number

2.5. Effect of functionalization of graphene Chen et al. experimentally investigated TC of CVD fabricated SLG
with different diameter ranging from 2.9 to 9.7 μm, and obtained TC of
Functionalization of graphene is very important to improving its suspended graphene between 2600–3100 W/m·K with no clear size ef­
dispersion in composite materials and its affinity with polymer ma­ fect [16]. This observation is attributed to the relatively large mea­
trices, which may finally impart an effectively enhanced performance surement uncertainty as well as grain boundaries, wrinkles, defects,
[65]. However, functionalization is a double-edged sword because the and/or polymeric residue which are possibly present in the tested
functional groups attached on the graphene surface are found to reduce samples.
remarkably its inherently high TC [66]. This is expected since any Theoretically, the lateral size of a graphene ribbon is critical to
randomly distributed functional groups act as molecular doping to maintain high TC of graphene since the edge can scatter phonons. A
graphene, destroying the perfect 2D crystallinity of graphene and giving recent experimental study showed that for porous graphene nano-mesh
unavoidable phonon scattering caused by the sp3 bonds created by with feature size of ∼10 nm, the thermal conduction would be dra­
functionalization different from the typical sp2 bonds. matically altered due to phonon-edge scattering. The neck width be­
The smallest “functional group” is a hydrogen atom. Unfortunately, tween pores is critical. Specifically, the TC of graphene nano-mesh with
it is shown that even a hydrogen coverage of 2.5 % may decrease the TC 8 nm neck width can be as low as ∼78 W/m·K that is the lowest value
of graphene by ∼40 % [67]. Large size groups may be more severe to for suspended graphene obtained by MD simulation [72]. Mahdizadeh

Fig. 7. (a) TC versus measured temperature for suspended graphene on Si3N4 substrate in air with grain sizes of 0.5, 2.2 and 4.1 μm. (b) SEM images of graphene with
different grain size synthesized at controlled operating temperature and pressure. (Reproduced with permission from Ref. [63]).

6
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

Fig. 8. Effect of degree of (a) −CH3 and (b)


–C6H5 functionalization on TC. S is the degree
of functionalization and defined as the ratio of
sp3 hybridized carbon atoms to the total carbon
atoms. The error bars are obtained from mul­
tiple MD simulations of statistically in­
dependent functionalized GNRs for a given
functional group coverage regime.
(Reproduced with permission from [68]).

and Goharshadi studied the heat transport of a series of GNRs with 2 nm


width and 2−30 nm length [73]. Their results show that the TC decays
exponentially with GNR length from 3500 W/m·K at 2 nm to a limiting
value of ∼1500 W/m·K at 30 nm. Majee and Aksamija obtained TC of
suspended GNRs with a constant ribbon width, showing a logarithmic
divergence for ribbon length < 100 μm [74]. By contrast, Nika et al.
found that the long mean free path of long wave-length acoustic pho­
nons in graphene could lead to an unusual non-monotonic dependence
of TC on ribbon length (Fig. 9a) [75]. For a graphene ribbon with a
width of 5 μm, the maximum TC occurs at a ribbon length of ∼100 μm,
then gradually reduces to a steady constant value at a length of ∼1000
μm. Detailed discussion reveals that the TC-GNR length relationship is
actually associated with the ribbon width. With infinitely wide ribbons,
the monotonic dependence of TC on length is obtained (Fig. 9b). Be­
sides, the TC of graphene ribbon depends on its width. Unlike the length
effect, the mechanism of width effect in graphene is mainly due to the
limitation of the directional edge on phonon transmission [76].
Layer number is obvious for graphene to be distinct from graphite.
Fig. 10. Measured TC as a function of the number of atomic planes in FLG. The
Thus, it is believed that interlayer phonon coupling affects TC of gra­ dashed straight lines indicate the range of bulk graphite thermal conductivities.
phene. Cao et al. reported monotonously decreasing TC for graphene The blue diamonds were obtained from the first-principle theory of thermal
with increasing layers [77]. However, calculation gives an un­ conduction in FLG based on actual phonon dispersion and accounting for all
reasonably low TC of graphene with 500 W/m·K for single-layer gra­ allowed three-phonon umklapp scattering channels. The green triangles are
phene and ∼300 W/m·K for 5-layer graphene [77]. For bi-layer GNRs, Callaway–Klemens model calculations, which include the extrinsic effects
Liu et al. [78] showed that the TC depends weakly on the stacking characteristic of thicker films (Reproduced with permission from Ref. [19]).
pattern, with the thermal conductivities of AA-stacked bi-layer GNRs (For interpretation of the references to colour in this figure legend, the reader is
slightly higher than those of AB-stacked. However, again low TC values referred to the web version of this article.)
of 100−500 W/m·K of these graphene ribbons are obtained. Such small
values may be due to the small domains used in MD modelling. conduction crossover is clearly demonstrated as the number of atomic
In an experimental study, Ghosh et al. [19] found that, at room- planes changes from 2 to about 8. Sadeghi et al. conducted a systematic
temperature, the TC changes from ∼2800 to 1300 W/m·K as the investigation on TC of MLG on an amorphous substrate. It is found that
number of layers in graphene increases from 2 to 4 (see Fig. 10). This even after the graphene thickness is increased to 34 layers, its TC is still
phenomenon can be explained by the through-plane coupling of the higher than that of graphite [79]. Hence, it is believed that MLG can be
low-energy phonons, and changes in the phonon umklapp scattering. even better than graphite platelets in order to achieve high TC in
The TC obtained is reasonable and the 2D graphene to 3D graphite heat polymer composites.

Fig. 9. (a) Dependence of TC of rectangular


GNR on ribbon length L shown for different
specularity parameters p. The specularity
parameter determines the fraction of diffu­
sively scattered phonons at the edges and is
related to the edge roughness. The width is
fixed at d = 5 μm. (b) Dependence of TC of
rectangular graphene ribbon on the ribbon
length L shown for different ribbon width d.
The specularity parameter is fixed at p = 0.9.
Note in both (a) and (b) an unusual non-
monotonic length dependence of TC, which
results from the exceptionally long mean free
path of the low-energy phonons and their
angle-dependent scattering from the ribbon
edges (Reproduced with permission from Ref.
[75]).

7
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

2.7. Effects of folding, twisting and wrinkles 3. Modelling TC of graphene/polymer composites

The common deformations of graphene (e.g., folding, twisting and Developing composite materials with desirable properties through
wrinkles) may remarkably affect its TC. Higher flatness ratios lead to traditional trial-and-error method is not only time-consuming and
higher TC. A low flatness ratio causes substantial degradation in TC costly but also difficult to obtain valuable insights on materials design.
[80], which in turn may markedly limit the thermal conductivity en­ An optimum way toward real high-performance composites is to clearly
hancement (TCE) in graphene/polymer composites since twisting, define the factors determining the properties of a composite through
folding and wrinkles of high aspect ratio graphene cannot be totally analytical and numerical simulations and then to fabricate the com­
avoided in the matrix. posites by taking all the factors into full account. That is, analytical and
It has been shown that folding can severely modulate the thermal numerical simulations can inform what should be done to optimize and
transport properties of GNRs [81]. TC may be greatly decreased up to what is not necessary from the scientific perspective.
70 % in comparison with that of its flat counterpart although no defects Traditionally, the TC of composite materials can be modelled using
are actually introduced to the graphene during folding. The more the some homogenization approach, e.g., effective medium theory. The
ribbons are folded, the more the thermal conductivities are reduced effective medium theory was first developed analytically to obtain ef­
because of the strong scattering of low frequency phonon modes at the fective TC values of homogeneous spherical particle filled composites,
folds (see Fig. 11). In a study by NEMD simulation [82], a more drastic and then gradually extended to model different types of fillers [71].
effect is illustrated. Although perfect graphene may possess a TC of up Although graphene/polymer composites did not attract much interests
to 2212 W/m·K, TC of folded graphene sheets with macroscopic thick­ until 2006 [88], Nan et al. developed an analytical model/formula
ness is only 71.4 W/m·K, which is ∼3.3 % of perfect graphene struc­ which can consider the filler shape, orientation and interface thermal
ture. resistance. It is most suitable to estimate the effective TC (kc ) of com­
Other deformations, such as twisting and wrinkles, will undoubtedly posites with graphene as filler [89]. That is,
reduce substantially the TC of graphene [81–87]. Using the non-equi­
3 + f [2 11 (1 L11) + 33 (1 L33 )]
librium Green’s function, which is another atomic simulation method k c = km
for phonon transport, it is shown that the thermal conductance of GNRs 3 f (2 11 L11 + 33 L33 ) (1)
can be, controllably and reversibly, modulated over more than 55 % at with
room temperature by adjusting the twisted angle. The tuning range of
c c
the twisted angle depends on the width and length of the graphene k11 km k33 km
11 = , 33 = ,
ribbon under study. The reduction of the thermal conductance mainly
c
km + L11 (k11 km ) c
km + L 33 (k33 km ) (2)
comes from the phonon scattering due to the inhomogeneous force
constant in the twisted nanoribbons [83]. NEMD simulations on gra­ c
kf c
kf
k11 = , k33 =
phene with wrinkles and pristine graphene show lower TC for the 1 + L11 kf / km 1 + L33 kf / km (3)
former owing to the increased phonon scattering and decreased phonon
lifetime. Also, the TC of graphene wrinkles is insensitive to temperature 2
1
L33 = 2 L11 = 1 2 cosh
in the range of 200−600 K caused by the dominance of low frequency 2( 2 1) 2( 2 1)3/2 (4)
phonon modes in the vertical direction along the wrinkle textures,
which is distinctly different to pristine graphene [86]. where kc is isotropic TC due to random dispersion of graphene in a
polymer matrix, km and kf are thermal conductivities of matrix and

Fig. 11. (A) Simulation structures of multiple folded GNRs. The structures in (a), (c), (e) and (g) are GNRs before relaxation and have 0.74 nm in inter-lamellar space.
(a) GNR with 3 folds; (b) relaxed GNR with 3 folds and 0.24 nm inter-lamellar space; (c) GNR with 4 folds; (d) relaxed GNR with 4 folds and 0.37 nm inter-lamellar
space; (e) GNR with 5 folds; (f) relaxed GNR with 5 folds and 0.34 nm inter-lamellar space; (g) GNR with 6 folds; (h) relaxed GNR with 6 folds and 0.27 nm inter-
lamellar space. The heat bath with high temperature and low temperature are shown in red and blue, respectively. There are two substrates for each GNR. One is on
the top layer and the other is below the bottom layer. (B) Relative TC modulation by compressing the inter-lamellar space with different folds in GNRs. The value 1.0
of relative TC corresponds to 111.5 W/m·K which is TC of the flat zigzag GNR. The sizes of all GNRs are the same, which are 0.71 nm in width and 21.3 nm in length
(Reproduced with permission from Ref. [81]). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this
article.)

8
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

graphene, respectively, and kiic = (i = 1, 3) is equivalent TC of the thermal conductance becomes significant, and the initial morphology of
graphene composite along the x- and z- directions. f is graphene volume the grafting chains gives a large difference on enhancement of inter­
fraction and ρ (=a3/ a1) is aspect ratio, defined as the ratio of thickness facial thermal conductance. In addition, grafting of aligned long PE
to diameter, of graphene. = (1 + 2 ) , where = ak /a3 and chains onto graphene gives a much higher improvement of the inter­
ak = km/Gk ; Gk is interfacial thermal conductance of graphene/polymer facial thermal conductance compared to the grafting of random long PE
matrix. The thermal conductivities of graphene/polymer composites chains on graphene. As revealed by Shen et al. [106], these results are
can be predicted using Eqs. (1)–(4) with kf , f and Gk known. due to the aligned PE chains having more segments which interact
Although Nan et al.’s model can estimate TC of graphene compo­ strongly with the polymer matrix and such aligned PE chains also
sites, it needs the intrinsic TC value of graphene as an input. As de­ possess intrinsically higher TC than the random PE chains.
scribed in Section 2, unlike bulk materials, the external factors induced MD simulations show that the grafting of polymer chains onto the
during composites processing can affect the intrinsic properties of graphene surface can significantly enhance the interfacial thermal
graphene. As such, in theory, an atomic scale approach has to be used to conductance across the graphene/polymer interface. However, it is
model the thermal transport in graphene/polymer composites. Hence, difficult in practice to graft polymer chains onto the basal plane of
MD simulations have been extensively used to evaluate TC of graphene/ graphene because all the bonds of graphene are saturated on this plane.
polymer composites [67,79,85,90–103]. In the following subsections, Moreover, MD simulations demonstrate that the covalent grafting of
the main factors which determine TC of graphene/polymer composites polymer chains on 5 % carbon atoms of the basal plane of graphene
by MD simulations will be discussed. gives ∼60 % decrease of TC of graphene due to the destruction of the
perfect 2D crystallinity of graphene [92].
3.1. Interfacial thermal transport and interfacial thermal conductance An alternative way to solve this problem is, instead of grafting
polymer chains onto the basal plane of graphene, through covalent
It is well established that the filler/matrix interfacial thermal con­ functionalization at the edges of graphene, which will not compromise
ductance is critical to determine the TC of a composite. Poor thermal the intrinsic high TC of graphene but increase the thermal conductance
coupling between filler and matrix usually results in limited thermal across the graphene/polymer interface. Konatham and Striolo ex­
conductance enhancement of the composite due to the serious mis­ amined the effect of grafting branched alkanes on the interfacial
match of the phonon spectra of the interface components [90]. The thermal conductance across a graphene/octane interface [91]. MD si­
degree of mismatch of the phonon vibration at the interface can be mulations show that the graphene/ octane interfacial thermal con­
evaluated by the calculated vibration power spectrum (VPS) which ductance increases by ∼3 times, benefiting from the improved coupling
reflects the atomic vibration energy (that is, the essential part of of the vibrational mode between functionalized graphene and octane.
thermal energy). At the interface between two dissimilar materials, the As displayed in Fig. 14, the VPS traces of graphene and octane show
similarity of VPS usually determines the interfacial thermal con­ that the spectra of pristine graphene and octane do not couple. But the
ductance. Larger overlap between VPS gives larger interfacial thermal VPS plots of functionalized graphene by short branched alkanes and
conductance (or smaller resistance). As an example, Fig. 12 shows the octane overlap at several frequencies.
VPS of a composite of graphene and paraffin matrix (C30H60). The As mentioned above, covalent functionalization techniques can ef­
coupling between the through-plane of graphene and polymer in the fectively reduce the interfacial thermal resistance (that is, the inverse of
low frequency range (2−16 THz) contributes to the main interfacial conductance) between graphene and polymer matrix in the composite.
thermal transport; however, there is minimal VPS spectra overlap, in­ Simultaneously, the diverse covalent functional groups give different
dicating weak vibrational match or high interfacial thermal resistance
in the graphene/paraffin composite.
Interfacial interaction enhancement effectively increases the inter­
facial thermal conductance across the graphene/polymer interface. The
functionalization of graphene with organic molecules facilitates strong
interaction with the polymer matrix, which, in turn, greatly restrains
the vibrational mismatch between graphene and polymer matrix. In
general, there are two approaches to functionalize graphene: covalent
functionalization and non-covalent functionalization [104].

3.1.1. Effect of covalent functionalization on interfacial thermal


conductance
Wang et al. investigated the effect of grafting of polyethylene (CnH2n
+1, PE) chains on the interfacial thermal conductance of a graphene/PE
interface [105]. It is concluded from Fig. 13 that grafting of PE chains
can enhance the interfacial thermal conductance across the graphene/
polymer interface. For example, at a grafting density of ∼0.0144/Å2,
the interfacial thermal conductance increases to 144 ± 15 MW/m2K
from 56 ± 4 MW/m2K of the pristine graphene/polymer interface. In
addition, the increase of grafting density improves the enhancement of
the interfacial thermal conductance. It is also concluded that the length
and initial morphology (Fig. 13c,d) of grafting chains show important
effects on the interfacial thermal conductance. When short PE chains
(i.e., n = 8, 12) are grafted onto graphene, the increase of interfacial
thermal conductance is not significant; and the initial morphology of
the grafting chains shows a marginal difference in the improvement of
the interfacial thermal conductance. This is because the grafted PE
short chains have weak interaction with the polymer matrix, resulting Fig. 12. VPS spectra of carbon atoms of paraffin and graphene. The VPS
in poor thermal coupling between the graphene and polymer. With spectrum of graphene is decomposed into in-plane and through-plane direc­
longer grafted PE chains (i.e., n = 16), the increase of interfacial tions. (Reproduced with permission from Ref. [90]).

9
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

Fig. 13. Scheme illustrating the initial mor­


phology of PE chain grafted onto the surface of
a graphene sheet (a) aligned and (b) random.
Effects of grafting density on interfacial
thermal conductance across the graphene/
polymer interface with different initial PE
chain morphologies: (c) aligned and (d)
random. (Reproduced with permission from
Ref. [105]).

amount of reduction on interfacial thermal resistance in the graphene/ graphene. For example, the TC of a pristine graphene sheet embedded
polymer composites. Wang et al. studied the effect of different covalent in epoxy resin can be reduced by ∼50 % compared to the intrinsic TC
functional groups [e.g., butyl (−C4H9), carboxyl (−COOH) and hy­ of free-standing graphene owing to the large phonon energy reduction
droxyl (−OH)] on the reduction of interfacial thermal resistance of weakened by the matrix. TC reduction of functionalized graphene was
graphene/epoxy composites [107]. These functional groups were con­ relatively small (e.g., ∼10 %) when embedded; but the functionaliza­
sidered randomly dispersed on both sides of graphene with coverage of tion of graphene may deteriorate its TC by over one order of magnitude,
carbon atoms varying from 1.19 % to 9.52 %. Defining the interfacial whether the graphene is free-standing or embedded in the matrix, due
thermal resistance across the pristine graphene and epoxy as RK0, the to the increased phonon scattering from the functional groups. Shen
simulation results showed that the interfacial thermal resistance ratio et al. discovered [109] that graphene functionalization is effective in
(RK/RK0) between functionalized graphene and epoxy was less than 1, enhancing TC of composites with small lateral sizes of graphene, while
indicating reduction of interfacial thermal resistance between functio­ TCE is ineffective when graphene has a large size. Therefore, a critical
nalized graphene and epoxy resin. The reduction increased with in­ size theory was proposed based on the effective medium theory, which
creasing coverage of covalent functionalization and the butyl group could identify the predominant factors controlling the TC of graphene/
gave the largest reduction among the three functional groups. In an­ polymer composites. In this theory, it is necessary to functionalize
other study, Zabihi and Araghi compared the effects of three functional graphene to improve the interface conductance and TC of the composite
groups (phenyl, methyl and hydrogen) on the interfacial thermal re­ when the size of graphene is smaller than the critical value (e.g., a few
sistance of graphene/paraffin composites [108]. It found that each μm). However, the intrinsic high TC of pristine graphene is more im­
covalent functionalization group reduced substantially the interfacial portant when the size of graphene is larger than the critical value, and,
thermal resistance, and the phenyl group was most effective to enhance in this case, functionalization is unnecessary. Shen et al. compared the
thermal transport across the graphene/ paraffin interface [108]. effect of four types (oxygenated group [−OH], fluorine [-F], amine
Although functionalization of graphene can effectively decrease the [-NH2] and triethylenetetramine [TETA]) of graphene functionalization
interfacial thermal resistance, it may not enhance TC of the composites on the TCE of epoxy resin [109]. It was found that functionalization by
because the functionalization can also deteriorate the intrinsic TC of amino groups, particularly TETA, was most effective among the four

Fig. 14. VPS spectra of carbon atoms in octane, (a) pristine graphene (GS 216, which means the graphene sheet has 216 carbon atoms) and (b) short branched
alkanes functionalized graphene. (Reproduced with permission from Ref. [91]).

10
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

functional groups. This can be understood since covalent bonding im­ in reduced graphene/epoxy interfacial thermal resistance, and the re­
proves the interfacial thermal conductance and the long TETA chains duction increased markedly with increasing coverage of functionaliza­
enhance the vibrational coupling of phonons because of their penetra­ tion. However, the three types of polycyclic aromatic hydrocarbons did
tion into the epoxy. The enhanced interfacial thermal conductance in not show different effect on the reduction of interfacial thermal re­
functionalized graphene compensated its low TC, resulting in higher TC sistance. The vibrational density of states (VDOS) of graphene was re­
of the composite when the lateral size of graphene was smaller than the distributed after non-covalent functionalization and became better
critical value [109]. coupled with the VDOS of epoxy, which was the main mechanism for
the reduced graphene/epoxy interfacial thermal resistance [110].
3.1.2. Effect of non-covalent functionalization on interfacial thermal Van der Waals interaction between graphene and polymer matrix
conductance also affects the contact strength of graphene and polymer, giving dif­
Compared to covalent functionalization, non-covalent functionali­ ferent interfacial thermal conductance values across the graphene/
zation can be realized much easier. Also, no defects or adatoms are polymer interface. Luo and Lloyd examined the effect of van der Walls
introduced to the basal plane of graphene during the functionalization, interaction between graphene and polymer on their interfacial thermal
which is conducive to preserving the intrinsic high TC of graphene. conductance [90], which displayed significant enhancement as the van
These features make non-covalent functionalized graphene sheets pro­ der Walls interaction became stronger (Fig. 15a). This result indicates
mising to prepare polymer composites with enhanced TC. Typical non- that high TC graphene composites can be fabricated by selecting a
covalent functionalization of graphene is through the absorption of suitable polymer matrix that has an intense interaction with graphene
molecules containing aromatic structures on the basal surface of gra­ through van der Walls forces.
phene via π-π interaction. For example, pyrene derivatives have been Increasing the polymer density provides stronger van der Waals
widely used to non-covalently functionalize graphene. Lin and Buehler forces at the graphene/polymer interface and consequently stronger
studied the effect of non-covalent functionalization of graphene by alk- thermal coupling at the interface is expected. Luo and Lloyd also stu­
pyrene molecules on the interfacial thermal conductance of graphene/ died the role of polymer density on interfacial thermal conductance
polymer interface [95]. Three alk-pyrene derivatives were used as non- [90]. Fig. 15b shows that there is a fairly rapid increase of thermal
covalent linkers between graphene and model matrix (i.e., octane). conductance across the graphene/polymer interface with increasing
Simulation results showed that only C8-pyrene could increase the in­ polymer density.
terfacial thermal conductance across the graphene/octane interface (22
% increment). This result was caused by the different orientation of the
alkyl chains of pyrene derivatives. The simulation results showed that 3.1.3. Effect of defect/doping on interfacial thermal conductance
C8-pyrene had the highest degree of parallel alignment along the basal Defects of graphene affect the graphene/polymer interfacial thermal
surface of graphene, hence maximizing the through-plane phonon conductance. Li et al. studied the effect of four different types of defects
transport via carbon-hydrogen bond vibrations. In another MD simu­ on graphene/epoxy interfacial thermal conductance [101]. Fig. 16
lation investigation, functional pyrenes, such as 1-pyrenebutylamine, 1- shows that SW and MV defects can markedly enhance the interfacial
pyrenebutyric acid and 1-pyrenebutyl were found to have a similar thermal conductance. When the SW defect concentration is increased
effect on reducing the interfacial thermal resistance between graphene from no defects to 13 %, the interfacial thermal conductance is in­
and paraffin [110]. In addition, it was found that the reduction of creased from 135.54 ± 5.82 to 162.64 ± 4.84 MW/m2K; and as the
thermal resistance increases as the coverage of functional molecules removed atoms of MV defects is increased from no defects to 56, the
becomes higher. At 26−28 wt.% of functional pyrene molecules, the interfacial thermal conductance is increased from 135.54 ± 5.82 to
interfacial thermal resistance between graphene and paraffin can be 163.78 ± 7.10 MW/m2K [101]. These results can be understood be­
reduced by 16–17 %. cause the introduced defects in graphene facilitated the VDOS curves to
The effects of non-covalent functionalization of graphene by three overlap. The impact of carbon isotope doping on graphene/polymer
polycyclic aromatic hydrocarbons on the graphene/epoxy interfacial interfacial thermal resistance was also studied using MD simulation. It
thermal resistance were studied [107]. In this MD simulation study, found that the graphene/epoxy interfacial thermal resistance was in­
functional molecules were attached to both sides of graphene and the sensitive to 13C isotope doping in graphene [17,111] since the VDOS for
13
functionalization coverage was defined as the ratio of total number of C-modified graphene exhibited only small differences when compared
functional molecules to the number of carbon atoms in the graphene. It to graphene based on 12C [107].
found that the presence of non-covalent functional molecules resulted

Fig. 15. Interfacial thermal conductance (Gκ) plotted against (a) interfacial van der Waals interaction, and (b) across graphene/polymer interface versus polymer
density. (Reproduced with permission from Ref. [90]).

11
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

Fig. 16. Interfacial thermal conductance be­


tween graphene and epoxy resin with different
defects. The dashed blue horizontal lines in­
dicate interfacial thermal conductance of the
composites with defect-free graphene.
(Reproduced with permission from Ref. [101]).
(For interpretation of the references to colour
in this figure legend, the reader is referred to
the web version of this article.)

3.2. Effect of interfacial thermal conductance on TC of composites are in the majority; second, the graphene platelets have many defects,
resulting in lower TC than expected; and third, the graphene platelets
Larger interfacial thermal conductance indicates less thermal energy prepared in this study have higher interfacial thermal resistance with
loss at the graphene/polymer interface, and thus higher TC of compo­ the polymer matrix than that used in the calculations.
sites can be obtained. Wang et al. studied the effect of interfacial Although many studies have revealed that an increased number of
thermal conductance on TC of graphene composites [105]. As shown in graphene layers would result in a reduction of TC of suspended or free-
Fig. 17, the TC of graphene/polymer composites clearly increases with standing graphene [19,77], this is quite different with graphene/
increasing interfacial thermal conductance Gκ. Further, TC due to the polymer composites in which MLG sheets possess large lateral sizes but
increase of interfacial thermal conductance is stronger at higher gra­ similar aspect ratios as SLG sheets. MD simulation results (Fig. 19a)
phene volume fraction (e.g., f > 4 vol.%). have shown that the thermal conductivity of graphene/epoxy compo­
sites are enhanced significantly with increasing number of graphene
3.3. Effects of graphene size, dispersion and volume fraction on TC of layer that is consistent with experimental results [114]. Specifically, as
shown in Fig. 19b, the TCE factor (i.e., c k m ) of epoxy composites
k k
composites m
fabricated with 20-layer GNPs, at similar aspect ratio (∽200), is higher
Because of the usually low intrinsic phonon scattering, the TC of than those of composites containing 1-layer and 5-layer GNPs. It is
graphene generally increases with sample size. This has also been noted that compared to TC values of MLG and SLG epoxy composites
confirmed by NEMD simulation, which shows that the TC of graphene given in Fig. 18, the simulated TC values of these composites obtained
increases almost linearly with size below a critical value [93]. This in Fig. 19 show an opposite trend. This is because the TC of embedded
critical value depends on the sample quality. Thus, TC of graphene/ SLG and MLG are assumed to be the same in the effective medium
polymer composites increases rapidly with the size of graphene sheets model in Ref. [93], which we believe is not accurate.
when the size is smaller than this critical value. Hu et al. investigated
the graphene size and volume fraction dependence of the composite TC 3.4. Graphene-polymer interactions induced reduction of TC of graphene
and evaluated the interfacial thermal conductance [93]. As displayed in
Fig. 18, the TC of graphene/polymer composites increases rapidly with Apart from the aforementioned factors influencing the TC of gra­
graphene size and then reaches an upper limit, which is consistent with phene/polymer composites, a factor which can be inadvertently ig­
the size dependence of TC of pristine graphene sheets [75]. However, in nored is the variation of TC of graphene itself after being introduced
graphene framework/polymer composites, higher TC can be achieved into a polymer matrix. This should be considered in the MD
by size-dependent modulation of graphene sheets at a relative lower
graphene loading. Hou et al. reported that a quasi-isotropic graphene
framework fabricated by vacuum filtration could impart high in-plane
and through-plane thermal conductivities to epoxy composites. More­
over, the quasi-isotropic arrangement can be achieved by modulating
smaller lateral size graphene sheets [112].
Graphene/polymer interfacial area increases with graphene volume
fraction. However, the increase of thermal energy loss in graphene/
polymer interface is usually much lower than the heat transport en­
hancement, finally resulting in the increase of TC in the composites
[93,113]. As shown in Fig. 18b, at a given graphene volume fraction,
the MLG sheets (5-layer) can be regarded as the aggregation of SLG (1-
layer), leading to reduced thermal conductive paths in the composite.
Also, the TC of graphene decreases with increasing number of graphene
layers. Together, these factors explain why higher TC of composites can
be achieved with SLG fillers instead of MLG. In Fig. 18a, it is noted that
the upper limit of TC is achieved at a smaller size of SLG than that of
MLG, thus confirming that well-dispersed graphene sheets have higher
TCE efficiency in polymer composites. In the investigation by Yu et al.
[94], MLG/polymer composites displayed TC values much lower than
the expected values calculated using Eqs. (1–4). Three factors cause Fig. 17. TC of graphene composites versus graphene/polymer interfacial
these large discrepancies. First, graphene platelets have a wide layer thermal conductance (Gκ). Here L is the length of graphene and f is the graphene
number distribution and the platelets with layer number larger than 5 volume fraction. (Reproduced with permission from Ref. [105]).

12
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

Fig. 18. (a) TC of polymer composites with 1.0 vol.% graphene versus graphene sheet size and (b) TC of graphene composites versus graphene volume loading. TC
values of graphene and polymer matrix were set as 1000 W/m·K and 0.298 W/m·K, respectively. The interfacial thermal conductance was set at 30 MW/m2K, which is
calculated from MD simulation. (Reproduced with permission from [93]). The experimental data are for MLG/polymer composites reproduced from Ref. [94].

calculations. The simulation results obtained by Mortazavi et al. show


that, compared to free SLG, the TC of SLG in epoxy can be decreased by
30 % due to the interfacial interaction between graphene and epoxy
[92]. The main mechanism underlying this phenomenon is the overlap
of the phonon spectra of carbon atoms for graphene and epoxy, which
causes phonon scattering of graphene in a wide range of frequencies
and hence reduction of its intrinsic thermal conductivity.

4. TC of graphene/polymer composites: Influence of graphene


types

4.1. TC of GO/polymer composites

Apart from their high TC, graphene sheets also have high electrical
conductivity so that graphene/polymer composites usually exhibit sig­
nificantly enhanced electrical conductivity even at very low graphene
loadings [118,119]. This restrictively limits their applications to parts
and devices requiring a high level of electrical insulation [120,121].
As a derivative of graphene, GO is an electrically insulating material
possessing much higher TC in comparison with polymers. Hence, it can
be used to enhance TC of polymer composites yet retaining the elec­
trically insulating properties of the composites [122–124]. Kim et al.
studied the thermal and electrical conductivities of GO (prepared by the
Hummers method and the C:O ratio is 1.8) filled epoxy composites
[125]. As shown in Fig. 20, the GO fillers enhance the TC of epoxy
matrix and retain the electrical insulating properties. At 3 wt.% GO, the
TCE factor is ∼0.9 and the electrical conductivity is of the same order
of magnitude as neat epoxy. Wang et al. found that incorporation of 5 Fig. 20. TC, TCE factor and electrical conductivity of GO/epoxy composites.
Results shown are taken from test data provided in Refs. [125,127].
wt.% GO into epoxy increased the composite TC to ∼1.0 W/m·K, giving

Fig. 19. (a) Simulation results showing the ef­


fects of layer number (n) and lateral size of
graphene on TC of epoxy composites with 1.0
vol.% graphene. (b) Experimental data (unfilled
symbols: blue star from [114], red circle from
[94], blue diamond from [115], red triangle
from [116], and black right-pointing triangle
from [117]) and MD simulation results (solid
curves) showing TCE (in %) of epoxy composites
with graphene aspect ratio and the effect of
layer number (n = 1, n = 5 and n = 20) but
similar lateral size. (Reproduced with permis­
sion from Ref. [114]). (For interpretation of the
references to colour in this figure legend, the
reader is referred to the web version of this ar­
ticle.)

13
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

a TCE factor of ∼4.0 [126]. However, detailed structural analyses of resistivity in comparison with neat epoxy [122].
the GO fillers and electrical measurements of the composites were not
given so that improved understanding of the mechanisms involved 4.2. TC of RGO/polymer composites
could not be obtained. For example, the GO fillers might be partially
reduced during the fabrication process, leading to further increase of 4.2.1. Effect of RGO loading
the TCE of composites. GO is highly oxidized graphene, where oxygen atoms are bound to
It should be noted that even at moderate temperatures, the reduc­ carbon in the basal plane and at the edges of graphene, resulting in
tion of GO can happen within the polymer matrix [128–130]. This issue large amounts of hydroxyl, epoxy and carboxyl groups. The oxygen-
should be considered because the reduction of GO not only enhances containing functional groups form numerous tetrahedrally bonded sp3
the TC more effectively but also gives a significant variation of elec­ carbon clusters, yielding strong phonon scattering and hence GO
trical properties of the composites (e.g., electrical conductivity, di­ usually exhibits low TC. Reduction of GO produces a high fraction of
electric constant, etc.), particularly those highly loaded with GO. An large-size sp2 carbon clusters, in which the phonons travel with slight
investigation on TC of GrO-NH2/epoxy composites revealed that the scattering and therefore RGO exhibits higher TC compared to GO. Thus,
oxidation degree of GrO-NH2 played a critical role [131]. GrO-NH2 the TC of RGO composite is usually determined by several factors:
refers to (3-aminopropyl) triethoxysilane modified graphite oxide, loading level, surface functionality and structural property [133].
where -NH2 groups were introduced to establish covalent interactions Using thermally reduced GO, Yavari et al. prepared 1-octadecanol-
between GrO and epoxy resin. In detail, Fig. 21 indicates that the mildly based phase change composites (PCCs) [134]. At 4 wt.% fraction of
oxidized GrO-NH2 exhibits edge functionalization and a strong TCE RGO, TCE factor of PCCs is ∼1.4. Xue et al. revealed that TC of RGO/
[131]. This is caused by the edge-functionalized GrO sheets which porous poly(vinyl alcohol) (PVA) composites shows an almost linear
provide a preferential thermal conduction pathway for high efficiency increase from 0.31 W/m·K for neat PVA to 0.54 W/m·K at 2 wt.% RGO.
heat transfer in epoxy. Over-oxidation of GrO results in weak TCE in It is also shown that RGO disrupts the chain configuration of PVA
GrO/epoxy composites, while TCE can become stronger when the over- molecules, resulting in free volume reduction that may be responsible
oxidated GrO is reduced. for the increase in TC [135]. In addition, RGO has been used to improve
GO cannot be dissolved in many organic solvents including N, N- the TC of silicone grease and a TCE factor of 1.0 is obtained at 1.0 wt.%
dimethylformamide (DMF), tetra-hydrofuran (THF) and N-methyl-2- RGO [136]. The reduction of GO and hence the TCE of RGO composites
pyrrolidone (NMP), which makes the preparation of their polymer depend strongly on the reduction method used and the reaction con­
composites difficult [132]. To disperse GO in organic solvents, func­ ditions of temperature and chemical agent employed. For example, with
tionalization is required. This will also contribute to the TC value of 3.0 wt.% RGO, reduced by the microwave treatment of GO (C:O =
their composites [122]. Tseng et al. studied glycidyl methacrylate- 3.2:1), in polycarbonate (PC), only a small TCE factor of 0.20 is ob­
functionalized GO on the TC of polyimide (PI) composites [130]. As tained [137]. Further, marginal increase of TC is found in poly(vinyli­
shown in Fig. 22a, composites with functionalized GO (f-GO) exhibit dene fluoride) (PVDF) filled with hydrazine hydrate reduced GO (< 7.0
significantly enhanced TC in comparison to those with as-prepared GO. wt.%) [138]. The low TCE may be attributed to the weak reduction
Although the electrical resistance of the composites with high f-GO capability of the method/reagent used, resulting in low C:O ratio and
loading decreases more when compared with the corresponding as- high defect density [137–139]. In fact, annealed composites at high
prepared GO composites, both types of composites still have high temperatures show high TC since RGO is further reduced within the
electrical resistance (see Fig. 22b). polymer matrix [137]. Lee et al. revealed that the thermal conductance
Aminopropylisobutyl polyhedral oligomeric silsesquioxane (Ap- of RGO/silicone composites not only is affected by thermal reduction
POSS) has been used to modify GO for improved solubility in organic induced morphological and structural changes of RGO but is also sig­
solvents and enhanced dispersion in epoxy matrices. Ap-POSS-GO nificantly affected by the phenolic group density on the basal plane of
shows great influence on TCE but only marginal impact on electrical RGO. The phenolic groups can decrease the interfacial thermal re­
insulation [122]. For instance, adding 0.5 wt.% Ap-POSS-GO improves sistance between RGO and silicone [133].
the TC of epoxy by 57.5 %, but the same content of GO only yields a
16.7 % increase on TC of epoxy. However, the epoxy composite with 4.2.2. Effect of covalent functionalization of RGO
0.5 wt.% Ap-POSS-GO displays a slight decrease of the electrical RGO has the tendency to agglomerate and is generally incompatible

Fig. 21. TC of GrO-NH2/epoxy composites with GrO fabricated with different ratio of KMnO4 to graphite. Here, GrO refers to graphite oxide. The dash line
corresponds to TC (0.157 W/m·K) of neat epoxy resin. (Reproduced with permission from Ref. [131]).

14
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

Fig. 22. (a) TC and (b) surface electrical resistance of GO/PI and f-GO/PI composites based on data provided in Ref. [130].

with most polymers, exhibiting poor dispersion and high interfacial density due to the competing effects between the improvement of in­
thermal resistance. To overcome these disadvantages, covalent func­ terfacial thermal conductance and the reduction of in-plane TC of
tionalization of RGO is usually used to enhance its dispersion and re­ graphene.
duce the interfacial phonon scattering [140]. To achieve strong RGO/ Compared to grafting of organic polymers, the functionalization of
polymer interaction, RGO functionalized by polymer chains have often graphene by small molecules such as coupling agents are more con­
been adopted. Both the edges and basal plane of RGO are functionalized venient. Thus far, silane coupling agents (SCA), titanate coupling agents
to enhance the dispersion and interfacial interaction with the polymer (TCA) and antioxidants have been used to functionalize RGO in order to
matrix. Polyester functionalized RGO (f-RGO) is prepared by grafting improve the thermal conduction of RGO/polymer composites
polyester chains onto the edges of GO via esterification between hy­ [145,146]. Fig. 23 shows the effect of TCA functionalized RGO on TCE
droxyl terminations of polyester and carboxyl groups of GO edges [141] of RGO/PA composites. Clearly, at 5 wt.% RGO and over, TC of f-RGO/
and then reduced. By this method, the sp2 conjugated structure is lar­ PA composites reaches 5.10 W/m·K from 3.34 W/m·K of RGO/PA
gely preserved. The TCE factor is 1.90 at 1.5 vol.% RGO loading, re­ composites, yielding in a 53 % increase.
vealing a great enhancement of polyester grafted RGO (C:O = 5.5:1).
The influence of polymer chain length grafted on the RGO edges on TCE
is also studied [141]. Amino-terminated polyethylene-glycol (PEG) of 4.2.3. Effect of non-covalent functionalization of RGO
different chain length is used to functionalize RGO to enhance the TC of Although covalent functionalization of RGO can improve greatly
polyamide 6 (PA6) composites. The through-plane TC increases with RGO/polymer thermal interfacial conduction, it may also result in
increasing PEG chain length. significant reduction of the TC of the f-RGO due to the introduction of
Bis(3-aminopropyl)-terminated poly(ethylene glycol) (NH2-PEG- defect-sites with the conjugated basal plane of RGO. By contrast, non-
NH2) has been used to disperse RGO in cellulose to prepare composites covalent functionalization of RGO does not generate defects within the
with significantly improved in-plane TC which increases to 9.0 W/m·K graphene basal plane structure. As a result, the intrinsic high TC of
at 6.0 wt.% RGO from 2.47 W/m·K of neat cellulose, yielding a TCE graphene can be well retained.
factor of ∼2.60. The high TCE is attributed to the hydrogen bonding Aromatic molecules are usually used to non-covalently functionalize
between cellulose and f-RGO, the uniform dispersion of f-RGO, and the RGO via strong π-π or van der Waals interactions with the basal plane of
alignment of RGO along the in-plane direction [142]. It is also reported RGO. Pyrene (Py) derivatives are often used to functionalize RGO
that the PEG chain length has significant impacts on both through-plane through π-π interactions between Py and RGO [147]. For example, Py
and in-plane thermal conductivities of RGO/PA6 nanocomposites terminated poly(glycidyl methacrylate) (Py-PGMA) prepared by ATRP
[143]. The through-plane TC increases with increasing NH2–PEG–NH2 is used to functionalize RGO for enhancing TC of epoxy composites. The
chain length; but the in-plane TC has a maximum value of 9.71 W/m·K pendant oxirane functional groups react with epoxy, thus improving the
at a certain length of the NH2–PEG–NH2 chain [143]. compatibility between RGO and epoxy matrix. Fig. 24 shows epoxy
Functionalization of RGO at the basal plane is also found to be
useful to improve the TCE of polymer composites. Typically, the in­
troduction of polystyrene (PS) chains onto the basal plane of RGO is
conducted via diazonium addition and succeeding atom transfer radical
polymerization (ATRP) [144]. The merit of this method is that the
grafting density and grafted chain length can be well controlled, facil­
itating the investigation of their effects on TC of PS composites. Addi­
tion of PS-f-RGO provides strong TCE which can be 1.60 even at a low f-
RGO loading of 2.0 wt.%. In general, the grafted chains in functiona­
lized graphene can act as thermal conductive pathways in the gra­
phene/polymer interface. However, as mentioned before, functionali­
zation can also deteriorate the intrinsic TC of RGO [109]. Hence, there
may exist an optimum grafting density for the highest TCE in the
composites. It is found in experiments that RGO with a lower PS
grafting density (Mn = 81600, Mn/Mw = 1.56, weight loss at 600 °C =
55.4 %) results in a higher TCE [144]. At a higher grafting density (Mn
= 21300, Mn/Mw = 1.60, weight loss at 600 °C = 49.6 %), TCE due to
the improved graphene/matrix interface may be offset by the reduced
TC of f-RGO [97,144]. Hence, it is expected that the maximum TC of Fig. 23. TC of PA composites filled with rGO or TCA functionalized rGO.
composites with polymer grafted graphene occurs at a medium grafting (Reproduced with permission from [145]).

15
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

composites with Py-PGMA-RGO having much higher TC values com­ GO is removed by centrifugation from the pores. This process results in
pared to the epoxy composites filled with pristine RGO, particularly at coating of the foam skeleton by GO. After chemical reduction of GO, a
relatively high filler loading. At 4 parts filler per hundred parts (phr) of 3D RGO continuous network is obtained based on the skeleton. Note the
epoxy, for example, Py-PGMA-RGO/epoxy composite has a TC of uncoated skeleton surface is very smooth, but GO and RGO coated
1.91 W/m·K, giving a thermal enhancement factor of ∼8.6. Also, this skeleton surfaces become rough and exhibit crinkled texture due to the
TC is ∼20 % higher than that of RGO/epoxy composite with the same flexibility of GO and RGO sheets. Finally, thermally conductive com­
filler content of 4 phr. Another example is RGO-based linear low-den­ posites are prepared by pouring PDMS pre-polymer into the RGO coated
sity polyethylene (LLDPE) nanocomposites, wherein Py-grafted LLDPE foam. The merit of this method is that the orientation and density of
(Py-LLDPE) and maleated LLDPE (MA-LLDPE) were chosen as two RGO can be easily controlled in the composites by controlling the foam
different matrices for comparison. The π-π interaction between Py and dipping time in GO suspension and by pressing the RGO coated foam
RGO yielded much improved dispersion of RGO in Py-LLDPE, and from different directions. As the unidirectional compression ratio of the
hence RGO/Py-LLDPE displayed higher TCE values [148]. RGO coated foam is increased to 95 % from 0 %, the in-plane and
through-plane (or cross-plane in Fig. 26c) TCs increase to 1.68 W/m·K
4.2.4. Effect of composites processing methods and 0.76 W/m·K, respectively, from 0.175 W/m·K at zero compression,
The thermal conductivity of polymer composites depends closely on corresponding to an increase of RGO to 2.6 wt.% from 0.13 wt.%, and
the composite morphology, in particular, the filler dispersion state an increase of orientation to 20° from 1° [155]. In the case of 3D
[1,149]. Thus, the TC of polymer composites can be enhanced by compression and a compression ratio of 70 %, the isotropic TC increases
controlling the filler dispersion. Past investigations on the processing- to a maximum of 2.19 W/m·K, equivalent to a RGO loading of 4.82 wt.
structure-property relationships of RGO/natural rubber (NR) compo­ % [155].
sites showed that the TCE is dramatically affected by the composite Compared to constructing 3D interconnected RGO structures,
microstructure (i.e., the dispersion state of RGO), which is mainly aligning RGO in one direction is more effective to improve the heat
controlled by the processing method [150]. Fig. 25 presents the mi­ transfer in the aligned direction. Zhao et al. reported the preparation of
crostructure and TC of RGO/NR composites prepared by latex mixing, high in-plane thermally conductive poly(p-phenylene benzobisoxazole)
solution curing and mill mixing methods. The composites prepared by (PBO) composite films in which RGO was self-aligned along the in-
latex mixing exhibited a “web-like” dispersion morphology of RGO in plane direction. The RGO/PBO composite films exhibit an in-plane TC
NR matrix (images a, b), and this “web-like” morphology was preserved of 50 W/m·K at 5.0 vol.% RGO [156]. Luo et al. prepared RGO/epoxy
during the curing process in a solution containing peroxide (images c, composite films with high anisotropic TC. In their study, bulk RGO with
d), whereas the RGO platelets were uniformly dispersed in the milled an aligned laminated structure in the in-plane direction was first pre­
composites (images e, f). The network morphology facilitated phonon pared by vacuum-assisted self-assembly of modified RGO in suspension.
transport along the RGO, thus giving higher TC in the composites Then, the composite film was obtained by infiltrating epoxy resin into
prepared by latex mixing and then solution treated. the bulk RGO followed by thermal curing. The composite film exhibits
RGO coated PMMA (RGO@PMMA) balls with diameters of in-plane and through-plane TCs of 1.32 W/m·K and 0.17 W/m·K, re­
200−300 nm were used as core-shell fillers in epoxy matrix. The TC of spectively [157]. Kumar et al. fabricated thermoplastic RGO/PVDF-HFP
epoxy composites with 1.0 wt.% RGO@PMMA balls was increased 7- (poly(vinylidene fluoride-hexafluoropropylene) films by solution
fold compared to neat epoxy. By contrast, incorporation of 1.0 wt.% casting and chemical reduction, which exhibited an in-plane TC of
RGO only increased TC of epoxy composite by ∼ 3-fold [151]. Using ∼19.5 W/m·K at a RGO loading of 27.2 wt.% [158]. Recently, the
the π–π interaction between polydopamine (PDA) coated GO (GO-PDA) electrospinning technique has received considerable attention in pre­
and PS microsphere, Yuan et al. prepared electrically insulating but paring thermally conductive polymer composites since it can align fil­
thermally conductive GO/PS nanocomposites [152]. GO-PDA formed a lers along a certain direction [159]. RGO/PI nanocomposites were
3D interconnected structure in the nanocomposites, which resulted in prepared by electrospinning technique and the TC of PI with 5.0 wt.%
an in-plane TC of 4.13 W/m·K and a through-plane TC of 4.56 W/m·K at modified RGO is 4 times higher than that of pristine PI [160].
a low GO-PDA loading (i.e., 0.96 vol.%). Despite the high in-plane TC achieved, most of the aforementioned
Apart from using latex mixing and core-shell fillers, many other anisotropic composite films with aligned RGO display low through-
methods have also been developed to construct interconnected 3D RGO plane TC. To overcome this difficulty, Lian et al. developed a strategy to
structures in thermally conductive polymer composites [153]. In gen­ prepare composites with high through-plane TC [161]. They con­
eral, these preparation methods can be divided into two types: self- structed vertically aligned and interconnected RGO networks in epoxy
assembly method and template method. The self-assembly of GO starts
from the phase separation to develop an interconnected 3D structure,
and the liquid phase acts as a spacer to avoid parallel overlap of GO
sheets and to help form interlinked pores [153]. As an example, the
preparation of interconnected 3D RGO-based composites can be de­
scribed below: First, an isotropic RGO sponge is prepared by thermal
treatment of GO solution and the reduction of GO may be conducted by
thermal or chemical methods. Second, GO or RGO aerogel is obtained
by freeze drying. Third, thermally conductive composites are prepared
by infiltrating the aerogel with polymers such as epoxy resin and
polydimethyl-siloxane (PDMS) pre-polymer. Using this fabrication
technique, Conrado and Pavese prepared epoxy composites with 1.25
wt.% RGO and obtained an 80 % higher TC compared with neat epoxy
[154]. While the processing method is simple, the spontaneous self-
assembly of GO cannot produce efficient overlap of RGO, coupled with
the low RGO content, the TCE of the composites suffers.
The template strategy utilizes commercially available polymer or
metal foam as skeleton to prepare 3D RGO structure. A typical study
was demonstrated by Qin et al. [155] (see Fig. 26). Here, melamine- Fig. 24. TC of Py-PGMA-RGO/epoxy and RGO/epoxy composites versus RGO
formaldehyde foam is dipped into GO suspension and then the excess content. (Reproduced with permission from [147]).

16
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

Fig. 25. TEM images showing the dispersion of RGO in the RGO/NR nanocomposites: (a, b) uncured RGO/NR nanocomposites prepared by direct latex co-coa­
gulation; (c, d) solution-treated RGO/NR nanocomposites; and (e, f) milled RGO/NR nanocomposites. Here, “solution treatment” refers to a process to cure the rubber
but preserve the morphology obtained by latex co-coagulation. (g) TC of RGO/NR composites prepared by different processing methods. (Reproduced with per­
mission from Ref. [150]).

matrix by four steps: (a) formation of GO liquid crystals; (b) oriented at 1350 cm−1, which represents lattice detects of RGO, decreases with
freeze casting, (c) high-temperature reduction in Ar, and (d) infiltration increasing annealing temperature up to 2000 °C. Starting from 2200 °C,
with epoxy resin. The composites showed a high through-plane TC of the D band disappears and RGO becomes defect-free graphene. As a
2.13 W/m·K at 0.92 vol.% RGO, exhibiting a high TCE factor of ∼12.0 result, TCE in the composites increases with annealing temperature of
compared to neat epoxy [161]. RGO up to 2000 °C and becomes stable when the annealing temperature
is higher than 2200 °C. Compared to RGO (prepared by rapidly heating
4.2.5. Comparative study graphite oxide by Hummers’ method, e.g., > 2000 °C/min), defect-free
To gain deep insights into TCE of GO or RGO composites, Fig. 27 graphene (RGO annealed at > 2200 °C) exhibits much higher TCE in
summarizes reported TCE factors in different GO/RGO composites. It the composites. This study shows the importance of high intrinsic TC of
can be concluded that, apart from the RGO/PC composites [137], the filler to enhance the TC of polymer composites.
generally, TCE shows that f-RGO > RGO > f-GO > GO. In some cases,
RGO even exhibits comparable enhancement with f-RGO. This could be 4.3. TC of FLG/polymer and MLG/polymer composites
caused by the high reduction degree and/or large size of RGO [109].
High temperature annealing was found to be very effective to im­ Defect-free SLG have several merits when used as fillers in matrix
proving the quality of RGO [23,162–164]. Studies on the role of an­ for preparing highly thermally conductive polymer composites: ultra­
nealing temperature on the TC of RGO/1-octadanol composites showed high intrinsic TC, lightweight and high aspect ratio. However, defect-
that, as the annealing temperature was increased, the defects such as free SLG has not been widely used for this purpose owing to a few
oxygen-containing groups could be removed and the crystallinity of persistent difficulties. For example, (a) large-scale production of defect-
RGO fully recovered [165]. As shown in Fig. 28, the intensity of D band free SLG is too expensive; (b) compared to suspended pristine SLG, the

17
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

Fig. 26. (a) Schematic showing the


preparation process of RGO compo­
sites. SEM images and photos of (b)
melamine–formaldehyde foam [MF],
(c) GO coated foam [GO@MF], and (d)
RGO coated foam [RGO@MF]. (e) TC
(κ) and orientation degree versus com­
pression ratio of 1D compressed com­
posites. (f) TC and RGO content versus
compression ratio of 3D compressed
composite. (g) TC versus compression
ratio of 1D compressed composites
fabricated by different dipping times.
(Reproduced with permission from Ref.
[155]).

sheets are more suitable fillers to prepare highly thermal conductive


polymer composites because of their ease of production, high intrinsic
TC and low densities of wrinkles (i.e., difficult to fold).
The superiority of FLG sheets in enhancing TC of polymer compo­
sites was demonstrated by the experimental work of Shahil and
Balandin in which liquid-phase-exfoliated defect-free graphene/MLG
hybrids were used to enhance the TC of epoxy resin [116]. The absence
of a disorder-induced D band in the Raman spectra (see Fig. 29) has
confirmed a highly ordered structure of liquid phase exfoliated gra­
phene/MLG hybrids. At 10 vol.% graphene/MLG hybrid loading, TC of
epoxy composites reaches 5.1 W/m·K corresponding to a TCE factor of
∼23. By contrast, traditional fillers, such as graphite and carbon black,
can only yield a weak TCE factor of their composites (< 2.5). It is noted
that the graphene hybrids prepared by higher centrifugation rates gave
higher TC values of the composites since higher centrifugation rates
lead to higher fractions of few-layer-graphene in the hybrids, hence
forming more compact graphene networks in the composites.
Graphene produced by liquid phase exfoliation is almost defect free,
but the method is not scalable. To produce high-quality graphene filler
at large scale, a graphite intercalation compound method was devel­
Fig. 27. Reported TCE factors for different GO/RGO composites. h-PS-f-RGO1 oped. GFs with low oxygen content (∼2.1 %) were fabricated using this
and l-PS-f-RGO2 are RGO grafted with high and low density of PS chains and method and Raman spectra (Fig. 30) only displayed a weak D band
RGO was prepared via reduction of GO by hydrazine hydrate (100 °C, 14 h)
peak [117]. Non-covalently functionalized GFs (f-GFs) were used to
[144]. Polyester-f-RGO was polyester grafted chemically reduced GO [141].
enhance the TC of epoxy resin. Fig. 30 shows composites with 10 wt.%
RGO was prepared by rapid thermal extension of graphite oxide [126,134]. GO
was prepared by modified Hummers method [125,130]. Py-PGMA-RGO was
f-GFs have a TC of 1.53 W/m·K, which is much higher than the values of
non-covalent f-RGO by Py-PGMA [147]. PGMA-f-RGO was PGMA grafted RGO composites filled with GO, graphite, CNTs and carbon black. Compar­
[130]. ison of Figs. 29 and 30 shows that, compared to liquid-phase-exfoliated
defect-free graphene/MLG hybrids, the f-GFs produced by the graphite
intercalation compound method yield much lower TCE factors in the
SLG added to a polymer matrix shows significantly reduced TC due to
composites. Although slight oxidation of the f-GFs may be an important
the flexural (through-plane acoustic) ZA (Z: through-plane transverse
reason, the effect of filler size, size distribution and aspect ratio cannot
and A: acoustic) phonon suppression; and (c) high flexibility can give
be excluded.
substantial wrinkling of graphene sheets in polymer matrix, leading to
Another way that can be scaled up easily to fabricate defect-free FLG
an increased number of interfaces across the graphene. By contrast, FLG

18
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

Fig. 28. (a) Raman spectra of RGO and an­


nealed RGO at different temperature. TC and
TCE factors of RGO/1-octadanol composites:
(b) composites with different loading of pris­
tine RGO, (c) composites with 5% RGO an­
nealed at different temperature, and (d) com­
posites with different loading of RGO annealed
at 2200 °C. (Reproduced with permission from
Ref. [165]).

for polymer composites is by ball milling. Using this well-established 23.6. By contrast, composites with the same loading of sonication ex­
industrial grinding technique, FLG dispersion can be quickly obtained foliated expanded graphite has a > 50 % lower TC of 1.84 W/m·K.
from graphite with high yield [166]. Using melamine and Darvan-I A proper comparative study on TCE of different graphene filler is
(sodium polynaphthalene sulphonate) as exfoliating agent and surfac­ highly desirable to understand the factors that determine TC of gra­
tant, George et al. produced defect-free FLGs (2∼5 layers) by planetary phene/polymer composites. However, such a study cannot be done
ball milling of graphite in water and dispersed them as fillers in NR properly with so many factors influencing the final TC of a composite,
latex to fabricate composites with high TC [167]. At 37 ℃, the resulting including graphene size, size distribution, thickness, thickness dis­
composites with < 1.43 wt.% original graphite exhibit TCE factors of tribution, aspect ratio and defects, etc. Even if the same method was
4.86 and 9.83 when 200 g and 500 g weights are applied, respectively used, the geometric properties of graphene might be strongly affected
[167]. by the different preparation conditions, e.g., sonication time, etc.
At a given graphite content, reduction of layer number can facilitate Despite this, many efforts have been made to explore the factors af­
the formation of more compact thermal conductive paths in composites. fecting TC of graphene/polymer composites. Warzoha and Fleischer
Based on this idea, re-exfoliation of re-expanded graphite was used to attempted to reveal the effects of geometric properties of graphene on
produce SLG and FLG. Using graphene produced by this method, a large TC of paraffin composites [169]. As shown in Fig. 32, the TCE of the
TCE was found in epoxy composites [168]. As shown in Fig. 31, while composites increases with increasing graphene thickness and diameter.
graphene prepared by this method gives defects (Raman spectra of It is believed the results can be explained in terms of the graphene
graphene shows a strong D band), the composites show much higher bending stiffness that depends on its thickness. Thinner graphene tends
TCE compared with those filled with sonication exfoliated expanded to form more interfaces within the paraffin due to the ease of folding
graphite and graphite powder. At a graphene loading of ∼10.0 wt.%, (i.e., thin graphene has low bending stiffness). This is a reasonable
TC of epoxy composites reaches 4.0 W/m·K, exhibiting a TCE factor of explanation. Similarly, Kim et al. also revealed that graphene with large

Fig. 29. (a) Raman spectra of graphite, bi-layer graphene (BLG) and MLG. (b) TCE factors of epoxy composites with graphite, carbon black and graphene/MLG
hybrids. In Graphene-MLG-Hybrid Epoxy A and B, the sonication time used for graphene preparation is 12 h and 10 h, respectively, and centrifugation rates for
graphene preparation are 15,000 rpm and 5000 rpm, respectively. (Reproduced with permission from Ref. [116]).

19
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

Fig. 30. (a) Raman spectra of graphite, graphene and f-GFs. (b) TC of epoxy composites with GO, graphite, MWNT (multi-walled CNT), carbon black and f-GFs.
(Reproduced with permission from Ref. [117]).

lateral size and thickness could enhance TC of composites more than solution to form a proper concentration; (c) epoxy adhesion agent was
small-sized graphene because of the reduced filler/matrix interface gently added and the solution stirred at room temperature; and (d) the
[170]. However, Goli et al. reported that TC of graphene/paraffin final solution was drop-cast on a Teflon mould and heated at step-in­
composites with 1.0 wt.% SLG can reach ∼15 W/m·K from 0.25 W/m·K creased temperatures. The resultant graphene/OC nanocomposite films
of pristine paraffin, i.e., a TCE factor of 60 (Fig. 32b) [171]. To date, comprise long-range well-aligned rGO layers, which are separated by
this is one of the reported maximum TCEs in composites with low the OC planar layers. The nanocomposites films exhibit an in-plane TC
graphene loading. Moreover, paraffin composites with 20 wt.% of dif­ of 25.66 W/m·K at 4.1 vol.% rGO loading, resulting in a TCE factor of
ferent graphene thickness exhibit comparable TC values. These results 72.35 [177].
further reflect the complexity of TCE of polymer/graphene composites. Owing to the lack of functional groups on FLGs, their interaction with
That is, the TCE of graphene-based polymer composites is influenced by most polymers is low, which results in poor dispersion and weak heat
multiple factors. transfer across the FLG/polymer interface. To overcome these problems,
As TC of graphene is highly anisotropic and the in-plane TC is much efforts were made to form interconnected 3D FLG structures in the
higher than the through-plane TC, it is expected that, at the same polymer matrices, which usually displayed much higher TCE factors
graphene loading, the formation of aligned graphene can enhance in- compared to randomly dispersed FLG. Li et al. developed a spray drying
plane or through-plane TC of composites more efficiently as confirmed method to prepare FLG/SBR (styrene-butadiene rubber) composites that
by many studies [156,158,172–175]. Thus, Li et al. prepared aligned could not be processed by traditional methods due to the weak FLG/SBR
MLG/epoxy composites through two steps (Fig. 33): (a) An ethanol/ interactions [178]. FLG loading up to 15.0 vol.% was introduced to form
water solution of graphene was vacuum-filtrated, yielding self-assembly an interconnected 3D FLG structure in SBR, yielding not only a high TCE
of aligned graphene; and (b) epoxy resin and curing agent were im­ factor of 13.30 (that is, TC increased from 0.20 W/m·K of SBR to 2.92 W/
pregnated into the aligned graphene and cured [176]. The aligned m·K of FLG/SBR) but also excellent electromagnetic shielding of 45 dB at
graphene/epoxy composites exhibit strong anisotropic thermal con­ 8–12 GHz [178]. Fang et al. developed a method to increase the interac­
ductivities. For example, at 40 °C, the TCs of in-plane and through- tion between FLG foam and PDMS matrix [179]. PDA was first coated onto
plane are, respectively, 16.75 and 5.43 W/m·K. For comparison, com­ the FLG foam skeleton and then amino-silane was introduced by reaction
posites having randomly dispersed graphene have a low TC of 1.84 W/ with PDA. The modified FLG foam was compressed before infiltration with
m·K. Using an evaporation induced self-assembly method, Zeng et al. PDMS and the obtained PDMS composites with 11.62 wt.% FLG exhibited
prepared graphene/oxidized cellulose (OC) nanocomposite films with an in-plane TC of 28.77 W/m·K and a through-plane TC of 1.62 W/m·K
highly oriented structures [177]. The preparation process includes the [179]. Using Ni foam as a template and CVD, Shen et al. [180] prepared
following 4 steps: (a) GO/OC solution was heated at 180 °C for 6 h in a epoxy composites with 3D interconnected FLG through several steps in­
Teflon-lined autoclave; (b) DMF was added into the cooled rGO/OC cluding Ni foam stacking and compression, FLG growth by CVD, Ni

Fig. 31. (a) Raman spectra of graphite and graphene prepared by re-exfoliation of re-expanded graphite. (b) TC of epoxy composites with graphene prepared by re-
exfoliation of re-expanded graphite, sonication exfoliated expanded graphite, and graphite. (Reproduced with permission from Ref. [168]).

20
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

Fig. 32. TC of graphene/paraffin composites. Tables below the figures show the geometric parameters of the corresponding graphene filler. (Reproduced with
permission from Ref. [169,171]).

etching, infiltration by epoxy and then curing. The interconnected FLG 4.4. TC of 3D structured graphene/polymer composites
exhibited certain orientation degree along the in-plane direction de­
pending on the applied compressive stress when they were packed densely When dispersed in a composite, graphene sheets may be pre­
through the plane, yielding high loading (up to 8.3 wt.%) but uniform ferentially aligned in the direction of applied load, usually resulting in
dispersion of FLG within epoxy. With only 2.2 wt.% FLG, the composites low through-plane TC of the composite [181]. To obtain a composite
exhibited a TC higher than 5.0 W/m·K along the in-plane direction, and with high isotropic TC, random dispersion of highly filled graphene is
when FLG was increased to 8.3 wt.% this became 8.8 W/m·K. Such a high preferred. However, the loading of randomly dispersed graphene sheets
TC is ∼43 times higher than that of neat epoxy resin (∼0.2 W/m·K) is usually limited because of its influence on other properties (e.g.,
[180]. mechanical, processing, etc.) of the composite [174]. Large 3D

Fig. 33. (a) Schematic illustrating the preparation process of epoxy composites with aligned graphene (AG) and dispersed graphene (DG). (b) Raman spectrum of the
graphene used and (c) temperature dependence of TC of graphene/epoxy composites. (Reproduced with permission from Ref. [176]).

21
Table 1
Summary of TC and TCE factor of GNP-based polymer composites.
Graphene Raw materials Preparation method Lateral size Thickness Aspect ratio Loading Matrix TC TCE factor Ref.
X. Huang, et al.

filler (W/m K)

GNP NFG, Thermal exfoliation of intercalated NG at 800 °C 0.35 μm 1.7 nm 25 vol.% Epoxy resin 6.44 30.00 [94]
500 μm
GNP NFG, Thermal exfoliation of intercalated NG at 800 °C 0.35 μm 1.7 nm 5.4 vol.% Epoxy resin 1.45 6.21 [94]
500 μm
GNP NFG, Thermal exfoliation of intercalated NG at 800 °C ∼ 1 μm ∼ 25 ± 5 10 wt.% Epoxy resin ∼ 0.42 1.09 [184]
< 43 μm
GNP NFG, Thermal exfoliation of intercalated NG at 800 °C ∼ 1 μm ∼ 35 ± 10 10 wt.% Epoxy resin ∼ 0.88 3.38 [184]
< 78 μm
GNP NFG, Thermal exfoliation of intercalated NG at 800 °C Several to > 15 1−100 nm ∼ 290 ± 10 wt.% Epoxy resin ∼ 2.0 8.95 [184]
180–420 μm μm 50
GNP EG Chemical exfoliation & mechanical exfoliation 25 wt.% Cellulose 30.25 [211]
0.53
GNP NFG, Thermal exfoliation of intercalated NG at 800 °C Several to > 15 1−100 nm 350 ± 100 10 wt.% Epoxy resin ∼ 1.52 6.56 [217]
500–850 μm μm
Oxidized NFG, Thermal exfoliation of intercalated NG at 800 °C plus nitric acid Several to > 15 350 ± 100 10 wt.% Epoxy resin ∼ 2.08 9.03 [217]
GNP 500–850 μm treatment μm
GNP NFG Mechanical cleavage of NG ∼5 μm ∼ 60 nm 10 wt.% PE 5.9 11.00 [199]
GNP NFG Thermal exfoliation of intercalated NG at 600 °C 3.9 μm Several to 100 nm 16 wt.%, Epoxy resin 2.27, 10.35, [115]
20 wt.% 4.27 20.35
GNP NFG Thermal exfoliation of intercalated NG at 600 °C plus silane 3.9 μm Several to 100 nm 16 wt.%, Epoxy resin 3.27, 15.35, [115]
functionalization 20 wt.% 5.86 28.30
GNP EG In situ intercalated EG by caprolactam onium ion 12 wt.% PA 6 2.49 6.78 [205]
GNP NFG, Thermal exfoliation of intercalated NG at 900 °C plus separation 16 wt.%, Epoxy resin ∼ 3.55, 15.21 [218]
500 μm by a 50 mesh sieve 20 wt.% ∼ 4.3 18.63

22
GNP NFG, Thermal exfoliation of intercalated NG at 900 °C plus separation 16 wt.%, Epoxy resin ∼ 2.3, 9.50, [218]
500 μm by a 100 mesh sieve 20 wt.% ∼ 4.3 18.63
GNP NFG, Thermal exfoliation of intercalated NG at 900 °C plus separation 16 wt.%, Epoxy resin ∼ 1.5, 5.84 [218]
500 μm by a 150 mesh sieve 20 wt.% ∼ 4.3 18.63
GNP NFG Thermal exfoliation of intercalated NG at 800 °C 0.5−2 μm 25 nm 4.25 vol.% Silicone grease 1.05 7.07 [136]
GNP GO Thermal exfoliation of graphite oxide at 48 μm ∼3 nm 5.3 wt.% PEG 1.35 4.63 [219]
1050℃ followed by annealing at 2200℃under argon
atmosphere
GNP Commercially available 5 μm 6–8 nm 20 wt.% PCa 1.13 ∼4.90 [220]
GNP NFG Thermal exfoliation of intercalated NG 15 μm 5−10 nm ∼ 1500 20 vol.% HDPE ∼ 1.1 [221]
GNP NFG Thermal exfoliation of intercalated NG 15 μm 5−10 nm ∼ 1500 4.25 vol.% Nylon 66 ∼ 4.1 ∼ 12.66 [221]
GNP NFG Thermal exfoliation of intercalated NG 5 μm ∼ 5 – 10 nm 5 wt.% Epoxy resin ∼ 0.45 1.14 [188]
GNP NFG Thermal exfoliation of intercalated NG < 1 μm ∼ 5−10 nm 5 wt.% Epoxy resin ∼ 0.25 0.19 [188]
GNP Graphite oxide Ball mill of GO for 10 h and heated in a microwave oven at 1000 1−5 nm 1.5 wt.% Silicone 2.7 12.00 [222]
W
GNP NFG, Thermal exfoliation of intercalated NG at 1100 °C 2.7 wt.% Paraffin 2.75 10.00 [223]
∼ 700 μm
GNP NG Microwave thermal exfoliation of intercalated NG 15 μm 10 nm 5 vol.% Paraffin 2.41 8.64 [187]
GNP Graphite oxide Mechanical exfoliation of GO < 10 μm 20 wt.% PVDF 0.562 2.12 [138]
f-GNP GNP Surface hydroxylation of GNP 40 μm 250 30 wt.% Epoxy resin 1.698 8.00 [192]
GNP Commercially available 10 ± 7 μm 5−20 nm 725 ± 125 17 vol.% Epoxy resin ∼ 3.28 17.22 [185]
GNP Commercially available 19 ± 9 μm 9−20 nm 1255 ± 145 17 vol% Epoxy resin 3.86 20.44 [185]
GNP Commercially available 20 ± 7 μm 15−100 nm 568 ± 298 17 vol.% Epoxy resin 4, 21.20 [185]
24 vol.% 12.40 68.00
GNP Commercially available 34 ± 20 μm > 100 nm 340 ± 200 17 vol.% Epoxy resin ∼ 2.72 30.0 [185]
GNP Commercially available 2 ± 1 μm 5−20 nm 175 ± 25 Epoxy resin ∼ 0.65 [223]
GNP Commercially available 15 μm 10 wt% PEIb 3.2 16.00 [209]
GNPb Commercially available 40 μm 250 29.3 vol.% PPSc 4.41 19.00 [210]
GNP Commercially available 3−7 μm 20−60 nm 10 wt.% Epoxy resin 0.668 2.69 [224]
(continued on next page)
Materials Science & Engineering R 142 (2020) 100577
Table 1 (continued)

Graphene Raw materials Preparation method Lateral size Thickness Aspect ratio Loading Matrix TC TCE factor Ref.
filler (W/m K)
X. Huang, et al.

GNP Commercially available 40 μm 250 25 wt.% Epoxy resin 2.67 14.22 [197]
GNP Commercially available 5−10 μm 4−20 μm 10 wt.% Eicosane 2.0 4.00 [225]
GNP Commercially available 5−10 μm 4−20 μm 5 wt.% Paraffin 0.7 1.69 [216]
GNP Commercially available 1.5−10 μm ∼ 12 nm 12 wt.% Epoxy resin 0.73 3.50 [226]
GNP Commercially available ∼25 μm ∼6 nm 0.53 wt.% PA 66 0.53 1.65 [227]
0.65 wt.% 0.65 2.25
0.94 wt.% 0.94 3.70
1.32 wt.% 1.32 5.60
GNP NFG Ultrasonic treatment and homogenized at high pressure by 5 wt.% Silica gel sheets 0.43 1.10 [228]
homogenizer, repeated 20 times.
GNP Commercially available 5−25 μm 6−10 nm 16.7 wt.% PPd 0.8−1.0 4.00 [229]
GNP Commercially available 50 μm 17.6 vol.% HDPE ∼4.13 [204]
GNP NFG, Liquid phase exfoliation of NG 1.5 μm 20 wt.% PTFEe 4.02 13.00 [230]
∼20 μm
Modified GNP Commercially available and pre-treated with a coupling agent 0.17 wt.% Silicone rubber 0.20 [231]
GNP KH-550 0.33 wt.% 0.40
0.67 wt.% 0.50
GNP Commercially available 5−7 μm < 30 layers 1 wt.% PVDF 0.27 0.17 [232]
5 wt.% 0.48 1.09
10 wt.% 0.75 2.26
20 wt.% 1.64 6.13
f
GNP Commercially available 5−10 μm 4−20 nm 1 wt.% Polymer bonded 2.799 5.40 [233]
explosives
GNP Commercially available 40 μm 250 30 wt.% Epoxy resin 1.35 5.70 [192]
GNP Commercially available 1.5−10 μm 0.35−12 nm 55 wt.% Epoxy resin ∼8 35 [234]

23
2−8 μm 0.35−3 nm ∼3.2
GNP Commercially available 3−5 μm < 10 layers 0.43 vol.% Natural rubber 0.79 5.08 [235]
f-GNP Graphite oxide Surface functionalization of GO 3 wt% Epoxy resin 0.67 4.03 [236]
GNP Commercially available 7 μm 7 vol.% Linear PP ∼0.225 [237]
7 μm Branched PP ∼0.24
100 μm Linear PP ∼0.4
100 μm Branched PP ∼0.35
Modified GNP EG Ultrasonic treatment and modification with atmospheric- 0−20 μm < 100 nm 20 wt.% Epoxy resin 1.73 8.11 [238]
pressure plasma
GNP Commercially available 1−5 μm 3 nm 1.51 wt% Polyethylene glycol 1.03 2.32 [239]
GNP Commercially available 8 μm 20 vol.% PBT/PCh 5.82 (in-plane) 24.3 [240]
1.06 (through- 3.61
plane)
Modified GNP GNP GNP modified with polydopamine 10 wt.% Polyvinyl alcohol 13.4 43 [241]
f-GNP GNP Decorated with silica nanoparticles 10−15 μm 1−3 nm 2 wt.% PDMS 0.497 1.55 [242]
GNP NFG Liquid-phase exfoliation ∼400 nm ∼1.6 nm 4 wt.% PDMS 0.93 4.00 [243]
GNP Commercially available 5 μm 1.5 nm 1 wt.% Glass fiber-epoxy resin ∼0.38 [244]

Note: aPolycarbonate; bPolyether imide; cPolyphenylene sulphide; dPolypropylene; ePolytetrafluoroethylene, fComposites consist of a majority of powerful explosive crystals and a small portion of polymer binder; gTCA
functionalized; hpolybutylene terephthalate/polycarbonate blend.
Materials Science & Engineering R 142 (2020) 100577
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

graphene structure, e.g., aerogels, can markedly increase the TC of temperature of intercalated graphite, size of starting NFG, surface
composites at a low graphene loading. However, the size of aerogels is modification and separation of GNPs, etc. could affect very much the TC
often too small for industrial applications and the preparation is usually of polymer composites. Table 1 summarizes reported TC values and
complicated. In addition, the composite shape may be limited by the TCE factors of GNP-based polymer composites. In the following, the
aerogel configuration and the infiltration process of the composite is influence of preparation method, size, aspect ratio, functionalization of
largely restricted by the pore structure of the aerogels and the viscosity GNP and the composite processing methods on TC of GNP composites
of the matrix or their precursors. 3D structured graphene fillers can will be discussed in detail.
overcome these shortcomings of the graphene sheets and large 3D
graphene structure. Using 3D interconnected graphene microsphere as
4.5.1. Effects of preparation method, size and aspect ratio of GNPs
filler, Li et al. prepared epoxy composites with high through-plane TC at
It has been shown from previous research that the preparation
low loading. Owing to the interconnected 3D network of graphene
method, size and aspect ratio of GNPs show significant influence on the
microspheres, the epoxy composite with 1.0 wt.% loading gave the
TCE of polymer composites. The effect of heat treatment temperature of
maximum through-plane TC of 0.96 W/m·K, yielding a TCE factor of
intercalated graphite on the exfoliation of NFG and TC of GNP/epoxy
4.31 [182]. Zhou et al. prepared thermally conductive epoxy compo­
composites was studied by Yu et al. [94]. Three temperatures (200, 400
sites using 3D graphene as filler, which was prepared by fragmenting
and 800 °C) were employed to exfoliate the intercalated graphite. As
the large graphene foam [183]. The TC of 3D graphene/epoxy com­
shown in Fig. 34a, epoxy composites with GNPs exfoliated at higher
posites increased to 0.52 W/m·K at 0.27 wt.% graphene loading, which
temperature yield greater enhancement of TC. This is caused by the
is 2.74 times that of neat epoxy resin. Fang et al. recently reviewed the
increase of heat treatment temperature, giving rise to reduced number
progress on the preparation strategies of 3D structured graphene for
of graphene layers, which, in turn, increases the thermal conductive
TCE of composites [174]. For details, interested readers may refer to
paths in the composites.
this article.
Because GNPs are made from NFGs, the effect of the starting gra­
phite flake size on the final lateral dimension of exfoliated GNPs and TC
4.5. TC of GNP/polymer composites of GNP/epoxy composites should be considered. Sun et al. [184] found
that, under the same exfoliation conditions, NFGs with larger lateral
GNPs (called graphene nanoplatelets or graphite nanoplates), re­ size undergo higher exfoliation degree, giving higher aspect ratios of
ferred to as 2D graphite materials with a thickness less than 100 nm GNPs. Hence, epoxy composites with GNPs prepared from large-sized
[15], have been widely used as fillers to achieve high TC polymer NFGs produce high TC (Fig. 34b,c). However, as noted from Fig. 34c,
composites. The main advantage of utilizing GNPs is their ease of composites with GNPs having ultra-high aspect ratio (GNP-D, 500−850
preparation in comparison to SLG or MLG, particularly at large scale. μm) may exhibit low TC. This is explained by the fact that NFGs with
GNPs prepared by different methods have been used to prepare polymer too large size cannot be homogeneously intercalated, yielding in­
composites with high TC. complete exfoliation. Using commercially available GNPs as fillers,
Thermal exfoliation of intercalated natural graphite is the most Shtein et al. studied the effects of geometry parameters of GNPs on TC
common method for the preparation of GNPs. Here, natural flake gra­ of epoxy composites [185]. It is also found that the aspect ratio of GNPs
phite (NFG) was first intercalated by a mixture of sulphuric and nitric is the most important parameter controlling the composite TC. Higher
acids. Then, the intercalated graphite was subjected to a sudden high aspect ratios are more effective in enhancing TC of polymer composites.
temperature (e.g., several hundred to 900 °C) heat treatment. After To investigate the effect of GNP size on TC of polymer composites,
sonication of the thermal exfoliated graphite in a solvent, GNPs could Debelak and Lafdi prepared GNPs of different sizes by separating the
be successfully obtained. Several factors, e.g., heat treatment exfoliated GNPs by using sieves of 150, 100 and 50 mesh, respectively

Fig. 34. (a) TC at 30 °C for epoxy composites


with 5.4 wt.% GNPs. GMP (average lateral size
≈ 30 μm, average thickness ≈ 10 μm), GNP-
200: GNPs exfoliated at 200 °C (average lateral
size ≈ 1.7 μm, average thickness ≈ 60 nm),
GNP-400: GNPs exfoliated at 400 °C (average
lateral size ≈ 1.1 μm, average thickness ≈ 25
nm), and GNP-800: GNPs exfoliated at 800 °C
(average lateral size ≈ 0.35 μm, average
thickness ≈ 1.7 nm). (Reproduced with per­
mission from Ref. [94]). (b, c) Average-aspect-
ratios of GNPs prepared from different starting
natural graphite and TC of epoxy composites
with GNPs prepared from different starting
NFG. (Reproduced with permission from Ref.
[184]). (d) TC of epoxy composites versus gra­
phite content (i.e., GNPs) having different size.
(Reproduced with permission from Ref. [186]).

24
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

[186], and their products were labelled small, medium and large. dispersed in styrene-butadiene rubber (SBR) solution since the solution
Fig. 34d shows that except at very low and very high loadings, epoxy environment provided more interlayer spacing for elastomer molecules
composites with larger GNPs generally exhibit higher TC. Also, the TC to intercalate and retain the high aspect ratio of GNPs [195]. However,
of composites with large GNPs increases almost linearly with increasing in highly filled composites, melt compounding under high shear forces
GNP loading. These results are caused by large GNPs having high aspect are more effective to reduce the agglomeration of fillers. Hence, com­
ratios, thus enhancing the TC of the composites. Other researchers have bined processing techniques have been used to enhance the TC of
also confirmed that high aspect ratio GNPs are more effective to in­ polymer composites with high graphene loading. Ren et al. prepared
crease the TC of epoxy composites [187,188]. graphene/PA composites by combining the benefits of solution mixing
Using expanded graphite as starting materials, Song et al. [189,190] and melt compounding, which gave a through-plane TC of 3.55 W/m·K
and Veca et al. [191] developed a method to prepare GNPs whereby the in those composites with 20 wt.% graphene loading [196].
expanded graphite was prepared by three steps: (a) sonication treat­ Guo and Chen investigated the effects of processing methods on TC
ment in alcohol/water mixture, (b) intercalation in sulphuric acid/ni­ of commercially available GNP-based composites [197]. Ball milling
tric acid mixture, and (c) long-time sonication treatment. Using these and sonication were used to disperse GNPs in epoxy. As shown in
GNPs they prepared highly filled epoxy [191], ethylene-vinyl acetate Fig. 35, in general, the TC of the composites exhibits strong dependence
copolymer (EVA) [190], PI [190] and PVA [189] composites. Since only on processing time. Ball milling seems much more effective to achieve
thermal diffusivities of these composites were given, complete TC higher TC than sonication methods. The existence of a maximum in the
analysis was impossible to conduct. processing time dependent TC means that both processing methods not
only improve the dispersion of GNPs but also reduce the size of the
GNPs. These results indicate that maintaining a well-controlled balance
4.5.2. Effect of functionalization of GNPs
between the dispersion and large size of GNPs is important to maximize
Filler surface functionalization has been widely used in the pre­
the TC of GNP-based composites. Similar findings have been found in
paration of polymer composites with enhanced TC. But there is less
other studies [198], indicating that ball milling is promising for prac­
attention on the role of GNP functionalization on the TC of composites.
tical production of thermally conductive composites filled with GNPs.
SCAs have been used to functionalize GNPs. In the work of Ganguli
Aligning GNPs along a specific direction is an efficient way to en­
et al., at 8 wt.% GNPs, the thermal conductivities of epoxy composites
hance TC in that direction. Several methods have been developed to
filled with functionalized and non-functionalized GNPs are 1.24 and
prepare polymer composites with aligned GNPs through applications of
0.58 W/m·K, respectively [115]. Gu et al. also reported that at 30 wt.%
electric field, magnetic field, mechanical strain, melt flow and two-roll
GNPs, epoxy-terminated silane functionalized GNP/epoxy composite
milling. Mechanical strain with low drawing rate ∼40 μm/min was
has a TC of 1.70 W/m·K that is higher than 1.35 W/m·K of the pristine
used by Saeidijavash et al. to align GNPs in PE. At a strain of 400 % and
GNP composite [192]. Chen et al. designed a Diels-Alder reaction route
7.0 wt.% GNPs, a TC of 5.10 W/m·K is achieved for aligned GNP/PE
to functionalize pristine graphite by acrylic acid (AA) and they docu­
composites, which is ∼5 times higher than that of strain-free compo­
mented a TC of 21.3 W/m·K in PVA composite with 30 wt. % functio­
sites (1.10 W/m·K) [199]. Fig. 36a shows how two-roll milling is used to
nalized graphite. This is much higher than 14.1 W/m·K of composites
mix polyolefin elastomer (POE, ethylene content 8.5 wt.%) and natural
with the same loading of pristine graphite [193]. These results meant
graphite (NG) and through the strong shear force obtain the orientation
that surface functionalization of GNPs gives stronger interfacial adhe­
of NG flakes along the in-plane direction. After hot-pressing, the lami­
sion between graphite and epoxy and/or reduced acoustic impedance
nated samples (A and B) produce highly anisotropic thermal con­
mismatch at the graphite/epoxy interface.
ductivities, Fig. 36b. The TC along the aligned GNP direction reaches
13.27 W/m·K at 49.30 vol.% NG [200]. An L-shaped tube (like equal
4.5.3. Effect of composites processing methods channel angular extrusion used in metals) is used to prepare PVDF
Achieving excellent dispersion of GNPs in composites is not only composites with oriented GNFs (6−8 nm thickness), Fig. 36c. Sliced
important to enhance the TC but is also vital to obtain other desirable circular disks exhibit a TC along the oriented GNF direction of
attributes (e.g., mechanical properties). However, GNPs tend to ag­ ∼10.0 W/m·K at 25 vol.% as shown in Fig. 36d [201].
gregate during dispersion because of the strong van der Waals interac­ External compressive forces on the composites during curing
tion; and efforts have been made to obtain excellent dispersion of GNPs (thermoset matrices) or consolidation (thermoplastic matrices) can
by developing effective composite processing methods. align the fillers, reduce the inter-filler distance and effect of voids,
In general, melt compounding and solution mixing are two com­ hence the TC can be markedly increased and anisotropic. These results
monly used methods to fabricate graphene/polymer composites. clearly depend on the level of the applied compression. Ohayon-Lavi
Compared to melt compounding, solution mixing is more effective to et al. reported the effect of compression on TC of epoxy composites with
prevent stacking of graphene layers in polymer matrices [194]. Araby hybrid fillers comprising GNPs and graphite flakes [202]. A high TC of
et al. found that hydrophobic GNPs (∼3 nm thick) could be uniformly

Fig. 35. Effects of GNP loading and processing time on TC and TCE factors of epoxy composites prepared by ball milling and sonication methods. (Reproduced with
permission from Ref. [197]).

25
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

Fig. 36. (a) Schematic showing the prepara­


tion process of NG/POE composites. (b) NG
concentration dependent TC of POE/NG com­
posites. Inset in (b) shows the anisotropic
factor of TC as a function of NG content.
(Reproduced with permission from Ref. [200]).
(c) Schematic showing the fabrication process
of PVDF composites with aligned GNFs using
an L-shaped tube. Photos in the bottom right
side show a cylindrical PVDF composite with
orientated GNFs (oGNF), and the sliced and
polished disk-type samples for TC tests. (d) TC
of PVDF composites with randomly dispersed
GNF and oriented GNF and TC predicted by the
Hashin-Shtrikman (H-S) model. (Reproduced
with permission from Ref. [201]).

Fig. 37. (a) Illustrations for ideal GNP (GnP) exfoliation and dispersion in IMF process. (b) TC of GNP/HDPE composites prepared by IMS, IMF and HPIMF plotted
against GNP loading. (c) TC of IMF samples versus degree of foaming (DP). (Reproduced with permission from Ref. [204]).

26
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

27.5 W/m·K was obtained for the compressed (at 250 bar) epoxy composites could reach 2.49 W/m·K with 12 wt.% EG, much higher
composite with 30 wt.% GNPs and 40 wt.% graphite flakes. This value than that of composites fabricated by melt blending or without CL⊕
is much larger than the TC of 16.0 W/m·K for the non-compressed intercalation [205].
composite.
Applying an electric or magnetic field is the conventional way to 4.5.4. Percolation phenomenon in GNP/polymer composites
align solid inclusion in a dielectric liquid or polymer melt. The align­ The electrical percolation phenomenon of composites comprising
ment of GNPs in epoxy resin by an AC electric field was investigated by electrically conductive particles within an insulating matrix can be
Wu et al. [203]. However, the GNP alignment only exhibits marginal predicted and clearly observed in numerous experiments, which can be
enhancement of TC when the GNP content is lower than 0.8 vol.%. described by the abrupt change of electrical conductivity when the
Here, TC of epoxy composites with aligned GNPs is 0.45 W/m·K com­ conductive fillers begin to form conductive networks through electron
pared to ∼0.40 W/m·K of the same composites with randomly dis­ hopping or tunnelling [206]. However, the existence of a thermal
persed GNPs. Starting from 0.8 vol.%, TCE decreases in epoxy compo­ percolation in composites is still controversial because phonon heat
sites with aligned GNPs as the GNP loading is increased [203]. conduction contributes to the main heat transfer. Shtein et al. claimed
Magnetic field induced alignment of GNPs also gives very marginal that a thermal percolation can be found in GNP/epoxy composites
increase in the TC of GNP/PVDF composites [138]. The maximum TC of [185]. It can be seen from Fig. 38a that, below the percolation
the composites with aligned GNPs reaches 0.587 W/m·K at 20 wt.% threshold of ∼18 vol.%, the TC increases slowly with increasing GNP
GNP, which is only 4.5 % higher than that of composites with randomly loading caused by the isolated GNPs and the large GNP/epoxy inter­
dispersed GNPs. In both cases, the marginal increases in TC may be facial thermal resistance. Above this percolation threshold, most of the
caused by the low GNP loading and the high viscosity of the GNP/ heat is transferred via the connected GNP network, yielding a rapid
polymer systems (melt or solution), which have greatly limited the thermal conduction enhancement as the loading increases. In this case,
formation of connected thermally conductive pathways. the TC of the composites may be limited by the high thermal boundary
Hamidinejad et al. [204] demonstrated a supercritical fluid (SCF) resistance at the filler/filler interface or filler/matrix interface, while
assisted fabrication strategy for thermally conductive GNP/HDPE (high the through-plane TC of GNP is still one to three orders of magnitude
density polyethylene) composites by using injection moulding. Three higher than that of epoxy matrix. Considering the electrical con­
methods were used: (a) injection-moulded solid (IMS, traditional in­ ductivity of GNPs is more than 15 orders of magnitude higher than that
jection moulding process without SCF treatment and foaming), (b) in­ of epoxy, it is reasonable to observe in Fig. 38b a lower electrical
jection-moulded foam (IMF, 0.4 wt.% supercritical N2 injected into the percolation threshold at 5 vol.% in these GNP/epoxy composites [185].
barrel), and (c) high-pressure-injection-moulded foam (HPIMF, 0.4 wt. Kargar et al. [207] also claimed the existence of thermal percolations in
% supercritical N2 injected into the barrel). For IMF, the SCF can in­ graphene/epoxy composites and gave a thermal percolation threshold
tercalate the layered graphite and weaken the bonding force of GNPs, of ∼30 vol.% that is again much higher than the electrical percolation
making exfoliation easier. Also, the plasticizing effect of SCF can en­ threshold of ∼10 vol.%. As mentioned earlier, the large percolation
hance the likelihood for the macro-molecular chains to penetrate the threshold differences may be attributed to the different thermal and
interlayer regions of GNPs owing to the enhanced mobility of the electrical conduction mechanisms, and the different ratios of thermal
macromolecules and increased interlayer distances of the GNPs. During and electrical conductivity of fillers to matrix, and the large interfacial
the phase transition, expansion of SCF can further separate and ex­ thermal resistance. Later, Naghibi et al. reported that there exists a
foliate the platelets. Fig. 37 shows the ideal process for exfoliation and thermal percolation threshold of 1.9 vol.% in non-curing mixtures of
dispersion of GNPs for the IMF case [204]. The SCF-treated HDPE graphene (grade H-15, XG-Sciences) and mineral oil TIMs, wherein
composite with 17.6 vol.% GNP shows a TC of 3.75 ± 0.12 W/m·K, addition of 1.9 vol.% of graphene increases the TC to 1.2 W/m·K from
which is much higher than that of the IMS counterpart (2.09 ± 0.03 W/ 0.3 W/m·K of pure mineral oil [208]. Several studies also reported the
m·K). Note that the degree of foaming must be controlled due to the low existence of critical filler loadings in thermally conductive composites
TC (0.026 W/m·K) of air and its strong phonon scattering effect [88]. In- [185,209,210]. However, the linear increase of TC with filler loading
situ intercalation polymerization was also used by Meng et al. to pre­ was also frequently found in polymer composites
pare thermally conductive GNP/PA6 composites [205]. First, capro­ [94,116,192,211–214].
lactam onium ion (CL⊕) formed by -caprolactam and H+ of 6-ami­
nocaproic acid could intercalate into the interlayer of expanded
4.5.5. Comparative study
graphite (EG) via π-π interaction. Then, the GNP/PA6 composites were
Besides 2D nano-carbon fillers, 1D nano-carbon fillers, such as CNTs
synthesized by in situ polymerization. In this way, TC of GNP/PA6
and carbon nanofibres (CNFs), also have ultrahigh intrinsic thermal

Fig. 38. (a) TC and (b) electrical conductivity of epoxy composites versus GNP loading (lateral size: 20 ± 7 μm, thickness: 15-100 nm, aspect ratio: 568 ± 298).
(Reproduced with permission from Ref. [185]).

27
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

conductivity [215]. Fan et al. compared the TCE of commercially thermally conductive graphene fillers.
available GNPs, multi-walled CNTs (MWCNTs) and CNFs in eicosane (c) Use graphene fillers to improve the TC of polymer composites and
(C20H42) composite materials at loading levels lower than 1.6 vol.% simultaneously use other fillers to suppress electrical conduction,
[216]. GNPs are found to show much higher efficiency of TCE when thus fabricating thermally conductive but electrically insulating
compared with other nano-carbon fillers, exhibiting promising poten­ polymer composites.
tial for the preparation of thermally conductive composite materials.
Compared to 1D CNTs and CNFs, the superiority of GNPs in enhancing 5.1. Synergetic effects of GNP and other fillers
the TC of composites originates from their unique 2D planar structure,
which gives reduced nano-carbon/polymer thermal interface re­ Synergetic effects are to utilize fillers with different sizes and/or
sistance. In another study [214], the effects of relatively high loadings different dimensions to realize the formation of long-range thermally
(< 20 wt.%) of several different carbon allotropes (i.e., CB, CNT, GNP, conductive paths at relatively low total filler loading. As a typical 2D
graphite, pitch-based carbon fiber [PCF] and EG) on the TC of poly­ filler, GNPs have been widely used together with other fillers, such as
merized cyclic butylene terephthalate composites were investigated. 0D nanoparticle, 1D CNT, 2D BN and RGO, to synergistically enhance
The results (Fig. 39) showed that the morphology (sizes and shapes) of the TC of polymer composites [245–251]. Yu et al. found synergetic
the fillers plays an important role on the TCE of polymer composites. effects on TCE in PC composites containing GNP/CNT hybrid fillers. At
However, among these diverse fillers, micro-sized 1D PCF and EG dis­ a filler loading of 20 wt.%, TC of composite with hybrid fillers is in­
play superiority in enhancing the TC of composites because of the de­ creased to 1.39 W/m·K from 1.13 W/m·K of GNP composite [220]. Si­
creased phonon scattering at their interfaces [214]. All the results in­ milar synergetic effect was found in a comparative study of epoxy
dicate that decreasing interfaces between filler and matrix reduces the composites with individual fillers and hybrid fillers (Fig. 40). At a filler
interface thermal resistance and weakens the phonon scattering. loading of 10 wt.%, the epoxy composites with single-walled CNT
(SWCNT) gave a TC of 0.85 W/m·K and with GNP showed a TC of
5. TC of graphene/polymer composites: Influence of hybrid fillers 1.49 W/m·K, while the composites with hybrid filler (i.e., GNP/SWCNT)
exhibited a maximum TC of 1.75 W/m·K at a GNP:SWCNT ratio of 3:1
Graphene-based fillers together with other fillers are often jointly (Fig. 40a) [245]. The synergetic effects are attributed to the replace­
used as hybrid fillers to enhance the TC of polymer composites. The ment of CNT/CNT or GNP/GNP point contacts by linear contacts be­
utilization of hybrid fillers has three main goals: tween CNT and GNP (Fig. 40b). However, the synergetic effects shown
in [245] can only be found in composites with hybrid loading from 5 to
(a) Maximize the TC and TCE at a certain filler loading to take ad­ 20 wt.%; and in the work of Huang et al., the synergetic effect of GNP
vantage of the synergetic effects. Here, composites processing may and CNT at high filler loading can be achieved [252] (see Fig. 40d). As
become much easier and other properties, such as mechanical displayed in Fig. 40c, the synergetic effect of TC is seen to be much
flexibility of polymer, can be largely preserved. stronger at higher hybrid filler loading (> 25 wt.%). The effects were
(b) Further improve the performance of functional polymer composites found to originate from the enhanced random dispersion of GNPs in­
by enhancing the TC through addition of graphene. For example, to duced by the presence of CNTs. That is, GNPs tend to orientate along
meet the requirements of thermal management, packaging mate­ the in-plane direction; but the introduction of MWCNTs restrains the
rials may need to be filled by both reinforcement fillers and orientation of GNPs and improves their dispersion. The realization of

Fig. 39. SEM images of (a) CB, (b) CNT, (c) GNP, (d) graphite, (e) PCF and (f) EG. TC of polymerized cyclic butylene terephthalate composites with different carbon
filler. (Reproduced with permission from Ref. [214]).

28
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

Fig. 40. (a) TC of epoxy composites with GNPs and GNP/SWCNT hybrid fillers. (b) Schematic illustrating the GNP/SWCNT network in composites. (c) TC of epoxy
composites with individual filler and hybrids. (d) Synergistic effect of hybrids reported in different investigations. (Reproduced with permission from Ref. [245,252]).

strong synergetic effects at high hybrid filler loading was attributed to [253]. The composites with 20 wt.% CNT@GNP hybrids show 300 %
the unique sample preparation method, which guaranteed high sample and 50 % enhancement of the through-plane TC compared to those
quality. composites with individual CNTs and GNPs, respectively. Additionally,
Unlike CNT/GNP mixtures as hybrids, CNT grafted GNP compared to the composites with mixed CNT/GNP (2.15 W/m·K), a
(CNT@GNP) is also utilized as hybrids to enhance the TC of epoxy higher TC (2.41 W/m·K) is obtained in CNT@GNP composites. This is

Fig. 41. (a) Schematic showing the prepara­


tion process of PS composites with three dif­
ferent structures. (b) TC of PS composites
versus filler loading. (Reproduced with per­
mission from Ref. [249].) SEM image showing
(c) the UGF-CNT hybrid foam structure and (d)
the tip of Al2O3 layer covered single CNT
bundle. (e) TC of different composites. (Re­
produced with permission from Ref. [254]).

29
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

because the simply mixed CNT/GNP are separated in the composites drying. Composites were subsequently obtained by infiltration of the 3D
due to the density difference, whereas CNT@GNP cannot separate in interconnected BN-GO skeleton with liquid epoxy and followed by
the composites. curing. N, N-dimethylbenzylamine was used as the curing and reducing
Constructing interconnected GNP networks can improve the TCE in agent, which simultaneously converted GO into RGO during the curing
composites. Here, by introducing other fillers, e.g., CNTs, in the matrix process. At 5.17 vol.% of skeleton (Fig. 42b), BN-GO plates in the
or growing thermally conductive fillers on GNPs, the TCE can be in­ skeleton curl because of the limited space for ice crystal growth. At
creased. Indeed, high TCE has been found in composites having a GNP higher GO loading, the increased solution viscosity restricts the move­
network and MWCNT/PS particles, designated (MWCNT/PS)@GNPs ment of BN-GO plates and some particles are embedded within the ice
[249]. Fig. 41a shows 3 different hybrid composites: (a) MWCNT/ crystals, finally producing a homogeneous interconnected structure
GNPs/PS with randomly dispersed MWCNT and GNPs in PS, (b) (Fig. 42c). Composites with 13.16 vol.% BN-RGO show a maximum
(MWCNT@GNPs)/PS with a GNPs/MWCNT hybrid network and seg­ through-plane TC of 5.05 W/m·K, which is much higher than those
regated PS particles, and (c) (MWCNT/PS)@GNPs. Compared to GNPs/ values of composites with randomly dispersed BN (R-h-BN) and BN-
PS, for composites with higher than 3.5 wt.% total loading of fillers as RGO (R-BN-RGO). These results also confirm that the added BN-RGO
in Fig. 41b, (MWCNT/PS)@GNPs exhibit synergistically enhanced TC; hybrids have contributed to forming thermally conductive pathways in
however, the other two hybrid filler cases, MWCNT/GNPs/PS and the through-plane direction. An et al. [259] prepared highly anisotropic
(MWCNT@GNPs)/PS, yield lower TC values. The synergistic effect RGO/BN hybrid aerogels by hydrothermally treating the GO/BN sus­
between MWCNT/PS particles and GNP network may be caused by the pension, air-drying and finally high-temperature annealing. RGO were
double-network bridging effect and the increased density of filler net­ self-assembled to a long-range ordered network with high anisotropy
work [249]. Kholmanov et al. showed that the TC of UGF (UGF: ultra- during hydrothermal treatment, while BN distributed between the
thin graphite foam) based erythritol composites could be increased by aligned RGO could avoid volume shrinkage of RGO aerogel during air-
growing long CNTs onto the graphite skeleton [254]. SEM images drying. The RGO/BN (GO:BN = 5:1) hybrid aerogel composites exhibit
shown in Fig. 41c,d reveal the UGF-CNT hybrid foam structure: CNT a TC of 3.06 W/m·K, which is much higher than 0.87 W/m·K of BN
bundles exhibit curved shapes and the bundle tips are covered by an composites. Also, TC of RGO/BN hybrid aerogel composites increases
Al2O3 layer. As shown in Fig. 41e, the TC of UGF-CNT/erythritol with increasing annealing temperature and gives a through-plane TC of
composites is increased from 2.26 ± 0.3 W/m·K of the UGF/erythritol 11.01 W/m·K at 2000 °C. By GO-assisted gelation and following hot-
composites to 4.09 ± 0.3 W/m·K. However, TC decreases to 2.90 W/ pressing, Li et al. prepared flexible and thermally conductive natural
m·K from 4.09 W/m·K when ground UGF-CNT powder is used as hybrid rubber composite films, which displayed an in-plane TC of 16.04 W/
filler [254]. The enhanced TC of UGF-CNT/erythritol composites is m·K when the ratio of natural rubber:GO:BN is 1:25:10 [260]. Shao
attributed to the lower interfacial thermal resistance between CNTs and et al. reported a large enhancement effect of BNNSs on TCE of RGO/
UGF, which can be achieved through direct connections between PA6 composites [261]. 3D BNNSs/RGO framework was self-assembled
themselves. by introducing BNNSs into the 3D RGO framework; and PA6 composites
were prepared by in-situ polymerization of PA6 in the 3D framework. It
5.2. GO- and RGO-based hybrid fillers was found that incorporating 1.6 wt.% BNNSs in the composites with
6.8 wt.% 3D RGO could increase TC by 87.6 %, reaching 0.89 W/m·K.
GO can be dissolved in water to form hydrogels via assembly and Zhao et al. reported a remarkable synergistic effect between FLG
thus 3D architecture (i.e., foam, skeleton, aerogel) of GO or RGO can be and RGO foam in enhancing the TC of PDMS composites [262]. FLGs
easily constructed by drying the hydrogels. Combining the lightweight were added to 0.2 vol.% RGO foam/PDMS composite. At 2.7 vol.%
and tuneable density, the interconnected structure makes GO or RGO FLGs, TC of FLG/RGO/PDMS composite reaches 1.08 W/m·K, which is
3D architecture attractive for high efficiency TCE in polymer matrix. In 80 % and 184 % higher than those of 2.7 vol.% FLG/PDMS and RGO/
addition, other highly thermally conductive fillers, such as BN, GNP, PDMS composites, respectively. It is, however, noted that graphene-
CNT and FLG, can be introduced to further increase the TC of compo­ based hybrid fillers may have a negative effect on TCE. Zhang et al.
sites [255–257]. Using ball milled GO-BN hybrid, Yao et al. prepared studied the thermal and electrical conductivities of epoxy composites
3D RGO-BN skeleton to study the TCE of epoxy (Fig. 42a) [258]. First, filled with fullerene decorated RGO (fullerene@RGO) [263]. Although
BN-GO hybrid suspension was directionally frozen and 3D inter-con­ the electrical conductivity increase of highly loaded fullerene@RGO
nected BN-GO skeleton ice was prepared by low-temperature vacuum composites can be more than 17 orders of magnitude (i.e., from 10−14

Fig. 42. (a) Illustration of the fabrication of 3D


BN-RGO/epoxy composites. (b) SEM images of
cross-section of 3D BN-GO skeleton with 5.17
vol.% and the content of GO is 3.4 wt.% of BN-
GO hybrid. (c) SEM images of cross-section of
3D BN-GO skeleton with 13.16 vol.% and the
content of GO is 3.4 wt.% of BN-GO hybrid. (d)
TC of epoxy composites with R-h-BN, R-BN-
RGO, 3D BN-RGO. (Reproduced with permis­
sion from Ref. [258]).

30
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

to ∼3 × 103 S/m), TC only gives a two-fold increase (i.e., from 0.30 to enhances the TC, and simultaneously, in most cases, the electrical
0.66 W/m·K). Such a TCE (i.e., 1.2) is much lower in comparison with conductivity will also be increased. However, for some TIMs under high
the various reported RGO/epoxy composites. It is believed that the low voltage and as underfills in electronic packaging, electrical insulation is
TCE is mainly caused by phonon scattering of fullerene/RGO interface an essential requirement [273,274]. Thus, any electrical conductivity
and an isolation effect of fullerene on graphene/graphene contact. The enhancement must be avoided. A conventional method is to use gra­
latter isolation effect has been indirectly proven in a comparative in­ phene-based hybrids as thermal conductive fillers for the polymer
vestigation on the TC of composites with fullerene@RGO and a physical composites. For example, alumina coated RGO [275,276], alumina
mixture of fullerene and RGO [263]. coated GNPs [277,278], silica coated RGO [279,280], silica coated
GO or RGO solution can be used to prepare anisotropic composite graphite [281,282], silicon carbide (SiC) coated graphite [283], alu­
films by filtration or solution casting. Other fillers like MWCNT and BN minium nitride (AlN) coated GO [284] and Al(OH)3 coated GO
nanosheet (BNNS) are also used to further enhance the TC or impart [125,127] have been reported as effective fillers for preparation of
multi-functionalities to the GO/RGO composites [264,265]. Hence, thermally conductive but electrically insulating polymer composites.
RGO-encapsulated SiO2 (RGO@ SiO2) were used as fillers to fabricate Silica nanoparticles coated RGO have been used to prepare ther­
multifunctional epoxy composites, which not only gave enhanced TC mally conductive but electrically insulating epoxy composites [279].
but also much improved dielectric and thermal-mechanical properties Composites with 1.0 wt.% silica coated RGO display a TCE factor of
[266]. Zhang et al. studied PVA composites with enhanced in-plane TC 0.61 in comparison with epoxy resin (from 0.20 W/m·K of epoxy to
and excellent mechanical properties based on the synergistic effects of 0.32 W/m·K of the composite) and well-preserved electrically insulating
functionalized BNNSs/GO network at relatively low loading [267]. properties (i.e., electrical resistivity > 109 Ω·cm). The silica layer ef­
Also, some GO/RGO hybrid systems were used to increase both flame fectively prevents the electrons from hopping or tunnelling from one
retardancy and TC of composites. Al2O3/f-RGO/epoxy ternary compo­ graphene sheet to another, resulting in high electrical resistivity of the
sites possessed both higher TC and improved flame retardancy perfor­ composites. However, it is noted that the nanoparticles (e.g., silica,
mance relative to Al2O3/epoxy composites [268]. Silver nanowires and alumina) used to coat graphene usually have much lower TC than
flame-retardant functionalized RGOs could also enhance both TC and graphene. Together with their abovementioned insulation effect, a de­
flame retardancy of epoxy composites owing to the effects of bridging creased efficiency of TCE is unavoidable. Kim et al. indeed found that
and catalytic charring of the functionalized RGOs [269]. epoxy composites with Al(OH)3 coated GO show much lower TC in
Vertically aligning graphene sheets in a composite can maximize the comparison with epoxy composites with uncoated GO [125,127]. By
TCE along the through-plane direction [270]. Using this strategy, Li contrast, however, several studies reveal that composites with coated
et al. prepared epoxy composites with ultrahigh through-plane TC graphene exhibit higher TC compared to those filled with uncoated
[271]. Typically, they prepared highly aligned and graphitized gra­ graphene [28,29,279,283]. Sun et al. reported an alumina coated GNP
phene aerogels by (a) directional-freezing GO hydrogels, (b) freeze- (Al2O3@GNP) network synthesized by a solution-assisted deposition
drying and (c) graphitizing at 2800 °C. The epoxy composite with 0.75 method and found significant TCE in Al2O3@GNP/epoxy composites
vol.% vertically aligned graphitized graphene aerogel exhibited a compared to GNP/epoxy composites [277]. Hsiao et al. [279] explained
through-plane TC of 6.57 W/m·K, which is 37 times higher in com­ these results by suggesting that the coatings improved the graphene
parison with neat epoxy resin. Taking this further with hybrids, as dispersion and decreased the elastic modulus mismatch between gra­
shown in Fig. 43a, An et al. prepared vertically aligned hybrid foams by phene and epoxy. In addition, whereas pristine RGO sheets might be
(a) hydrothermal reduction of a GO/GNP suspension, (b) air-drying and easily rippled and wrinkled, the nanoparticle coating would tend to
(c) annealing at different temperature [272]. During the hydrothermal flatten the graphene sheets. Therefore, long TC pathways could be
reduction step, the RGO sheets formed a vertically aligned inter­ formed along the nanoparticle coated graphene, giving high TC of the
connected network and prevented the shrinkage of the RGO foams composites. Indirect proof comes from studies on TC of composites
during the air-drying process. The graphitization of the hybrid foams at filled with uncoated graphite and silica coated graphite [285], where
2800 °C removed the oxygen-containing groups and healed the defects lower TC is obtained in epoxy composites with silica coated graphite
of RGO. These factors resulted in vertically aligned high-quality hybrid flakes. Note that both silica coated and uncoated graphite flakes have
graphene foams with an ideal porous architecture so that the epoxy intrinsic flat structures. This implies that the coating would increase the
composites with 19.0 vol.% hybrid graphene foam displayed an ultra­ thermal resistance between the graphite sheets. By dispersing PDA
high through-plane TC of 35.5 W/m·K (Fig. 43b) [272]. coated GO (GO@PDA) into a PS microsphere suspension, Yuan et al.
prepared 3D GO interconnected PS nanocomposites. As shown in
5.3. TC of electrically insulating polymer composites with graphene as filler Fig. 44a, the nanocomposite with 0.96 vol.% GO@PDA exhibits both
high through-plane (4.56 W/m·K) and in-plane (4.13 W/m·K) thermal
The introduction of graphene-based fillers into a polymer matrix conductivities. Moreover, the same nanocomposite shows an excellent

Fig. 43. (a) Schematic illustration showing the


fabrication process of epoxy composite with
high through-plane TC, including (a) hydro­
thermally assembling GO and GNP in the
aqueous suspension to form vertically aligned
RGO/GNP hybrid foams, (b) air-drying the
hybrid foams, (c) thermally annealing the hy­
brid foams to form aligned graphene aerogels,
and (d) infiltrating epoxy into the graphene
aerogels. (b) TC of annealed air-dried graphene
hybrid foams (GHF)/epoxy composites at dif­
ferent annealing temperature. (Reproduced
with permission from Ref. [272]).

31
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

Fig. 44. (a) TC and (b) electrical resistivity of GO@PDA/PS composites. (Figures were replotted using the test data provided by the authors of Ref. [152]).

electrically insulating property in Fig. 44b (> 1014 Ω·cm) [152]. 5 vol.% graphene, however, the TC of epoxy composites increases re­
In addition, as an insulating material, GO has been used to further markably to 9.9 W/m·K with a TCE factor of ∼5.0. Another example is
increase the TC of traditional thermally conductive but electrically in­ f-RGO filled BN/PI composites. Fig. 45b shows BN has low TCE factors
sulating composites [258,286,287]. For example, Huang et al. used GO up to 50 wt.% loading in the composites. The introduction of 1.0 wt.%
as a modifier of AlN and found the encapsulation of AlN by GO sig­ f-RGO results in a TC of 2.11 W/m·K from 0.74 W/m·K [290]. For ad­
nificantly enhances the TC of epoxy composites when the AlN loading is hesives, low filler content is more attractive because it almost does not
low (< 55 vol.%) [288]. But when the AlN loading is very high (i.e., change the viscosity and the curing conditions. Messina et al. [248]
beyond the percolation threshold), introducing GO imparts a sub­ investigated the effect of a small amount of GNPs on the thermal and
stantial increase of viscosity (i.e., low flowability) of the uncured epoxy electrical conductivities of a commercial conductive adhesive com­
composites. Hence, the composites contain pores and voids, yielding prising epoxy resin and ∼78 % sliver platelets and has a TC of 2.5 W/
low TC. m·K. Starting from 0.005 wt.% GNPs, TC is increased to 4.0 W/m·K and
further rises to a maximum value of 11.8 W/m·K at 0.01 wt.% GNPs
5.4. TC of conventional thermally conductive polymer composites by (Fig. 45c). A remarkable enhancement of TC can also be observed at
graphene-based fillers 0.03 and 0.06 wt.% GNPs. Beyond 0.2 wt.% GNPs, however, the TC is
significantly reduced. This trend agrees with that of electrical con­
Since traditional filler composites often have low TCE factors, gra­ ductivity (that is, the reciprocal of electrical resistivity) versus GNP
phene-based fillers can be used as a second filler to enhance their TC. loading (Fig. 45d) [248]. The decreased enhancement of TC at higher
The first successful example is the increased TCE of silver/epoxy GNP loading is associated with the agglomeration of excessive fillers
composites by graphene. It was shown that adding a small amount of and/or introduction of defects such as voids and pores.
graphene can result in significant enhancement of TC in silver/epoxy
composites [289]. As shown in Fig. 45a, it is clear that traditional
silver/epoxy composites have a TC of 1.67 W/m·K. After the addition of

Fig. 45. (a) TC of silver/epoxy composites


versus graphene loading. (Reproduced with
permission from Ref. [289]). (b) TC of PI
composites with individual BN or RGO (f-BN:
TCA modified BN). (Reproduced with permis­
sion from Ref. [290]). (c) Effects of GNP
loading on TC and (d) electrical resistivity of
the commercial conductive adhesive com­
prising epoxy resin and ∼78 % sliver platelets.
(Reproduced with permission from Ref. [248]).

32
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

6. Applications of thermally conductive graphene/polymer When graphene/polymer composites are used as TIMs, the most
composites popular matrix is epoxy resin. The TCE achieved is better than the
traditional carbon materials, such as graphite and CB. Besides, some
The high TC of graphene/polymer composites presents opportu­ flexible graphene-based polymer composites with high TC also have
nities for a wide range of different applications. Among these, elec­ great potential in thermal management in many fields [300]. In fact,
tronic packaging materials, PCMs, TIMs and circuit board materials are when SLG and MLG are jointly used, the TCE is much more effective
the most important, since they utilize high TC of the composites for due to the synergistic effects obtained (Fig. 29). Indeed, most graphene/
effective thermal dissipation. The effects can be clearly and visibly il­ polymer composites with high TC can be considered or used as TIMs.
lustrated when they are compared to non-thermally conductive mate­ Surprisingly, few studies have been conducted on grease/paste
rials. (See, for example, thermal images of LED on graphene-BN/ based graphene composites as TIMs despite demonstrations that gra­
polymer composites versus FR4 printed circuit board substrates in phene can be an excellent filler for TCE. At a low filler loading of 2.0
Fig. 46b). However, as most studies on graphene-based composites for vol.%, the TC of commercial thermal grease is increased from an initial
electronic packaging materials and TIMs are often similar to routine value of ∼5.8 W/m·K to 14 W/m·K [116]. Though the matrix is a
materials research which are mainly focused on materials character­ commercial thermal grease with decent high TC, the final TCE achieved
ization, measurements of thermal and electrical conductivities, and by adding graphene is still very remarkable. The graphene used is
other physical and mechanical properties, we will not adopt this ap­ fabricated from natural graphite via ultrasonic-assisted liquid exfolia­
proach to avoid overlaps with preceding sections. Since applications are tion, followed by centrifugation separation. Hence, it is believed to be
the ultimate goal of materials research, we will review recent in­ multi-layered and with high quality. In another study [136], RGO is
vestigations on graphene-based TIMs (i.e., polymer composites/ found to be the most effective filler to enhance the heat transfer
greases/fluids), PCMs and battery separators as three examples to shed properties of silicone, and GNP is slightly inferior. However, silicone
light on future potential use of graphene-based highly thermally con­ grease containing RGO exhibits a high viscosity because of its porous
ductive polymer composites. structure, thus GNP is still the best choice to fabricate grease with high
filler loading and a TC of ∼1.0 W/m·K can be achieved with 4.25 vol.%.
Therefore, thermally conductive grease with graphene as filler pos­
6.1. Thermal interface materials sesses better thermal performance than GNP with thickness of tens of
nanometers [301,302]. However, they also induce a high viscosity in
TIMs are used to reduce the thermal contact resistance between two the grease and seriously suppress their processibility.
surfaces, such as the surface of heat-sinks and the surface of a packed As for thermally conductive fluids, graphene materials are not an
Central Processing Unit (CPU) schematically shown in Fig. 47. They excellent choice for applications because their high-aspect-ratio will
include thermal fluids, thermal grease (pastes), resilient thermal con­ lead to significant unacceptable increase in viscosity of the fluids. But
ductors and solders (applied in the molten state) [292]. TC is a critical advantages do exist. For example, the high aspect ratio may enhance
parameter for TIMs [293–298]. When commercial TIMs are filled with the TCE, and GO can be well dispersed in fluids without even surfac­
∼4.0 wt.% graphene, the temperature rise can be reduced by ∼10 °C in tants. To maintain a stable dispersion, low filler loadings, e.g., 0.1 wt.
the CPU after 15 h operation. This temperature reduction is commer­ %, are normally used. Indeed, the stability of the thermally conductive
cially important for practical applications because a decrease of tem­ fluids obtained are decent: in 2 months (6, 5, 3 months are reported), no
perature rise of ∼20 °C can increase the device lifetime by an order of obvious concentration change is observed [303–305]. The base fluids
magnitude [299].

Fig. 46. (a) Photo of FR4 printed circuit board and graphene-BN/polymer composites having a LED mounted on top of each; and (b) thermal images of the LED
mounted on two different substrates. (Reproduced with permission from Ref. [291]).

33
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

Fig. 47. (a) Schematic illustrating the action of


thermal interface material, which fills the gaps
between two contacting surfaces. The heat re­
moval improves with higher TC, smaller bond
line thickness and lower contact resistance of
the material. (Reproduced with permission
from Ref. [116]). (b) TC of a commercial TIM
with graphene. (c) Temperature rise versus
time inside a computer CPU using the com­
mercial TIM or graphene filled commercial
TIM. Inset in (c) shows the backside of a CPU
with TIM. (Reproduced with permission from
Ref. [299]).

are usually water, ionic liquid, propyl glycol or liquid paraffin. Gen­ randomly dispersed graphene leads to only 2.5-fold TC improvement
erally, a low filler loading has a limited effect on the TCE. For example, [13]. Moreover, thermal annealing (> 2200 ℃) is very effective to
when graphene is 0.25 wt.%, the enhancement efficiency is only 33.9 % obtain defect-free graphene [165]. A high TC of thermally annealed
at 20 ℃; and when the temperature is increased to 40 ℃, the en­ defect-free graphene/PCM can reach 3.55 W/m·K with 10 wt.% gra­
hancement is up to 47.5 % [306]. The enhancements achieved in other phene. This is an enhancement efficiency of > 600 % compared to un-
reports are close to these values [117,218,304,307–317]. Dramatically annealed graphene/PCM composites at a similar filler loading, and a
increased viscosity and instability at higher filler loadings are believed 16-fold increase to pure PCM (1-octadecanol). When defective pristine
to be the obstacle to achieve better performance [310]. Nonetheless, graphene is used as fillers, a TC of only 0.55 W/m·K is obtained with 10
Wang et al. reported the viscosity of an ionic liquid with graphene wt.% filler. This result clearly reveals the importance of graphene
added might be lower than that of the base fluid due to the self-lu­ quality.
brication effect of graphene [318]. Fortunately, the fluid viscosity Apart from the graphene quality, the number of graphene layers
would decrease sharply at high temperature [319]. also greatly affects the TC that can be obtained in graphene/PCM
It should be noted that currently studies on graphene involved TIMs composites [169]. Thus, the TCE factor varies from ∼2 to 25 at 20 vol.
focused primarily on material performance, while many related tech­ % graphene whose number of layers is between 3 and 44. These results
nical issues have not been fully considered. Some typical issues are: are not only due to the relatively lower intrinsic TC of MLG but are also
mismatch of thermal expansion coefficients, simplification of installa­ caused by the interface thermal resistance, morphologies of graphene
tion process and reduction of installation pressure of TIMs among with different layers in polymer matrix and the percolation threshold.
others. While common sense would suggest thin perfect graphene materials
are preferred fillers to achieve high TC in polymer composites, the
current best TCE is instead achieved with continuous 3D graphene
6.2. Phase change materials structures as fillers for PCM [13,338,348]. For example, with a very low
loading (0.8–1.2 vol.%) of ultrathin graphite foam, the TC of a PCM is
Organic PCMs are normally used as latent heat energy storage and enhanced up to 18 times (∼2.8 W/m·K) and the influences on the
release media for effective thermal management. Pristine PCMs usually melting temperature or mass specific heat of fusion are negligible [13].
have a very low TC, which results in a slow transient temperature re­ The ultrathin graphite foams were fabricated using Ni foam templates.
sponse and reduced heat transfer efficiency [320,321]. To enhance TCs The main point to achieve a very high TC at very low filler fraction is to
and facilitate phase change of PCMs, enclosing graphene materials is a utilize 3D continuous filler structure instead of dispersed fillers and to
means usually employed to increase the temperature during heating up anneal the fillers at very high temperature (3000 ℃). High temperature
and decrease the temperature more rapidly during cooling of the whole annealing makes the continuous graphite struts possessing TC up to
PCM [13,134,165,169,171,187,225,303,319,322–344]. Different from 840 W/m·K and the strut structure of the graphite fillers forms con­
conventional thermally conductive polymer composites, polymers with tinuous thermal conductive paths in the PCM matrix. It is noted that,
phase transition functions, such as PEG, n-hexadecane, 1-octadecanol, here, the graphite foam has a wall thickness of hundreds of nanometers,
n-eicosane, n-dodecanol and paraffin are often used as matrices which is much thicker than even MLGs. This claims to give a graphite
[165,335,337,345–347]. strut/PCM interface insensitive to phonon scattering that is crucial to
The quality of graphene varies with different preparation methods, achieving high TCE, since the interface phonon scattering greatly re­
which would result in diversified TCE of graphene/PCMs. For instance, duces the basal-plane TC of graphene materials. In another study, Min
0.8–1.2 vol.% of 3D ultrathin-graphite foams embedded in PCM can et al. showed that at 3.6 ± 0.2 wt.%, the anisotropic high-quality
increase its TC up to 18-fold, while in other experiments, 4.0 wt.%

34
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

graphene aerogel-based paraffin PCM composites exhibit a high reveal the excellent thermal management of such PCMs in power bat­
through-plane TC of 8.87 W/m·K, resulting in a TCE factor of 24.34 and tery. It can be seen that the maximum surface temperature of the
an exceptional latent heat retention of 98.7 % [349]. They prepared the wrapped power battery is 47.8 °C when the discharging rate is 2.3 C.
graphene aerogels in four steps: (a) fabricating vertically aligned However, the non-wrapped power battery reaches 61.0 °C.
monoliths by freezing the suspensions of polyamic acid (PAA) salt and
GO, (b) freeze-drying, (c) imidization at 300 °C and (d) graphitization at
2800 °C. 6.3. Battery separators
The application of graphene filled PCMs for battery pack thermal
management is illustrated in Fig. 48 by Goli et al. [171]. Clearly, gra­ Separator is an important component of battery systems that elec­
phene-filled PCMs show large reductions of temperature rise inside the trically isolate the cathode from the anode to avoid internal short cir­
battery pack. Computer simulation also confirms the significant effect cuits. However, low affinity to electrolyte and poor thermal properties
of graphene filled PCMs on thermal management of battery pack. In the of commercial polyolefin separators have seriously undermined the
case of graphene-filled PCMs, battery cylinders exhibit the lowest electrochemical behaviours and safety of batteries [355]. The applica­
temperature of ∼310 K–315 K. In combination with TC data shown in tion of graphene-based materials in battery separators has been in­
Fig. 32, it can be concluded that the increased TCs in graphene-filled vestigated due to its high TC and rich functional groups [356]. Typical
PCMs significantly improve the thermal management capability of Li- graphene/polymer composite separators can be divided into hybrid
ion battery packs. separators and coated separators.
In addition, graphene filled PCMs with high TC have great potential The polymers commonly used in developing hybrid separators in­
for high-power-density energy harvesting, such as solar-thermal energy clude PAN [357], PMMA [358], PVDF [359] and PU [360]. PAN and
conversion [350,351]. These PCMs not only can improve the thermal- PMMA nanofiber separators incorporated with graphene were fabri­
diffusion-based charging rate, but also enable cost-effective thermal cated by the electrospinning technique [358]. The TC of PAN and
management applications [352,353]. Wu et al. reported highly ther­ PMMA nanocomposite fibers increased with graphene loading (Fig. 50).
mally conductive PCMs with largely aligned GNP and their application At 8 wt.% graphene, the TC of PMMA and PAN composite fibers
in thermal energy storage [354]. As illustrated in Fig. 49a,b, the heat reached 5.0 W/m·K and 2.7 W/m·K, respectively, which could improve
capacity of the PCM-based thermal energy harvesting device remains thermal management of batteries [361,362]. Besides TC, electrolyte
quite stable after ∼100 cycles of charge and discharge. Moreover, the wettability, dielectric constant, and ionic conductivity of the nano­
TC of these PCMs can be tailored in the range of 4.4–35.0 W/m·K at composite separators also increasd with the addition of graphene. As for
filler loading below 40 wt.%. The IR images shown in Fig. 49c also coated separators, the batteries with graphene composite separators
display significant enhancement of electrochemical performance [363].

Fig. 48. (a) Setup for battery testing. Six 4 V Li-ion cells are placed in an Al battery pack, and each has a capacity of 3000 mA h. Graphene filled PCM (IGI-1260) is
used for cooling the Li-ion battery pack. (b) Measured temperature fluctuations at battery cathode, battery anode and battery pack and ambient temperature. (c) Rise
of temperature inside the battery pack of first ten charge-discharge cycles in the cases of no PCM, paraffin, PCM with 1 wt.% graphene and PCM with 20 wt.%
graphene. (d) Simulated temperature in the pack of a battery. (Reproduced with permission from Ref. [171]).

35
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

Fig. 49. (a) Schematic showing the heat design of the thermal energy harvesting system. (b) Thermal capacity versus cycle times of the thermal energy harvesting
system. The inset in (b) is a photo of the thermal energy harvesting system. (c) Surface temperature of two power battery units. The insets in (c) show the photos of
the tested power battery units and their IR images during the discharging process. (Reproduced with permission from Ref. [354]).

the superb high TC and very high aspect ratio. The high aspect ratio
makes the utilization of the high TC of graphene more effective to
achieve high TC composites. However, it is noted that many problems
in current graphene/polymer composites are also associated with these
two advantages. Some of these problems are identified below:

(1) The extremely high TC of graphene is very sensitive to many fac­


tors. The large TC can only be preserved in perfect graphene
structure, defects and modification to graphene and even the in­
teraction between graphene and matrix will unavoidably result in
remarkable reduction of TC. This incurs two sub-problems: (a)
Graphene derived from different fabrication methods actually pos­
sesses different TC. Graphene fabricated by CVD, peeling graphite
or sonication exfoliation may have very high quality, but they have
extremely low yield or multi-layered structures. Highly yielded GO
or RGO is known to possess low TC caused by the defective struc­
tures and the quality of RGO which greatly depends on the reduc­
tion agents/methods used and high temperature post-annealing. (b)
The contradiction between surface modification and high TC of
graphene. Surface modification is of primary importance to im­
Fig. 50. TC of (a) PAN and (b) PMMA composite fibers plotted against gra­ prove the interfacial affinity between graphene and polymer and
phene nanoflake loading. (Reproduced with permission from Ref. [357]). graphene dispersion in matrix. However, any functional groups
attached on graphene will induce local phonon scattering and in
Li et al. studied a bi-layer composite separator by coating PA-grafted turn ruin the intrinsic TC. Detailed studies on balancing these two
GO molecular brushes onto the PP separator. The cells with these contradictions have been wanting to-date.
composite separators exhibited long-term stability for over 2600 h at a (2) High aspect ratio results in the serious problem of dispersion. While
high current density (2 mA/cm2) [364]. Zhu et al. explored the effect of theoretically high aspect ratio benefits TC achievable in composites
RGO reduction on the electrochemical characteristics of Li-S cells with at high filler loading, the high viscosity of graphene/polymer melt
coated separators [365]. It was found that RGO with higher reduction or solution mixture makes it very difficult to fabricate such gra­
degrees possessed better polysulfide rejection performance, which phene/polymer composites. In addition, high aspect ratio induces
provides a promising strategy in the RGO selection for battery separa­ folding and twisting of graphene sheets which are difficult to
tors. However, the detailed role and mechanism of the TC of separators overcome, and in turn they may lead to serious reductions of TC of
on battery performance was seldom reported. the graphene sheets [80,84]. High aspect-ratio MLG with high
flatness should be better than thinner graphene sheets and thicker
7. Summary and outlook graphite platelets to achieve high TC in polymer composites.
(3) There are no effective approaches to curb the interfacial thermal
For the development of highly thermally conductive polymer com­ resistance between graphene and polymer matrix as well as the
posites, graphene materials as fillers have at least two key advantages: contact resistance between graphene sheets in the composites. A

36
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

graphene/polymer composite fabrication approach that has a high unsolved interfacial and contact thermal resistances, the expected
potential to achieve high TC at low graphene loading is to use high TC has not been realized.
graphene aerogel which is a self-supported structure of graphene (4) Synergetic effects, which have been widely studied in polymer
sheets. Although it is expected that the pre-formed thermal con­ composites, must be well designed to overcome the weakness of
ductive paths will give high TC of fabricated composites, the TC graphene as a filler. First, synergetic effects greatly benefit forma­
values obtained by this method is still much lower than expected. tion of thermally conductive paths in polymer composites; second,
This is attributed to the high interfacial thermal resistance and hybridizing graphene and other fillers with different dimensions
contact resistance which are not very well managed. and particle sizes can improve the graphene dispersion in polymer
(4) Despite the high expectations, the advantages of utilizing graphene matrices by overcoming the adverse self-formed layered structure.
in highly thermally conductive polymer composites have not been Hence, combining graphene with other highly thermally conductive
well demonstrated. Actually, most data achieved with graphene are fillers may be the ultimate solution for current practical applica­
not much superior to those by CNTs or thin graphite platelets. tions of graphene-based materials as thermally conductive fillers.
Instead, an obvious observation is that the quality and formation of (5) Graphene has anisotropic TC and the in-plane TC can be more than
thermal conductive paths are more important. Some CNTs and thin two orders of magnitude higher than the through-plane TC. Thus,
graphite platelets possess much more perfect structure than GO and much research work had been done to obtain polymer composites
RGO, which lead to even better performance. Also, 3D structure of with anisotropic TC. Such methods include aligning graphene by
graphite/graphene platelets may be more effective than 2D thin applying electric/magnetic field, electrospinning, mechanical
layer graphene for TCE [13,161,179,366–371], indicating the for­ straining, injection moulding, two-roll milling, self-assembly, ice
mation of 3D thermally conductive paths in polymer composites is template, filtration and solution casting. Using these methods, it is
more important than the thin structure of graphene. possible to design and fabricate graphene/polymer composites with
(5) Graphene is highly electrically conductive and they also give sig­ either high in-plane TC or high through-plane TC. In the future,
nificant electrical conductivity enhancement of the polymer com­ particular attention should be given to practical industrial appli­
posites when increasing the TC. In this case, the development of cations of these methods, but not just seeking the highest TC values.
thermally conductive but electrically insulating polymer compo­ Researchers should address the critical issue of what is the TC re­
sites is curtailed. Coating a high-electrical-resistivity layer, such as quired for the particular end-use.
inorganic nanoparticles and dielectric polymers, on graphene-based (6) The role of TC of graphene and its composites in practical appli­
fillers retains the electrical insulation of the composites. However, cations should be investigated in depth. Hence, new measurement
TCE of the composites is also influenced by the additional inter­ and characterization techniques must be developed, which should
facial and contact thermal resistances. In addition, these composites not only give TC of materials in service but should also reflect the
cannot withstand high voltage applications due to the electrically performance variation induced by the enhanced TC. One example is
conductive nature of graphene [121,372]. graphene nanocomposite separators for batteries. While high TC
separators may improve the thermal management of batteries, the
Looking to the future, to achieve the desired high TC in polymer detailed role and mechanism of how the TC improves the battery
composites with graphene as filler, the following issues should be ad­ performance has not been revealed.
dressed with good engineering solutions:
Declaration of Competing Interest
(1) High yield and large-scale production of high-quality graphene is
still on demand and is critical for the development of polymer The authors declare that they have no known competing financial
composites with high qualities. Today, almost all the methods de­ interests or personal relationships that could have appeared to influ­
veloped for large-scale fabrication of graphene materials involve ence the work reported in this paper.
severe conditions, yielding defective samples. The TC of these de­
fective graphene materials may be low to tens of W/m·K in spite of Acknowledgements
the high intrinsic TC up to thousands of W/m·K of the perfect
graphene lattice. Therefore, most graphene currently used as fillers X.H. and G.W. gratefully acknowledge the supports from the
for polymer composites actually does not truly justify the name of National Natural Science Foundation of China (Nos. U19A20105). Y.-
“graphene” as far as TC is concerned. W.M. thanks the support from the Australian Research Grants Council
(2) For processing and fabrication of graphene/polymer composites, it (DP130104648).
is widely agreed that some methods, e.g., surface modification,
using graphene with large lateral sizes, etc., are very effective; References
however, studies to avoid their negative effects are much fewer. For
example, surface modification on graphene is a typical double-edge [1] X. Huang, P. Jiang, T. Tanaka, IEEE Electr. Insul. Mag. 27 (2011) 8–16.
sword: while it can solve the problems of dispersion and interfacial [2] Y. Huang, C. Ellingford, C. Bowen, T. McNally, D. Wu, C. Wan, Int. Mater. Rev. 65
phonon scattering, it can also destroy the high intrinsic TC of gra­ (2020) 129–163.
[3] C. Huang, X. Qian, R. Yang, Mater. Sci. Eng. R Rep. 132 (2018) 1–22.
phene. High aspect ratio graphene may increase the TCE factor, it [4] G. Pandey, E.T. Thostenson, Polym. Rev. 52 (2012) 355–416.
also imparts dramatically higher mixture viscosity, poor dispersion, [5] A.A. Balandin, Nat. Mater. 10 (2011) 569–581.
folding and twisting of graphene sheets, etc. Hence, it is possible to [6] M. Mu, C. Wan, T. McNally, 2d Mater. 4 (2017) 042001.
[7] E. Pop, V. Varshney, A.K. Roy, MRS Bull. 37 (2012) 1273–1281.
use low aspect ratio graphene to achieve high TC in polymer [8] Y. Su, J.J. Li, G.J. Weng, Carbon 137 (2018) 222–233.
composites with better dispersibility than high aspect ratio gra­ [9] Y. Sun, L. Chen, L. Cui, Y. Zhang, X. Du, Comput. Mater. Sci. 148 (2018) 176–183.
phene. More basic work is needed on these issues to balance the [10] J. Song, C. Chen, Y. Zhang, Compos. Part A Appl. Sci. Manuf. 105 (2018) 1–8.
[11] Y.S. Yeom, K.Y. Cho, H.Y. Seo, J.S. Lee, D.H. Im, C.Y. Nam, H.G. Yoon, Compos.
processing conditions. Sci. Technol. 186 (2020) 107915.
(3) Great efforts are needed on 3D graphene/polymer composites. It is [12] Y. Fu, J. Hansson, Y. Liu, S. Chen, A. Zehri, M.K. Samani, N. Wang, Y. Ni, Y. Zhang,
a promising approach to achieve high TC with preformed 3D Z.-B. Zhang, Q. Wang, M. Li, H. Lu, M. Sledzinska, C.M. Sotomayor Torres, S. Volz,
A.A. Balandin, X. Xu, J. Liu, 2D Mater. 7 (2020) 012001.
thermal conductive paths. However, at present, as a result of

37
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

[13] H. Ji, D.P. Sellan, M.T. Pettes, X. Kong, J. Ji, L. Shi, R.S. Ruoff, Energy Environ. Sci. Adv. Funct. Mater. 28 (2018) 1706954.
7 (2014) 1185–1192. [66] T. Wang, J. Tsai, Comput. Mater. Sci. 122 (2016) 272–280.
[14] L. Cui, Y. Zhang, X. Du, G. Wei, J. Mater. Sci. 53 (2017) 4242–4251. [67] S. Chien, Y. Yang, Co. Chen, Appl. Phys. Lett. 98 (2011) 033107.
[15] A. Bianco, H. Cheng, T. Enoki, Y. Gogotsi, R.H. Hurt, N. Koratkar, T. Kyotani, [68] S. Chien, Y. Yang, Co. Chen, Carbon 50 (2012) 421–428.
M. Monthioux, C.R. Park, J.M.D. Tascon, J. Zhang, Carbon 65 (2013) 1–6. [69] A.H. Aref, H. Erfan-Niya, A.A. Entezami, J. Mater. Sci. 51 (2016) 6824–6835.
[16] S.S. Chen, A.L. Moore, W.W. Cai, J.W. Suk, J.H. An, C. Mishra, C. Amos, [70] W. Huang, Q. Pei, Z. Liu, Y. Zhang, Chem. Phys. Lett. 552 (2012) 97–101.
C.W. Magnuson, J.Y. Kang, L. Shi, R.S. Ruoff, ACS Nano 5 (2011) 321–328. [71] Y.Y. Zhang, Q.X. Pei, X.Q. He, Y.-W. Mai, Chem. Phys. Lett. 622 (2015) 104–108.
[17] S. Chen, Q. Wu, C. Mishra, J. Kang, H. Zhang, K. Cho, W. Cai, A.A. Balandin, [72] J. Oh, H. Yoo, J. Choi, J.Y. Kim, D.S. Lee, M.J. Kim, J.-C. Lee, W.N. Kim,
R.S. Ruoff, Nat. Mater. 11 (2012) 203–207. J.C. Grossman, J.H. Park, S.-S. Lee, H. Kim, J.G. Son, Nano Energy 35 (2017)
[18] A.A. Balandin, S. Ghosh, W. Bao, I. Calizo, D. Teweldebrhan, F. Miao, C.N. Lau, 26–35.
Nano Lett. 8 (2008) 902–907. [73] S.J. Mahdizadeh, E.K. Goharshadi, J. Nanoparticle Res. 16 (2014) 2553.
[19] S. Ghosh, W. Bao, D.L. Nika, S. Subrina, E.P. Pokatilov, C.N. Lau, A.A. Balandin, [74] A.K. Majee, Z. Aksamija, Phys. Rev. B 93 (2016) 235423.
Nat. Mater. 9 (2010) 555–558. [75] D.L. Nika, A.S. Askerov, A.A. Balandin, Nano Lett. 12 (2012) 3238–3244.
[20] D.L. Nika, S. Ghosh, E.P. Pokatilov, A.A. Balandin, Appl. Phys. Lett. 94 (2009) [76] H.D. Wang, H. Takamatsu, X. Zhang, Width-Dependent Thermal Conductivity of
203103. Suspended Single Layer Graphene Proceedings of the Asian Conference on
[21] S. Ghosh, I. Calizo, D. Teweldebrhan, E.P. Pokatilov, D.L. Nika, A.A. Balandin, Thermal Sciences 2017, Jeju Island, Korea, 2017, pp. 1–4.
W. Bao, F. Miao, C.N. Lau, Appl. Phys. Lett. 92 (2008) 151911. [77] H. Cao, Z. Guo, H. Xiang, X. Gong, Phys. Lett. A 376 (2012) 525–528.
[22] N. Wang, M.K. Samani, H. Li, L. Dong, Z. Zhang, P. Su, S. Chen, J. Chen, S. Huang, [78] Y. Liu, H. Yang, N. Liao, P. Yang, RSC Adv. 4 (2014) 54474–54479.
G. Yuan, X. Xu, B. Li, K. Leifer, L. Ye, J. Liu, Small 14 (2018) 1801346. [79] M.M. Sadeghi, I. Jo, L. Shi, Proceedings of the National Academy of Sciences of the
[23] L. Peng, Z. Xu, Z. Liu, Y. Guo, P. Li, C. Gao, Adv. Mater. 29 (2017) 1700589. United States of America, (2013), pp. 16321–16326 110.
[24] D.L. Nika, A.A. Balandin, Rep. Prog. Phys. 80 (2017) 036502. [80] K. Chu, W. Li, F. Tang, Phys. Lett. A 377 (2013) 910–914.
[25] T.R. Anthony, W.F. Banholzer, J.F. Fleischer, L. Wei, P.K. Kuo, R.L. Thomas, [81] N. Yang, X. Ni, J. Jiang, B. Li, Appl. Phys. Lett. 100 (2012) 093107.
R.W. Pryor, Phys. Rev. B 42 (1990) 1104–1111. [82] S.-P. Ju, K.Y. Chen, M.C. Lin, Y.R. Chen, Y.L. Lin, J.-W. Chang, S.-C. Huang,
[26] J. Gong, Z. Liu, J. Yu, D. Dai, W. Dai, S. Du, C. Li, N. Jiang, Z. Zhan, C.-T. Lin, Carbon 77 (2014) 36–43.
A. Composites Part, Applied Science and Manufacturing 87 (2016) 290–296. [83] X. Wei, G. Guo, T. Ouyang, H. Xiao, J. Appl. Phys. 115 (2014) 154313.
[27] Y. Guo, C. Dun, J. Xu, J. Mu, P. Li, L. Gu, C. Hou, C.A. Hewitt, Q. Zhang, Y. Li, [84] H. Li, H. Ying, X. Chen, D.L. Nika, A.I. Cocemasov, W. Cai, A.A. Balandin, S. Chen,
D.L. Carroll, H. Wang, Small 13 (2017) 1702645. Nanoscale 6 (2014) 13402–13408.
[28] D.L. Nika, E.P. Pokatilov, A.S. Askerov, A.A. Balandin, Phys. Rev. B 79 (2009) [85] A.I. Cocemasov, D.L. Nika, A.A. Balandin, Phys. Rev. B 88 (2013) 035428.
155413. [86] L. Cui, X. Du, G. Wei, Y. Feng, J. Phys. Chem. C 120 (2016) 23807–23812.
[29] E. Munoz, J. Lu, B.I. Yakobson, Nano Lett. 10 (2010) 1652–1656. [87] C. Zhang, X.L. Hao, C.X. Wang, N. Wei, T. Rabczuk, Sci. Rep. 7 (2017) 41398.
[30] L. Lindsay, D.A. Broido, N. Mingo, Phys. Rev. B 83 (2011) 235428. [88] S. Stankovich, D.A. Dikin, G.H. Dommett, K.M. Kohlhaas, E.J. Zimney, E.A. Stach,
[31] L. Lindsay, D.A. Broido, N. Mingo, Phys. Rev. B 82 (2010) 115427. R.D. Piner, S.T. Nguyen, R.S. Ruoff, Nature 442 (2006) 282–286.
[32] C. Shao, X.X. Yu, N. Yang, Y.A. Yue, H. Bao, Nanoscale Microscale Thermophys. [89] C.-W. Nan, R. Birringer, D.R. Clarke, H. Gleiter, J. Appl. Phys. 81 (1997)
Eng. 21 (2017) 201–236. 6692–6699.
[33] Y. Xu, Z. Li, W. Duan, Small 10 (2014) 2182–2199. [90] T. Luo, J.R. Lloyd, Adv. Funct. Mater. 22 (2012) 2495–2502.
[34] X. Li, J. Chen, C. Yu, G. Zhang, Appl. Phys. Lett. 103 (2013) 013111. [91] D. Konatham, A. Striolo, Appl. Phys. Lett. 95 (2009) 163105.
[35] J. Hu, S. Schiffli, A. Vallabhaneni, X. Ruan, Y.P. Chen, Appl. Phys. Lett. 97 (2010) [92] B. Mortazavi, O. Benzerara, H. Meyer, J. Bardon, S. Ahzi, Carbon 60 (2013)
133107. 356–365.
[36] N. Khosravian, M.K. Samani, G.C. Loh, G.C.K. Chen, D. Baillargeat, B.K. Tay, [93] L. Hu, T. Desai, P. Keblinski, J. Appl. Phys. 110 (2011) 033517.
Comput. Mater. Sci. 79 (2013) 132–135. [94] A. Yu, P. Ramesh, M.E. Itkis, E. Bekyarova, R.C. Haddon, J. Phys. Chem. C 111
[37] D. Yang, F. Ma, Y. Sun, T. Hu, K. Xu, Appl. Surf. Sci. 258 (2012) 9926–9931. (2007) 7565–7569.
[38] G. Rajasekaran, P. Narayanan, A. Parashar, Crit. Rev. Solid State Mater. Sci. 41 [95] S. Lin, M.J. Buehler, Nanotechnology 24 (2013) 165702.
(2015) 47–71. [96] D. Konatham, K.N.D. Bui, D.V. Papavassiliou, A. Striolo, Mol. Phys. 109 (2011)
[39] H.F. Zhan, Y.Y. Zhang, J.M. Bell, Y.-W. Mai, Y.T. Gu, Carbon 77 (2014) 416–423. 97–111.
[40] J.J. Yeo, Z. Liu, T.Y. Ng, Nanotechnology 23 (2012) 385702. [97] Y. Gao, F. Muller-Plathe, J. Phys. Chem. B 120 (2016) 1336–1346.
[41] Y.Y. Zhang, Y. Cheng, Q.X. Pei, C.M. Wang, Y. Xiang, Phys. Lett. A 376 (2012) [98] A. Lakshmanan, S. Srivastava, A. Ramazani, V. Sundararaghavan, Appl. Phys. Lett.
3668–3672. 112 (2018) 151902.
[42] B. Mortazavi, S. Ahzi, Carbon 63 (2013) 460–470. [99] Y. Zhang, Y. Zhao, S. Bai, X. Yuan, Compos. Part B Eng. 106 (2016) 324–331.
[43] H. Yang, Y. Tang, J. Gong, Y. Liu, X. Wang, Y. Zhao, P. Yang, S. Wang, J. Mol. [100] L. Zhang, L. Liu, ACS Appl. Mater. Interfaces 9 (2017) 28949–28958.
Model. 19 (2013) 4781–4788. [101] M. Li, H. Zhou, Y. Zhang, Y. Liao, H. Zhou, Carbon 130 (2018) 295–303.
[44] T.Y. Ng, J.J. Yeo, Z.S. Liu, Carbon 50 (2012) 4887–4893. [102] C. Diao, Y. Dong, J. Lin, Int. J. Heat Mass Transf. 112 (2017) 903–912.
[45] H. Malekpour, P. Ramnani, S. Srinivasan, G. Balasubramanian, D.L. Nika, [103] T. Wang, P. Tseng, J. Tsai, J. Compos. Mater. 53 (2018) 835–847.
A. Mulchandani, R.K. Lake, A.A. Balandin, Nanoscale 8 (2016) 14608–14616. [104] J.W. Liu, Y.S. Ye, Y. Xue, X.L. Xie, Y.-W. Mai, Journal of Polymer Science Part A-
[46] B. Mortazavi, A. Rajabpour, S. Ahzi, Y. Rémond, S. Mehdi Vaez Allaei, Solid State Polymer Chemistry 55 (2017) 622–631.
Commun. 152 (2012) 261–264. [105] M. Wang, N. Hu, L. Zhou, C. Yan, Carbon 85 (2015) 414–421.
[47] H. Yang, Y. Tang, Y. Liu, X. Yu, P. Yang, React. Funct. Polym. 79 (2014) 29–35. [106] S. Shen, A. Henry, J. Tong, R. Zheng, G. Chen, Nat. Nanotechnol. 5 (2010)
[48] B.S. Lee, J. Phys. Condens. Matter 30 (2018) 295302. 251–255.
[49] C. Si, L. Li, G. Lu, B. Cao, X. Wang, Z. Fan, Z. Feng, J. Appl. Phys. 123 (2018) [107] Y. Wang, C. Yang, Q.X. Pei, Y. Zhang, ACS Appl. Mater. Interfaces 8 (2016)
135101. 8272–8279.
[50] P. Plachinda, D. Evans, R. Solanki, J. Heat Transfer 134 (2012) 122401. [108] Z. Zabihi, H. Araghi, Phys. Lett. A 380 (2016) 3828–3831.
[51] C. Shao, Q.Y. Rong, N.B. Li, H. Bao, Phys. Rev. B 98 (2018) 155418. [109] X. Shen, Z. Wang, Y. Wu, X. Liu, J.-K. Kim, Carbon 108 (2016) 412–422.
[52] C. Shao, Q.Y. Rong, M. Hu, H. Bao, J. Appl. Phys. 122 (2017) 155104. [110] Y. Wang, C.H. Yang, Y.-W. Mai, Y.Y. Zhang, Carbon 102 (2016) 311–318.
[53] A.V. Savin, Y.S. Kivshar, B. Hu, Phys. Rev. B 82 (2010) 195422. [111] Q. Pei, Y. Zhang, Z. Sha, V.B. Shenoy, Appl. Phys. Lett. 100 (2012) 101901.
[54] C. Yu, G. Zhang, J. Appl. Phys. 113 (2013) 214304. [112] H. Hou, W. Dai, Q. Yan, L. Lv, F.E. Alam, M. Yang, Y. Yao, X. Zeng, J.-B. Xu, J. Yu,
[55] Z. Guo, J.W. Ding, X. Gong, Phys. Rev. B 85 (2012) 235429. N. Jiang, C.-T. Lin, J. Mater. Chem. A 6 (2018) 12091–12097.
[56] J. Chen, G. Zhang, B.W. Li, Nanoscale 5 (2013) 532–536. [113] X. Xu, J. Chen, J. Zhou, B. Li, Adv. Mater. 30 (2018) 1705544.
[57] A.N. Sidorov, D.K. Benjamin, C. Foy, Appl. Phys. Lett. 103 (2013) 243103. [114] X. Shen, Z. Wang, Y. Wu, X. Liu, Y.B. He, J.K. Kim, Nano Lett. 16 (2016)
[58] I. Vlassiouk, S. Smirnov, I. Ivanov, P.F. Fulvio, S. Dai, H. Meyer, M. Chi, 3585–3593.
D. Hensley, P. Datskos, N.V. Lavrik, Nanotechnology 22 (2011) 275716. [115] S. Ganguli, A.K. Roy, D.P. Anderson, Carbon 46 (2008) 806–817.
[59] A. Reina, X. Jia, J. Ho, D. Nezich, H. Son, V. Bulovic, M.S. Dresselhaus, J. Kong, [116] K.M. Shahil, A.A. Balandin, Nano Lett. 12 (2012) 861–867.
Nano Lett. 9 (2009) 30–35. [117] S.H. Song, K.H. Park, B.H. Kim, Y.W. Choi, G.H. Jun, D.J. Lee, B.S. Kong,
[60] A. Sood, R. Cheaito, T. Bai, H. Kwon, Y. Wang, C. Li, L. Yates, T. Bougher, K.W. Paik, S. Jeon, Adv. Mater. 25 (2013) 732–737.
S. Graham, M. Asheghi, M. Goorsky, K.E. Goodson, Nano Lett. 18 (2018) [118] C. Wu, X. Huang, G. Wang, L. Lv, G. Chen, G. Li, P. Jiang, Adv. Funct. Mater. 23
3466–3472. (2013) 506–513.
[61] B. Mortazavi, M. Potschke, G. Cuniberti, Nanoscale 6 (2014) 3344–3352. [119] X.Y. Huang, B.Y. Sun, C.Y. Yu, J.D. Wu, J. Zhang, P.K. Jiang, High Volt. 5 (2020)
[62] S.N. Raja, D. Osenberg, K. Choi, H.G. Park, D. Poulikakos, Nanoscale 9 (2017) 387–396.
15515–15524. [120] M. Ghassemi, High Volt. 5 (2020) 7–14.
[63] W. Lee, K.D. Kihm, H.G. Kim, S. Shin, C. Lee, J.S. Park, S. Cheon, O.M. Kwon, [121] Y. Lin, X.Y. Huang, J. Chen, P.K. Jiang, High Volt. 2 (2017) 139–146.
G. Lim, W. Lee, Nano Lett. 17 (2017) 2361–2366. [122] P. Zong, J. Fu, L. Chen, J. Yin, X. Dong, S. Yuan, L. Shi, W. Deng, RSC Adv. 6
[64] D. Lee, S. Lee, B.-S. An, T.-H. Kim, C.-W. Yang, J.W. Suk, S. Baik, Chem. Mater. 29 (2016) 10498–10506.
(2017) 10409–10417. [123] M.C. Vu, G. Park, Y. Bae, M.J. Yu, T.K. An, S. Lee, S. Kim, Macromol. Res. 24
[65] M.M. Bernal, A. Di Pierro, C. Novara, F. Giorgis, B. Mortazavi, G. Saracco, A. Fina, (2016) 1070–1076.

38
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

[124] W. Du, Z. Zhang, H. Su, H. Lin, Z. Li, Ind. Eng. Chem. Res. 57 (2018) 7146–7155. [175] F. Zhang, C. Ye, W. Dai, L. Lv, Q. Yuan, K.W.A. Chee, K. Yang, N. Jiang, C.-T. Lin,
[125] J. Kim, H. Im, J.-m. Kim, J. Kim, J. Mater. Sci. 47 (2011) 1418–1426. Z. Zhan, D. Dai, H. Li, Chinese Chem. Lett. 31 (2020) 244–248.
[126] S. Wang, M. Tambraparni, J. Qiu, J. Tipton, D. Dean, Macromolecules 42 (2009) [176] Q. Li, Y. Guo, W. Li, S. Qiu, C. Zhu, X. Wei, M. Chen, C. Liu, S. Liao, Y. Gong,
5251–5255. A.K. Mishra, L. Liu, Chem. Mater. 26 (2014) 4459–4465.
[127] J. Kim, B.-s. Yim, J.-m. Kim, J. Kim, Microelectron. Reliab. 52 (2012) 595–602. [177] H.X. Zeng, J.Y. Wu, Y.P. Ma, Y.S. Ye, J.W. Liu, X.W. Li, Y. Wang, Y.G. Liao,
[128] A.J. Glover, M. Cai, K.R. Overdeep, D.E. Kranbuehl, H.C. Schniepp, X.B. Luo, X.L. Xie, Y.-W. Mai, ACS Appl. Mater. Interfaces 10 (2018)
Macromolecules 44 (2011) 9821–9829. 41690–41698.
[129] S. Ye, J. Feng, Polym. Chem. 4 (2013) 1765–1768. [178] Y. Li, F. Xu, Z. Lin, X. Sun, Q. Peng, Y. Yuan, S. Wang, Z. Yang, X. He, Y. Li,
[130] I.H. Tseng, J. Chang, S. Huang, M. Tsai, Polym. Int. 62 (2013) 827–835. Nanoscale 9 (2017) 14476–14485.
[131] W. Sun, L. Wang, Z. Yang, T. Zhu, T. Wu, C. Dong, G. Liu, Chem. Mater. 30 (2018) [179] H. Fang, Y. Zhao, Y. Zhang, Y. Ren, S.L. Bai, ACS Appl. Mater. Interfaces 9 (2017)
7473–7483. 26447–26459.
[132] J.I. Paredes, S. Villar-Rodil, A. Martı´nez-Alonso, M.D. Tasco´n, Langmuir 24 [180] X. Shen, Z. Wang, Y. Wu, X. Liu, Y. He, Q. Zheng, Q. Yang, F. Kang, J. Kim, Mater.
(2008) 10560–10564. Horiz. 5 (2018) 275–284.
[133] S.Y. Lee, P. Singh, R.L. Mahajan, Carbon 145 (2019) 131–139. [181] D.C. Zhu, Y.Y. Ren, G.X. Liao, S.L. Jiang, F.H. Liu, J.J. Guo, G.J. Xu, J. Appl.
[134] F. Yavari, H.R. Fard, K. Pashayi, M.A. Rafiee, A. Zamiri, Z. Yu, R. Ozisik, T. Borca- Polym. Sci. 134 (2017) 45332.
Tasciuc, N. Koratkar, J. Phys. Chem. C 115 (2011) 8753–8758. [182] C. Li, X.-L. Zeng, L.-Y. Tan, Y.-M. Yao, D.-L. Zhu, R. Sun, J.-B. Xu, C.-P. Wong,
[135] G. Xue, J. Zhong, S. Gao, B. Wang, Carbon 96 (2016) 871–878. Chem. Eng. J. 368 (2019) 79–87.
[136] W. Yu, H. Xie, L. Chen, Z. Zhu, J. Zhao, Z. Zhang, Phys. Lett. A 378 (2014) [183] H.Z. Zhou, H.J. Wang, X.S. Du, Y.Y. Zhang, H.M. Zhou, H. Yuan, H.Y. Liu, Y.-
207–211. W. Mai, Carbon 139 (2018) 1168–1177.
[137] J.R. Potts, S. Murali, Y. Zhu, X. Zhao, R.S. Ruoff, Macromolecules 44 (2011) [184] X. Sun, P. Ramesh, M.E. Itkis, E. Bekyarova, R.C. Haddon, J. Phys. Condens. Matter
6488–6495. 22 (2010) 334216.
[138] H. Guo, X. Li, B. Li, J. Wang, S. Wang, Mater. Des. 114 (2017) 355–363. [185] M. Shtein, R. Nadiv, M. Buzaglo, K. Kahil, O. Regev, Chem. Mater. 27 (2015)
[139] F. Luo, K. Wu, J. Shi, X. Du, X. Li, L. Yang, M. Lu, J. Mater. Chem. A 5 (2017) 2100–2106.
18542–18550. [186] B. Debelak, K. Lafdi, Carbon 45 (2007) 1727–1734.
[140] Y. Wang, H.F. Zhan, Y. Xiang, C. Yang, C.M. Wang, Y.Y. Zhang, J. Phys. Chem. C [187] J. Xiang, L.T. Drzal, Sol. Energy Mater. Sol. Cells 95 (2011) 1811–1818.
119 (2015) 12731–12738. [188] F. Wang, L.T. Drzal, Y. Qin, Z. Huang, J. Mater. Sci. 50 (2015) 1082–1093.
[141] Z. Tang, H. Kang, Z. Shen, B. Guo, L. Zhang, D. Jia, Macromolecules 45 (2012) [189] W. Song, W. Wang, L.M. Veca, C.Y. Kong, M. Cao, P. Wang, M.J. Meziani, H. Qian,
3444–3451. G.E. LeCroy, L. Cao, Y. Sun, J. Mater. Chem. 22 (2012) 17133–17139.
[142] N. Song, X. Hou, L. Chen, S. Cui, L. Shi, P. Ding, ACS Appl. Mater. Interfaces 9 [190] W.-L. Song, L.M. Veca, C.Y. Kong, S. Ghose, J.W. Connell, P. Wang, L. Cao, Y. Lin,
(2017) 17914–17922. M.J. Meziani, H. Qian, G.E. LeCroy, Y.-P. Sun, Polymer 53 (2012) 3910–3916.
[143] N. Song, J. Yang, P. Ding, S. Tang, L. Shi, Compos. Part A Appl. Sci. Manuf. 73 [191] L.M. Veca, M.J. Meziani, W. Wang, X. Wang, F. Lu, P. Zhang, Y. Lin, R. Fee,
(2015) 232–241. J.W. Connell, Y.-P. Sun, Adv. Mater. 21 (2009) 2088–2092.
[144] M. Fang, K. Wang, H. Lu, Y. Yang, S. Nutt, J. Mater. Chem. 20 (2010) 1982–1992. [192] J. Gu, X. Yang, Z. Lv, N. Li, C. Liang, Q. Zhang, Int. J. Heat Mass Transf. 92 (2016)
[145] E. Cho, J. Huang, C. Li, C. Chang-Jian, K. Lee, Y. Hsiao, J. Huang, Carbon 102 15–22.
(2016) 66–73. [193] W.L. Chen, K. Wu, Q. Liu, M.G. Lu, Polymer 186 (2020) 122075.
[146] B. Zhong, H. Dong, Y. Luo, D. Zhang, Z. Jia, D. Jia, F. Liu, Compos. Sci. Technol. [194] A.A. Tarhini, A.R. Tehrani-Bagha, Compos. Sci. Technol. 184 (2019) 107797.
151 (2017) 156–163. [195] S. Araby, Q.S. Meng, L.Q. Zhang, H.L. Kang, P. Majewski, Y.H. Tang, J. Ma,
[147] C. Teng, C.M. Ma, C. Lu, S. Yang, S. Lee, M. Hsiao, M. Yen, K. Chiou, T. Lee, Carbon Polymer 55 (2014) 201–210.
49 (2011) 5107–5116. [196] Y.J. Ren, Y.F. Zhang, H.C. Guo, R.C. Lv, S.L. Bai, Composites Part A-Applied
[148] A.A. Vasileiou, M. Kontopoulou, A. Docoslis, ACS Appl. Mater. Interfaces 6 (2014) Science and Manufacturing 126 (2019) 105578.
1916–1925. [197] W. Guo, G. Chen, J. Appl. Polym. Sci. 131 (2014) 40565.
[149] Y. Guo, K. Ruan, X. Shi, X. Yang, J. Gu, Compos. Sci. Technol. 193 (2020) 108134. [198] Y. Li, H. Zhang, H. Porwal, Z. Huang, E. Bilotti, T. Peijs, Compos. Part A Appl. Sci.
[150] J.R. Potts, O. Shankar, L. Du, R.S. Ruoff, Macromolecules 45 (2012) 6045–6055. Manuf. 95 (2017) 229–236.
[151] O. Eksik, S.F. Bartolucci, T. Gupta, H. Fard, T. Borca-Tasciuc, N. Koratkar, Carbon [199] M. Saeidijavash, J. Garg, B. Grady, B. Smith, Z. Li, R.J. Young, F. Tarannum, N. Bel
101 (2016) 239–244. Bekri, Nanoscale 9 (2017) 12867–12873.
[152] H. Yuan, Y. Wang, T. Li, Y.J. Wang, P.M. Ma, H.J. Zhang, W.J. Yang, M.Q. Chen, [200] C.P. Feng, L. Bai, Y. Shao, R.Y. Bao, Z.Y. Liu, M.B. Yang, J. Chen, H.Y. Ni, W. Yang,
W.F. Dong, Nanoscale 11 (2019) 11360–11368. Adv. Mater. Interfaces 5 (2018) 1700946.
[153] W. Lv, C. Zhang, Z.J. Li, Q.H. Yang, J. Phys. Chem. Lett. 6 (2015) 658–668. [201] H. Jung, S. Yu, N.S. Bae, S.M. Cho, R.H. Kim, S.H. Cho, I. Hwang, B. Jeong,
[154] F. Conrado, M. Pavese, Journal of Nanomaterials 2017 (2017) 8974174. J.S. Ryu, J. Hwang, S.M. Hong, C.M. Koo, C. Park, ACS Appl. Mater. Interfaces 7
[155] M. Qin, Y. Xu, R. Cao, W. Feng, L. Chen, Adv. Funct. Mater. 28 (2018) 1805053. (2015) 15256–15262.
[156] W. Zhao, J. Kong, H. Liu, Q. Zhuang, J. Gu, Z. Guo, Nanoscale 8 (2016) [202] A. Ohayon-Lavi, M. Buzaglo, S. Ligati, S. Peretz-Damari, G. Shachar, N. Pinsk,
19984–19993. M. Riskin, Y. Schatzberg, I. Genish, O. Regev, Carbon 163 (2020) 333–340.
[157] F. Luo, K. Wu, H. Guo, Q. Zhao, M. Lu, Compos. Sci. Technol. 132 (2016) 1–8. [203] S.Y. Wu, R.B. Ladani, J. Zhang, E. Bafekrpour, K. Ghorbani, A.P. Mouritz,
[158] P. Kumar, S. Yu, F. Shahzad, S.M. Hong, Y.-H. Kim, C.M. Koo, Carbon 101 (2016) A.J. Kinloch, C.H. Wang, Carbon 94 (2015) 607–618.
120–128. [204] S.M. Hamidinejad, R.K.M. Chu, B. Zhao, C.B. Park, T. Filleter, ACS Appl. Mater.
[159] J. Chen, X. Huang, B. Sun, Y. Wang, Y. Zhu, P. Jiang, ACS Appl. Mater. Interfaces 9 Interfaces 10 (2018) 1225–1236.
(2017) 30909–30917. [205] F. Meng, F. Huang, Y. Guo, J. Chen, X. Chen, D. Hui, P. He, X. Zhou, Z. Zhou,
[160] Y. Guo, G. Xu, X. Yang, K. Ruan, T. Ma, Q. Zhang, J. Gu, Y. Wu, H. Liu, Z. Guo, J. Compos. Part B Eng. 117 (2017) 165–173.
Mater. Chem. C 6 (2018) 3004–3015. [206] H.M. Kim, M.S. Choi, J. Joo, S.J. Cho, H.S. Yoon, Phys. Rev. B 74 (2006) 054202.
[161] G. Lian, C.-C. Tuan, L. Li, S. Jiao, Q. Wang, K.-S. Moon, D. Cui, C.-P. Wong, Chem. [207] F. Kargar, Z. Barani, R. Salgado, B. Debnath, J.S. Lewis, E. Aytan, R.K. Lake,
Mater. 28 (2016) 6096–6104. A.A. Balandin, ACS Appl. Mater. Interfaces 10 (2018) 37555–37565.
[162] J.D. Renteria, S. Ramirez, H. Malekpour, B. Alonso, A. Centeno, A. Zurutuza, [208] S. Naghibi, F. Kargar, D. Wright, C.Y.T. Huang, A. Mohammadzadeh, Z. Barani,
A.I. Cocemasov, D.L. Nika, A.A. Balandin, Adv. Funct. Mater. 25 (2015) R. Salgado, A.A. Balandin, Adv. Electron. Mater. 6 (2020) 1901303.
4664–4672. [209] H. Wu, L.T. Drzal, Polym. Compos. 34 (2013) 2148–2153.
[163] G. Xin, T. Yao, H. Sun, S.M. Scott, D. Shao, G. Wang, J. Lian, Science 349 (2015) [210] J. Gu, C. Xie, H. Li, J. Dang, W. Geng, Q. Zhang, Polym. Compos. 35 (2013)
1083–1087. 1087–1092.
[164] V. Cecen, R. Thomann, R. Mülhaupt, C. Friedrich, Polymer 132 (2017) 294–305. [211] G. Li, X. Tian, X. Xu, C. Zhou, J. Wu, Q. Li, L. Zhang, F. Yang, Y. Li, Compos. Sci.
[165] G. Xin, H. Sun, S.M. Scott, T. Yao, F. Lu, D. Shao, T. Hu, G. Wang, G. Ran, J. Lian, Technol. 138 (2017) 179–185.
ACS Appl. Mater. Interfaces 6 (2014) 15262–15271. [212] F.E. Alam, W. Dai, M. Yang, S. Du, X. Li, J. Yu, N. Jiang, C.-T. Lin, J. Mater. Chem.
[166] C. Teng, D. Xie, J. Wang, Z. Yang, G. Ren, Y. Zhu, Adv. Funct. Mater. 27 (2017) A 5 (2017) 6164–6169.
1700240. [213] G. Gedler, M. Antunes, T. Borca-Tasciuc, J.I. Velasco, R. Ozisik, Eur. Polym. J. 75
[167] G. George, S.B. Sisupal, T. Tomy, B.A. Pottammal, A. Kumaran, V. Suvekbala, (2016) 190–199.
R. Gopimohan, S. Sivaram, L. Ragupathy, Carbon 119 (2017) 527–534. [214] Y.J. Noh, H.S. Kim, B. Ku, M. Khil, S.Y. Kim, Adv. Eng. Mater. 18 (2016)
[168] Y. Fu, Z. He, D. Mo, S. Lu, Int. J. Therm. Sci. 86 (2014) 276–283. 1127–1132.
[169] R.J. Warzoha, A.S. Fleischer, ACS Appl. Mater. Interfaces 6 (2014) 12868–12876. [215] Z. Han, A. Fina, Prog. Polym. Sci. 36 (2011) 914–944.
[170] H.S. Kim, H.S. Bae, J. Yu, S.Y. Kim, Sci. Rep. 6 (2016) 26825. [216] L.-W. Fan, X. Fang, X. Wang, Y. Zeng, Y. Xiao, Z.-T. Yu, X. Xu, Y. Hu, K. Cen, Appl.
[171] P. Goli, S. Legedza, A. Dhar, R. Salgado, J. Renteria, A.A. Balandin, J. Power Energy 110 (2013) 163–172.
Sources 248 (2014) 37–43. [217] X. Sun, A. Yu, P. Ramesh, E. Bekyarova, M.E. Itkis, R.C. Haddon, J. Electron.
[172] Y. Zhang, D. Han, Y. Zhao, S. Bai, Carbon 109 (2016) 552–557. Packag. 133 (2011) 020905.
[173] Y. Zhang, Y. Ren, S. Bai, Int. J. Heat Mass Transf. 118 (2018) 510–517. [218] Z. Sun, S. Pöller, X. Huang, D. Guschin, C. Taetz, P. Ebbinghaus, J. Masa, A. Erbe,
[174] H.M. Fang, S.L. Bai, C.P. Wong, Composites Part A-Applied Science and A. Kilzer, W. Schuhmann, M. Muhler, Carbon 64 (2013) 288–294.
Manufacturing 112 (2018) 216–238. [219] J. Yang, E. Zhang, X. Li, Y. Zhang, J. Qu, Z.-Z. Yu, Carbon 98 (2016) 50–57.

39
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

[220] J. Yu, H.K. Choi, H.S. Kim, S.Y. Kim, Compos. Part A Appl. Sci. Manuf. 88 (2016) Interfaces 10 (2018) 21628–21641.
79–85. [270] Y. Lin, J. Chen, P. Jiang, X. Huang, Chem. Eng. J. 389 (2020) 123467.
[221] M.E. Kompan, F.M. Kompan, P.V. Gladkikh, E.I. Terukov, V.G. Rupyshev, [271] X.H. Li, P.F. Liu, X.F. Li, F. An, P. Min, K.N. Liao, Z.Z. Yu, Carbon 140 (2018)
Y.V. Chetaev, Tech. Phys. 56 (2011) 1074–1078. 624–633.
[222] H. Zhang, Y. Lin, D. Zhang, W. Wang, Y. Xing, J. Lin, H. Hong, C. Li, Curr. Appl. [272] F. An, X.F. Li, P. Min, P.F. Liu, Z.G. Jiang, Z.Z. Yu, ACS Appl. Mater. Interfaces 10
Phys. 16 (2016) 1695–1702. (2018) 17383–17392.
[223] J. Shi, M. Ger, Y. Liu, Y. Fan, N. Wen, C. Lin, N. Pu, Carbon 51 (2013) 365–372. [273] J.J. Sun, D. Wang, Y.M. Yao, X.L. Zeng, G.R. Pan, Y. Huang, J.T. Hu, R. Sun,
[224] S.G. Prolongo, R. Moriche, A. Jiménez-Suárez, M. Sánchez, A. Ureña, J. Adhes. 90 J.B. Xu, C.P. Wong, High Volt. 2 (2017) 147–153.
(2014) 835–847. [274] W. Yang, W.J. Chen, Y. Zhou, C. Zhang, Z. Zhang, X. Chen, Y.S. Zhao, X.M. Bian,
[225] X. Fang, L. Fan, Q. Ding, X. Wang, X. Yao, J. Hou, Z.-T. Yu, G. Cheng, Y. Hu, High Voltage To be published (2020), https://doi.org/10.1049/hve.2019.0125.
K. Cen, Energy Fuels 27 (2013) 4041–4047. [275] R. Qian, J. Yu, C. Wu, X. Zhai, P. Jiang, RSC Adv. 3 (2013) 17373–17379.
[226] T.D. Hoang, P. Joonkyu, H. Sang, A. Muneer, S. Yongho, S. Koo, J. Korean Phys. [276] J. Li, X. Zhao, Z. Zhang, Y. Xian, Y. Lin, X. Ji, Y. Lu, L. Zhang, Compos. Sci.
Soc. 59 (2011) 2760–2764. Technol. 186 (2020) 107930.
[227] J. You, H.-H. Choi, J. Cho, J.G. Son, M. Park, S.-S. Lee, J.H. Park, Compos. Sci. [277] R.H. Sun, H. Yao, H.B. Zhang, Y. Li, Y.-W. Mai, Z.Z. Yu, Compos. Sci. Technol. 137
Technol. 160 (2018) 245–254. (2016) 16–23.
[228] J. Lv, X. Cai, Q. Ye, H. Zhang, Z. Ruan, J. Cai, Mater. Res. Express 5 (2018) [278] M.W. Akhtar, Y.S. Lee, D.J. Yoo, J.S. Kim, Compos. Part B Eng. 131 (2017)
055606. 184–195.
[229] K.A. Imran, J. Lou, K.N. Shivakumar, J. Appl. Polym. Sci. 135 (2018) 45833. [279] M.C. Hsiao, C.C. Ma, J.C. Chiang, K.K. Ho, T.Y. Chou, X. Xie, C.H. Tsai, L.H. Chang,
[230] X. Cai, Z. Jiang, X. Zhang, T. Gao, K. Yue, X. Zhang, RSC Adv. 8 (2018) C.K. Hsieh, Nanoscale 5 (2013) 5863–5871.
11367–11374. [280] Z. Liu, H. Zhang, S. Song, Y. Zhang, Compos. Sci. Technol. 150 (2017) 174–180.
[231] L. Tian, E. Jin, H. Mei, Q. Ke, Z. Li, H. Kui, J. Bionic Eng. 14 (2017) 130–140. [281] Y. Kim, M. Kim, H.G. Seong, J.Y. Jung, S.H. Baeck, S.E. Shim, Polymer 148 (2018)
[232] Y. Xiao, W. Wang, X. Chen, T. Lin, Y. Zhang, J. Yang, Y. Wang, Z. Zhou, Compos. 295–302.
Part A Appl. Sci. Manuf. 90 (2016) 614–625. [282] Y. Noma, Y. Saga, N. Une, Carbon 78 (2014) 204–211.
[233] G. He, Z. Yang, X. Zhou, J. Zhang, L. Pan, S. Liu, Compos. Sci. Technol. 131 (2016) [283] X. Zhang, J. Zhang, X. Zhang, C. Li, J. Wang, H. Li, L. Xia, H. Wu, S. Guo, Compos.
22–31. Sci. Technol. 150 (2017) 217–226.
[234] F. Kargar, Z. Barani, M. Balinskiy, A.S. Magana, J.S. Lewis, A.A. Balandin, Adv. [284] W. Yuan, Q. Xiao, L. Li, T. Xu, Appl. Therm. Eng. 106 (2016) 1067–1074.
Electron. Mater. 5 (2019) 1800558. [285] W. Li, W. Feng, H.Y. Huang, Journal of Materials Science-Materials in Electronics
[235] Z.H. Wu, C. Xu, C.Q. Ma, Z.B. Liu, H.M. Cheng, W.C. Ren, Adv. Mater. 31 (2019) 27 (2016) 6364–6370.
1900199. [286] M. Li, C. Tang, L. Zhang, B. Shang, S. Zheng, S. Qi, J. Mater. Sci. Mater. Electron.
[236] H. Oh, K. Kim, S. Ryu, J. Kim, Composites Part A-Applied Science and 29 (2017) 4948–4954.
Manufacturing 116 (2019) 206–215. [287] X. Zhang, X. Zhang, M. Yang, S. Yang, H. Wu, S. Guo, Y. Wang, Compos. Sci.
[237] M. Ajorloo, M. Fasihi, M. Ohshima, K. Taki, Mater. Des. 181 (2019) 108068. Technol. 136 (2016) 104–110.
[238] Y.H. Zhang, S.J. Park, Polymer 168 (2019) 53–60. [288] X. Huang, T. Iizuka, P. Jiang, Y. Ohki, T. Tanaka, J. Phys. Chem. C 116 (2012)
[239] X. Wei, F. Xue, X.D. Qi, J.H. Yang, Z.W. Zhou, Y.P. Yuan, Y. Wang, Appl. Energy 13629–13639.
236 (2019) 70–80. [289] V. Goyal, A.A. Balandin, Appl. Phys. Lett. 100 (2012) 073113.
[240] X.L. Zheng, B.Y. Wen, RSC Adv. 9 (2019) 36316–36323. [290] M.H. Tsai, I.H. Tseng, J.C. Chiang, J.J. Li, ACS Appl. Mater. Interfaces 6 (2014)
[241] Y.C. Liu, K. Wu, F.B. Luo, M.P. Lu, F. Xiao, X.X. Du, S.H. Zhang, L.Y. Liang, 8639–8645.
M.G. Lu, Composites Part A-Applied Science and Manufacturing 117 (2019) [291] H. Liem, H.S. Choy, Solid State Commun. 163 (2013) 41–45.
134–143. [292] Y. Zhang, Y.J. Heo, Y.R. Son, I. In, K.H. An, B.J. Kim, S.J. Park, Carbon 142 (2019)
[242] C.X. Shen, H. Wang, T.X. Zhang, Y. Zeng, J. Mater. Sci. Technol. 35 (2019) 36–43. 445–460.
[243] H.J. Ye, B. Han, H.Y. Chen, L.X. Xu, Nanotechnology 30 (2019) 355602. [293] M.A. Raza, A.V.K. Westwood, A.P. Brown, C. Stirling, J. Mater. Sci. Mater.
[244] M. Rafiee, F. Nitzsche, J. Laliberte, S. Hind, F. Robitaille, M.R. Labrosse, Electron. 23 (2012) 1855–1863.
Composites Part B-Engineering 164 (2019) 1–9. [294] D.D.L. Chung, J. Mater. Eng. Perform. 10 (2001) 56–59.
[245] A. Yu, P. Ramesh, X. Sun, E. Bekyarova, M.E. Itkis, R.C. Haddon, Adv. Mater. 20 [295] R. Prasher, Proceedings of the IEEE, (2006), pp. 1571–1586 94.
(2008) 4740–4744. [296] K.M.F. Shahil, A.A. Balandin, Solid State Commun. 152 (2012) 1331–1340.
[246] S. Yang, W. Lin, Y. Huang, H. Tien, J. Wang, C.M. Ma, S. Li, Y. Wang, Carbon 49 [297] F. Streb, M. Mengel, D. Schweitzer, C. Kasztelan, P. Schoderbock, G. Ruhl,
(2011) 793–803. T. Lampke, IEEE Trans. Compon. Packaging Manuf. Technol. 8 (2018) 1024–1031.
[247] M. Safdari, M.S. Al-Haik, Carbon 64 (2013) 111–121. [298] C. Li, L. Tan, X. Zeng, D. Zhu, R. Sun, J. Xu, C. Wong, Compos. Sci. Technol. 188
[248] E. Messina, N. Leone, A. Foti, G. Di Marco, C. Riccucci, G. Di Carlo, F. Di Maggio, (2020) 107970.
A. Cassata, L. Gargano, C. D’Andrea, B. Fazio, O.M. Marago, B. Robba, C. Vasi, [299] J. Renteria, S. Legedza, R. Salgado, M.P. Balandin, S. Ramirez, M. Saadah,
G.M. Ingo, P.G. Gucciardi, ACS Appl. Mater. Interfaces 8 (2016) 23244–23259. F. Kargar, A.A. Balandin, Mater. Des. 88 (2015) 214–221.
[249] K. Wu, C. Lei, R. Huang, W. Yang, S. Chai, C. Geng, F. Chen, Q. Fu, ACS Appl. [300] F. Zhang, Y.Y. Feng, M.M. Qin, L. Gao, Z.Y. Li, F.L. Zhao, Z.X. Zhang, F. Lv,
Mater. Interfaces 9 (2017) 7637–7647. W. Feng, Adv. Funct. Mater. 29 (2019) 1901383.
[250] Y.P. Chen, X. Hou, M.Z. Liao, W. Dai, Z.W. Wang, C. Yan, H. Li, C.T. Lin, N. Jiang, [301] M.A. Raza, A.V.K. Westwood, A.P. Brown, C. Stirling, Compos. Sci. Technol. 72
J.H. Yu, Chem. Eng. J. 381 (2020) 122690. (2012) 467–475.
[251] Z. Barani, A. Mohammadzadeh, A. Geremew, C.Y. Huang, D. Coleman, [302] M.A. Raza, A. Westwood, A. Brown, N. Hondow, C. Stirling, Carbon 49 (2011)
L. Mangolini, F. Kargar, A.A. Balandin, Adv. Funct. Mater. 30 (2020) 1904008. 4269–4279.
[252] X. Huang, C. Zhi, P. Jiang, J. Phys. Chem. C 116 (2012) 23812–23820. [303] H. Li, M. Jiang, Q. Li, D. Li, Z. Chen, W. Hu, J. Huang, X. Xu, L. Dong, H. Xie,
[253] L. Yu, J.S. Park, Y.S. Lim, C.S. Lee, K. Shin, H.J. Moon, C.M. Yang, Y.S. Lee, C. Xiong, Energy Convers. Manage. 75 (2013) 482–487.
J.H. Han, Nanotechnology 24 (2013) 155604. [304] A. Ijam, A. Moradi Golsheikh, R. Saidur, P. Ganesan, J. Mater. Sci. 49 (2014)
[254] I. Kholmanov, J. Kim, E. Ou, R.S. Ruoff, L. Shi, ACS Nano 9 (2015) 11699–11707. 5934–5944.
[255] X. Liang, F. Dai, ACS Appl. Mater. Interfaces 12 (2020) 3051–3058. [305] T.T. Baby, S. Ramaprabhu, J. Mater. Chem. 21 (2011) 9702–9709.
[256] D. An, S.S. Cheng, Z.Y. Zhang, C. Jiang, H.M. Fang, J.X. Li, Y.Q. Liu, C.P. Wong, [306] Z. Hajjar, Am. Rashidi, A. Ghozatloo, Int. Commun. Heat Mass Transf. 57 (2014)
Carbon 155 (2019) 258–267. 128–131.
[257] C. Zhang, R. Huang, Y. Wang, Z. Wu, H. Zhang, Y. Li, W. Wang, C. Huang, L. Li, [307] S.S. Jyothirmayee Aravind, S. Ramaprabhu, J. Appl. Phys. 110 (2011) 124326.
ACS Appl. Mater. Interfaces 12 (2020) 1436–1443. [308] C.M. Kim, Y.T. Kang, Energy 76 (2014) 468–476.
[258] Y. Yao, J. Sun, X. Zeng, R. Sun, J.B. Xu, C.P. Wong, Small 14 (2018) 1704044. [309] A. Ghozatloo, A. Rashidi, M. Shariaty-Niassar, Exp. Therm. Fluid Sci. 53 (2014)
[259] F. An, X. Li, P. Min, H. Li, Z. Dai, Z.-Z. Yu, Carbon 126 (2018) 119–127. 136–141.
[260] J. Li, X. Zhao, W. Wu, Z. Zhang, Y. Xian, Y. Lin, Y. Lu, L. Zhang, Carbon 162 (2020) [310] M. Kole, T.K. Dey, J. Appl. Phys. 113 (2013) 084307.
46–55. [311] T.D. Dao, H.I. Lee, H.M. Jeong, J. Colloid Interface Sci. 416 (2014) 38–43.
[261] L. Shao, L. Shi, X. Li, N. Song, P. Ding, Compos. Sci. Technol. 135 (2016) 83–91. [312] J. Liu, F. Wang, L. Zhang, X. Fang, Z. Zhang, Renew. Energy 63 (2014) 519–523.
[262] Y. Zhao, Y. Zhang, S. Bai, A. Composites Part, Applied Science and Manufacturing [313] M. Mehrali, E. Sadeghinezhad, S.T. Latibari, S.N. Kazi, M. Mehrali, M.N. Zubir,
85 (2016) 148–155. H.S. Metselaar, Nanoscale Res. Lett. 9 (2014) 15.
[263] K. Zhang, Y. Zhang, S. Wang, Carbon 65 (2013) 105–111. [314] S.D. Park, S. Won Lee, S. Kang, I.C. Bang, J.H. Kim, H.S. Shin, D.W. Lee, D. Won
[264] H. Im, J. Kim, Carbon 50 (2012) 5429–5440. Lee, Appl. Phys. Lett. 97 (2010) 023103.
[265] T. Pan, W. Kuo, N. Tai, Compos. Sci. Technol. 151 (2017) 44–51. [315] P.M. Sudeep, J. Taha-Tijerina, P.M. Ajayan, T.N. Narayanan, M.R. Anantharaman,
[266] L. Huang, P. Zhu, G. Li, D. Lu, R. Sun, C. Wong, J. Mater. Chem. A 2 (2014) RSC Adv. 4 (2014) 24887–24892.
18246–18255. [316] B. Wang, J. Hao, H. Li, Dalton Trans. 42 (2013) 5866–5873.
[267] J. Zhang, W. Lei, D. Liu, X. Wang, Compos. Sci. Technol. 151 (2017) 252–257. [317] W. Yu, H. Xie, X. Wang, X. Wang, Phys. Lett. A 375 (2011) 1323–1328.
[268] Y.Z. Feng, J. Hu, Y. Xue, C.G. He, X.P. Zhou, X.L. Xie, Y.S. Ye, Y.-W. Mai, J. Mater. [318] F. Wang, L. Han, Z. Zhang, X. Fang, J. Shi, W. Ma, Nanoscale Res. Lett. 7 (2012)
Chem. A 5 (2017) 13544–13556. 314.
[269] Y. Feng, X. Li, X. Zhao, Y. Ye, X. Zhou, H. Liu, C. Liu, X. Xie, ACS Appl. Mater. [319] M. Mehrali, S. Tahan Latibari, M. Mehrali, T.M.I. Mahlia, E. Sadeghinezhad,

40
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

H.S.C. Metselaar, Appl. Energy 135 (2014) 339–349. [368] A. Li, C. Zhang, Y.-F. Zhang, Compos. Part A Appl. Sci. Manuf. 103 (2017)
[320] H.Y. Wu, S. Deng, Y.W. Shao, J.H. Yang, X.D. Qi, Y. Wang, ACS Appl. Mater. 161–167.
Interfaces 11 (2019) 46851–46863. [369] Y. Jia, H. He, Y. Geng, B. Huang, X. Peng, Compos. Sci. Technol. 145 (2017)
[321] H.M. Weingrill, K. Resch-Fauster, T. Lucyshyn, C. Zauner, J. Appl. Polym. Sci. 137 55–61.
(2020) 48269. [370] Z. Liu, D. Shen, J. Yu, W. Dai, C. Li, S. Du, N. Jiang, H. Li, C. Lin, RSC Adv. 6
[322] Y. Zhong, M. Zhou, F. Huang, T. Lin, D. Wan, Sol. Energy Mater. Sol. Cells 113 (2016) 22364–22369.
(2013) 195–200. [371] X. Li, L. Shao, N. Song, L. Shi, P. Ding, A. Composites Part, Applied Science and
[323] Y. Zhang, X. Zheng, H. Wang, Q. Du, J. Mater. Chem. A 2 (2014) 5304–5314. Manufacturing 88 (2016) 305–314.
[324] J. Xu, T.S. Fisher, Int. J. Heat Mass Transf. 49 (2006) 1658–1666. [372] X.Y. Huang, High Volt. 2 (2017) 137–138.
[325] W. Wang, C. Wang, T. Wang, W. Li, L. Chen, R. Zou, J. Zheng, X. Li, Mater. Chem.
Phys. 147 (2014) 701–706.
[326] B. Li, T. Liu, L. Hu, Y. Wang, S. Nie, Chem. Eng. J. 215-216 (2013) 819–826. Xingyi Huang is a professor at the Research Center of
[327] H. Ke, Z. Pang, Y. Xu, X. Chen, J. Fu, Y. Cai, F. Huang, Q. Wei, J. Therm. Anal. Dielectrics and Electrical Insulation, Shanghai Jiao Tong
Calorim. 117 (2014) 109–122. University. Huang’s research focuses on polymers and their
[328] L.-W. Fan, Z.-Q. Zhu, Y. Zeng, Q. Lu, Z.-T. Yu, Int. J. Heat Mass Transf. 79 (2014) composites for dielectric, energy and thermal applications.
94–104. He received his Ph.D. from Shanghai Jiao Tong University
[329] R.J. Warzoha, A.S. Fleischer, Int. J. Heat Mass Transf. 79 (2014) 314–323. in 2008 and had postdoctoral experience in the same uni­
[330] G.-Q. Qi, C.-L. Liang, R.-Y. Bao, Z.-Y. Liu, W. Yang, B.-H. Xie, M.-B. Yang, Sol. versity. Currently, Huang serves as Associate Editor of two
Energy Mater. Sol. Cells 123 (2014) 171–177. journals (IEEE Transactions on Dielectrics and Electrical
[331] M. Mehrali, S.T. Latibari, M. Mehrali, T.M. Indra Mahlia, H.S. Cornelis Metselaar, Insulation and High voltage).
M.S. Naghavi, E. Sadeghinezhad, A.R. Akhiani, Appl. Therm. Eng. 61 (2013)
633–640.
[332] M. Mehrali, S.T. Latibari, M. Mehrali, H.S.C. Metselaar, M. Silakhori, Energy
Convers. Manage. 67 (2013) 275–282.
[333] J.F. Li, W. Lu, Y.B. Zeng, Z.P. Luo, Sol. Energy Mater. Sol. Cells 128 (2014) 48–51.
[334] J.M. Khodadadi, L. Fan, H. Babaei, Renewable Sustainable Energy Rev. 24 (2013)
418–444. Chunyi Zhi obtained his PhD degree in physics from the
[335] L. Fan, J.M. Khodadadi, Renewable Sustainable Energy Rev. 15 (2011) 24–46. Institute of Physics, Chinese Academy of Sciences. After
[336] M. Moeini Sedeh, J.M. Khodadadi, Carbon 60 (2013) 117–128. that, he started to work as a postdoctoral researcher in the
[337] W. Yang, L. Zhang, Y. Guo, Z. Jiang, F. He, C. Xie, J. Fan, J. Wu, K. Zhang, J. National Institute for Materials Science (NIMS) in Japan,
Mater. Sci. 53 (2017) 2566–2575. followed by a research fellowship in the International
[338] Y. Xia, W. Cui, H. Zhang, F. Xu, L. Sun, Y. Zou, H. Chu, E. Yan, J. Mater. Chem. A 5 Center for Young Scientists in NIMS and a permanent po­
(2017) 15191–15199. sition in NIMS as a senior researcher. He is currently pro­
[339] J. Yang, G. Qi, Y. Liu, R. Bao, Z. Liu, W. Yang, B. Xie, M. Yang, Carbon 100 (2016) fessor in the Department of Materials Science and
693–702. Engineering, City University of Hong Kong. Zhi’s research is
[340] G. Qi, J. Yang, R. Bao, D. Xia, M. Cao, W. Yang, M. Yang, D. Wei, Nano Res. 10 currently focused on flexible/wearable energy storage de­
(2016) 802–813. vices and sensors, etc.
[341] A. Hussain, I.H. Abidi, C.Y. Tso, K.C. Chan, Z. Luo, C.Y.H. Chao, Int. J. Therm. Sci.
124 (2018) 23–35.
[342] W. Cheng, W. Li, Y. Nian, W. Xia, Int. J. Heat Mass Transf. 116 (2018) 507–511.
[343] P.A. Advincula, A.C. de Leon, B.J. Rodier, J. Kwon, R.C. Advincula, E.B. Pentzer, J.
Mater. Chem. A 6 (2018) 2461–2467. Ying Lin obtained her BS (2013) and MS (2016) degrees
[344] Y.Q. Qian, N. Han, Z.X. Zhang, R.R. Cao, L.L. Tan, W. Li, X.X. Zhang, ACS Appl. from Anhui University (AHU) and University of Science and
Mater. Interfaces 11 (2019) 45832–45843. Technology of China (USTC), respectively. She is currently
[345] J. Yang, L. Tang, R. Bao, L. Bai, Z. Liu, B. Xie, M. Yang, W. Yang, Sol. Energy pursuing her PhD degree in Polymer Chemistry at Shanghai
Mater. Sol. Cells 174 (2018) 56–64. Jiao Tong University. Her current research interests focus
[346] J. Yang, L. Tang, R. Bao, L. Bai, Z. Liu, W. Yang, B. Xie, M. Yang, J. Mater. Chem. A on the structural design of 3D graphene-based aerogel and
4 (2016) 18841–18851. thermal management applications of its polymer compo­
[347] Z. Chen, J. Wang, F. Yu, Z. Zhang, X. Gao, J. Mater. Chem. A 3 (2015) sites.
11624–11630.
[348] M. Karthik, A. Faik, P. Blanco-Rodríguez, J. Rodríguez-Aseguinolaza,
B. D’Aguanno, Carbon 94 (2015) 266–276.
[349] P. Min, J. Liu, X. Li, F. An, P. Liu, Y. Shen, N. Koratkar, Z.-Z. Yu, Adv. Funct. Mater.
28 (2018) 1805365.
[350] Z.Y. Wang, Z. Tong, Q.X. Ye, H. Hu, X. Nie, C. Yan, W. Shang, C.Y. Song, J.B. Wu,
J. Wang, H. Bao, P. Tao, T. Deng, Nat. Commun. 8 (2017) 1478.
Hua Bao is an associate professor at the University of
[351] I. Gur, K. Sawyer, R. Prasher, Science 335 (2012) 1454–1455.
Michigan-Shanghai Jiao Tong University Joint Institute. He
[352] G.G.D. Han, H.S. Li, J.C. Grossman, Nat. Commun. 8 (2017) 1446.
received his B.S. in physics from Tsinghua University,
[353] P. Min, J. Liu, X.F. Li, F. An, P.F. Liu, Y.X. Shen, N. Koratkar, Z.Z. Yu, Adv. Funct.
China, in 2006, and his Ph.D. in Mechanical Engineering
Mater. 28 (2018) 1805365.
from Purdue University in 2012. His research interests are
[354] S. Wu, T.X. Li, Z. Tong, J.W. Chao, T.Y. Zhai, J.X. Xu, T.S. Yan, M.Q. Wu, Z.Y. Xu,
in micro/nano scale thermal energy transport, thermo­
H. Bao, T. Deng, R.Z. Wang, Adv. Mater. 31 (2019) 1905099.
physical properties, and applications on thermal manage­
[355] C.M. Costa, L.C. Rodrigues, V. Sencadas, M.M. Silva, J.G. Rocha, S. Lanceros-
ment and energy conversion.
Mendez, J. Memb. Sci. 407 (2012) 193–201.
[356] G.M. Zhou, S.F. Pei, L. Li, D.W. Wang, S.G. Wang, K. Huang, L.C. Yin, F. Li,
H.M. Cheng, Adv. Mater. 26 (2014) 625–631.
[357] J.D. Zhu, C. Chen, Y. Lu, J. Zang, M.J. Jiang, D. Kim, X.W. Zhang, Carbon 101
(2016) 272–280.
[358] W.S. Khan, R. Asmatulu, V. Rodriguez, M. Ceylan, Int. J. Energy Res. 38 (2014)
2044–2051.
[359] P. Zhu, J.D. Zhu, J. Zang, C. Chen, Y. Lu, M.J. Jiang, C.Y. Yan, M. Dirican, Guangning Wu PhD, born in 1969, is Distinguished
R.K. Selvan, X.W. Zhang, J. Mater. Chem. A 5 (2017) 15096–15104. Professor of Southwest Jiaotong University, IEEE Fellow,
[360] X. Liu, K.D. Song, C. Lu, Y.T. Huang, X.L. Duan, S. Li, Y.H. Ding, J. Memb. Sci. 555 IET Fellow, and a regular member of CIGRE SC B2. Prof. Wu
(2018) 1–6. has been focusing on research about high voltage traction
[361] H.S. Hamut, I. Dincer, G.F. Naterer, Int. J. Energy Res. 37 (2013) 1–12. power supply system. He is the author (or co-author) of
[362] G. Karimi, X. Li, Int. J. Energy Res. 37 (2013) 13–24. over 10 IEC/IEEE standards, more than 10 monographs and
[363] Y.B. Zhang, L.X. Miao, J. Ning, Z.C. Xiao, L. Hao, B. Wang, L.J. Zhi, 2d Mater. 2 exceeding 200 academic papers. These research achieve­
(2015) 024013. ments are of great significance to the development of high-
[364] C.F. Li, S.H. Liu, C.G. Shi, G.H. Liang, Z.T. Lu, R.W. Fu, D.C. Wu, Nat. Commun. 10 speed electrified railways.
(2019) 1363.
[365] P. Zhu, J. Zang, J.D. Zhu, Y. Lu, C. Chen, M.J. Jiang, C.Y. Yan, M. Dirican,
R.K. Selvan, D. Kim, X.W. Zhang, Carbon 126 (2018) 594–600.
[366] Z. Liu, Y. Chen, W. Dai, Y. Wu, M. Wang, X. Hou, H. Li, N. Jiang, C. Lin, J. Yu, RSC
Adv. 8 (2018) 1065–1070.
[367] L. Chen, X. Hou, N. Song, L. Shi, P. Ding, A. Composites Part, Applied Science and
Manufacturing 107 (2018) 189–196.

41
X. Huang, et al. Materials Science & Engineering R 142 (2020) 100577

Pingkai Jiang is a professor of Department of Polymer Yiu-Wing Mai is University Professor in Mechanical
Science and Engineering and the Director of Research Engineering at the University of Sydney. He obtained his
Center of Dielectrics and Electrical Insulation at Shanghai PhD on fracture mechanics from the University of Hong
Jiao Tong University. Jiang’s research focuses on polymer Kong. He previously worked in the US (University of
nanocomposites for energy storage and dielectric applica­ Michigan and NIST), the UK (Imperial College) and Hong
tion and fire retardant materials for electrical insulation. He Kong (HKUST, CityU, HKU and PolyU). His current research
received his Ph.D. in Electrical Materials and Insulation interest is on polymer nanocomposites and their functional
Technology in 1995 from Xi’an Jiao Tong University. applications. He is a Fellow of the Royal Society of London
and the Australian Academy of Science as well as a Foreign
Member of the Chinese Academy of Engineering.

42

You might also like