thermal analysis of epoxy-based nanocomposites: have solvent effects been overlooked?

9
Thermal analysis of epoxy-based nanocomposites: Have solvent effects been overlooked? Geoff Rivers Allan Rogalsky Pearl Lee-Sullivan Boxin Zhao Received: 11 November 2013 / Accepted: 15 December 2013 Ó Akade ´miai Kiado ´, Budapest, Hungary 2014 Abstract Analysis of the available published data calls into question some of the reported effects of silver, graphene, and boron nitride nanoparticles on the cure behavior and glass transition temperature (T g ) of epoxy- based conductive nanocomposites. The usefulness of most studies is limited, as the final T g value is only reported for a single composition. More useful studies provide a com- parison of T g across compositions and/or to the T g of the neat matrix material. However, the main focus is on nanoparticle effect, while the presence of residual solvent is generally overlooked. Although the data are too sparse for firm conclusions, reductions in T g seem to correlate with a known residual solvent or mild degas conditions (low temperature, unagitated), while increases are more likely when no solvent is used or after rigorous degassing. Using solvent control groups to differentiate solvent effects from filler effects, we conducted our own differential scanning calorimetry experiments. It was found that silver microparticles have no statistically significant effect on T g , but appeared to when solvent content was not accounted for. The erroneous apparent effect is very similar to reported effects for nanoparticles in the literature. If a solvent must be used, we recommend that the residual solvent content be quantified and provide the T g of a control sample with a representative solvent content with- out nanoparticles. Analysis of the present literature also suggests that (i) nanoparticle surface chemistry has a sig- nificant effect and (ii) poor dispersion/aggregation results in a reduction in T g . Keywords Epoxy Á Nanocomposite Á Solvent Á Glass transition Á Cure Introduction Silver, graphene, and hexagonal boron nitride (hBN) nanoparticles are three relatively new filler materials being explored for use in the development of polymer-based conductive nanocomposites. Even at low filler contents, the addition of nanoparticles to polymeric matrices is known to have large impact on mechanical properties [13]. If the particles have high aspect ratios and intrinsic conductivi- ties, there is the potential to achieve low bulk resistivity at low filler volume fractions [4, 5]. Consequently, this is a new class of adhesives being considered as potential replacements for lead-based solders, which are being phased out in the electronics industry due to toxicity [6]. Significant development studies, however, are still required before they can be applied. Notably, the high specific surface areas of nanoparticles, up to three orders of mag- nitude larger than that of microparticles, provide high surface energies and large filler–matrix interaction sur- faces, which magnify the particle influence on the polymer matrix [7]. For practical application in industry, we need to understand how this impacts the chemistry-driven solidi- fication ‘‘cure’’ process of these nanocomposites. If we consider the mechanisms that influence the cure behavior and final property of epoxy-based nanocompos- ites, there appear to be two bodies of work which are related but not yet fully connected: the established cure G. Rivers (&) Á A. Rogalsky Á P. Lee-Sullivan Department of Mechanical and Mechatronics Engineering, University of Waterloo, 200 University Avenue, Waterloo, ON N2L 3G1, Canada e-mail: [email protected] B. Zhao Department of Chemical Engineering, University of Waterloo, 200 University Avenue, Waterloo, ON N2L 3G1, Canada 123 J Therm Anal Calorim DOI 10.1007/s10973-013-3613-2

Upload: boxin

Post on 23-Dec-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

Thermal analysis of epoxy-based nanocomposites: Have solventeffects been overlooked?

Geoff Rivers • Allan Rogalsky •

Pearl Lee-Sullivan • Boxin Zhao

Received: 11 November 2013 / Accepted: 15 December 2013

� Akademiai Kiado, Budapest, Hungary 2014

Abstract Analysis of the available published data calls

into question some of the reported effects of silver,

graphene, and boron nitride nanoparticles on the cure

behavior and glass transition temperature (Tg) of epoxy-

based conductive nanocomposites. The usefulness of most

studies is limited, as the final Tg value is only reported for a

single composition. More useful studies provide a com-

parison of Tg across compositions and/or to the Tg of the

neat matrix material. However, the main focus is on

nanoparticle effect, while the presence of residual solvent

is generally overlooked. Although the data are too sparse

for firm conclusions, reductions in Tg seem to correlate

with a known residual solvent or mild degas conditions

(low temperature, unagitated), while increases are more

likely when no solvent is used or after rigorous degassing.

Using solvent control groups to differentiate solvent effects

from filler effects, we conducted our own differential

scanning calorimetry experiments. It was found that silver

microparticles have no statistically significant effect on Tg,

but appeared to when solvent content was not accounted

for. The erroneous apparent effect is very similar to

reported effects for nanoparticles in the literature. If a

solvent must be used, we recommend that the residual

solvent content be quantified and provide the Tg of a

control sample with a representative solvent content with-

out nanoparticles. Analysis of the present literature also

suggests that (i) nanoparticle surface chemistry has a sig-

nificant effect and (ii) poor dispersion/aggregation results

in a reduction in Tg.

Keywords Epoxy � Nanocomposite � Solvent � Glass

transition � Cure

Introduction

Silver, graphene, and hexagonal boron nitride (hBN)

nanoparticles are three relatively new filler materials being

explored for use in the development of polymer-based

conductive nanocomposites. Even at low filler contents, the

addition of nanoparticles to polymeric matrices is known to

have large impact on mechanical properties [1–3]. If the

particles have high aspect ratios and intrinsic conductivi-

ties, there is the potential to achieve low bulk resistivity at

low filler volume fractions [4, 5]. Consequently, this is a

new class of adhesives being considered as potential

replacements for lead-based solders, which are being

phased out in the electronics industry due to toxicity [6].

Significant development studies, however, are still required

before they can be applied. Notably, the high specific

surface areas of nanoparticles, up to three orders of mag-

nitude larger than that of microparticles, provide high

surface energies and large filler–matrix interaction sur-

faces, which magnify the particle influence on the polymer

matrix [7]. For practical application in industry, we need to

understand how this impacts the chemistry-driven solidi-

fication ‘‘cure’’ process of these nanocomposites.

If we consider the mechanisms that influence the cure

behavior and final property of epoxy-based nanocompos-

ites, there appear to be two bodies of work which are

related but not yet fully connected: the established cure

G. Rivers (&) � A. Rogalsky � P. Lee-Sullivan

Department of Mechanical and Mechatronics Engineering,

University of Waterloo, 200 University Avenue, Waterloo,

ON N2L 3G1, Canada

e-mail: [email protected]

B. Zhao

Department of Chemical Engineering, University of Waterloo,

200 University Avenue, Waterloo, ON N2L 3G1, Canada

123

J Therm Anal Calorim

DOI 10.1007/s10973-013-3613-2

research and newer nanocomposites’ research. Cure

research has focused on the polymer-originating influences,

while nanocomposites’ research is interested in the effect

and mechanism caused by the presence of nanoparticles in

these novel composites. The process of filling the polymer

resin typically involves multiple steps to disperse the

nanoparticles. Because the focus is mainly placed on the

filler material, the conclusions being drawn in studies fre-

quently neglect influences caused by factors that are

independent of the nanoparticles, such as dispersion

methods. As a result, there is strong likelihood for misin-

terpretation of the mechanisms influencing cure data.

Properties of nanomaterials

Nanoparticles composed of silver, graphene, and boron

nitride (BN) are available in high aspect ratio forms, making

them ideal for the formation of conductive nanoparticle

networks at low filler fractions [8–11]. Each has either high

electrical conductivity (Silver [bulk] 6.3 9 109 S cm-1

[12], graphene 9.6 9 105 S cm-1 [13]) or thermal conduc-

tivity (Silver [bulk] 425 W mk-1 [14], graphene

600–5,000 W mk-1 [13, 15], BN 400 W mk-1 [16]).

Nanoscale particles of metallic silver are known to have

very high surface energy, which can result in detrimental

aggregation and thermal instability of silver nanoparticles

[8, 17, 18]. As a result, much of the nanosilver research is

directed toward stabilizing agents. Without chemical

modification of its surface, silver is not expected to chem-

ically interact with the cross-linking reaction of an epoxy.

Graphene is a two-dimensional single layer of hexagonally

bonded sp2 carbon, which forms natural graphite when

stacked. The continuous delocalized p-bond network on the

surfaces of non-functionalized graphene sheets is not only

highly conductive but also strongly drives aggregation

through p–p stacking [5, 9, 15, 19]. hBN is composed of

two-dimensional, hexagonally bonded alternating boron

and nitrogen [20, 21]. These hBN nanosheets, or nanotubes,

have a localized semi-ionic p-bond network on their sur-

faces, driving aggregation of the nanoparticles though

mechanisms similar to p–p stacking [20, 21].

Each of these nanomaterials is several orders of mag-

nitude more rigid than cured epoxy, and can be surface

modified through purposeful functionalization, capping

agents, or surfactants to change its interaction with the

epoxy matrix. For example, graphene sheets can be oxi-

dized into graphene oxide or reduced graphene sheets,

producing various degrees of surface decoration by car-

bonyl, epoxide, hydroxyl, and other oxygen-derived func-

tional groups, which disrupt the p-bond network and

reduce conductivity in exchange for increased dispersion in

epoxy [5, 9, 15, 19]. Modifications are typically made with

the intent of reducing surface energy to increase particle

stability, to reduce aggregation, or to allow an epoxy

matrix to bond to the nanoparticle surface [5, 8, 9, 15, 17,

19, 22–24].

Properties of epoxy resin

Epoxies are thermosetting resins containing an epoxide

group (Fig. 1a). To cure the resin, a hardener is added,

which initiates polymerization and forms cross-links via

reaction with the epoxide group. Backbone chemistry (R

group in Fig. 1a) and choice of hardener set the expected

range of matrix cure temperatures, cure rates, and service

temperatures [25]. This makes it difficult to directly com-

pare cure data between epoxy systems. Among the many

hardeners used, amines and carboxylic acids are the most

relevant to the discussion of nanoparticle interaction

(Fig. 1b) [25, 26]. Many hardener reactions produce –OH

residues, which form further cross-links via an autocata-

lytic reaction (Fig. 1c) [25]. Fillers, process agents, or

contaminants bearing –OH groups can also react via the

same mechanism.

Predicting cure behavior of nanocomposites

The study of epoxy cure is the study of network formation

via a cross-linking reaction. From established cure

research, there are four potential influences on the cure

behavior in epoxy-based nanocomposites systems: (1) resin

O

C CHH2

H2N

R O

C

HO

R

RR1

R1

R2

R2R3

NH2

H2

H2H2 HHH

N C C

C

OH

C

O

H

ii)

i)

C C

O

(a) (b) (c)Fig. 1 Illustrations of a an

epoxide group, b amine and

carboxylic groups, c two

common epoxy reactions,

i primary amine forward

reaction, ii autocatalytic

hydroxy reaction. Note: Fig 2a

caption on next page with

Fig 2b

G. Rivers et al.

123

to hardener ratio; (2) system degree of cure; (3) the pre-

sence of residual solvents; and (4) interaction between the

matrix and nanoparticle surfaces. The relative magnitude

of these influences can be characterized using calorimetric

or spectroscopic methods to directly follow the cross-

linking reaction; however, indirect estimation via glass

transition measurement is more common.

The glass transition is important for the analysis of

epoxies because it is relatively simple to measure and is

influenced by a wide range of factors [27]. It is a thermal

transition occurring over a temperature range, in which the

epoxy mechanical response changes from stiff elastic

behavior to rubbery viscoelastic behavior as temperature

increases [27–29]. For analysis, the temperature range of a

glass transition is simplified to a representative ‘‘glass

transition temperature,’’ Tg. The glass transition is a

cooperative process related to the mobility available to

polymer chains on the molecular scale: anything that

influences molecular mobility can potentially impact Tg

[27–29]. It is well known that the higher cross-linking

densities result in reduced molecular mobility resulting in

the higher glass transition temperatures [27].

Resin to hardener ratio affects cure rate and maximum

obtainable cross-link density [27, 30]. Deviating from the

optimum ratio results in fewer cross-links and reduction of

‘‘final Tg’’ (Tg?). Overall reaction rates are generally

reduced by non-optimum reactant ratios, though excess

‘‘hardener-like groups’’ may increase initial reaction rates.

Degree of cure (a) approximates the fractional conver-

sion of the limiting reactant for a given resin hardener ratio

[25]. In neat epoxy systems, ‘‘instantaneous Tg’’ (Tga)

increases with a. Tga changes most rapidly as it approaches

Tg? at cure completion [31]. This means that when using

Tg as a probe for nanofiller effects, a is the most significant

as a source of error when cure is nearly complete and the

assumption of a similar degree of cure between composi-

tions is violated. Several calorimetric techniques exist to

verify that complete cure has been achieved, the simplest

of which is to verify that no further reaction occurs at

temperatures above the measured Tg [32]. Under isother-

mal cure conditions, cure rate data are unreliable as an

indicator of cure completion because cure rate decreases

asymptotically as Tga approaches the cure temperature.

Solvents present during cure separate reactive groups

reducing reaction rates and reducing Tg by forming a

looser, less entangled network [33–36]. This effect has

been observed for the very low solvent contents present

after vacuum degassing to equilibrium. Permanent changes

exist even if the solvent is driven off during early stages of

cure [34].

Surface effects have been shown to affect the Tg of

adjacent polymer [27]. For surfaces stiffer than the poly-

mer, molecular mobility decreases and Tg increases, while

flexible interfaces produce the opposite effect. Reduced

molecular mobility during cure is expected to result in

reduced overall reaction rates. Using ‘‘the rule of mixtures’’

and an estimate of interaction length scale from work

reported in [29], it is very likely that the entire matrix will

be affected by well-dispersed nanoparticles. Accordingly,

our basic understanding of cure would lead us to predict

large filler impact on nanocomposite cure behavior.

Perspectives from nanocomposite literature

In the vast majority of the nanocomposite studies, the

effects of silver, graphene, or BN nanoparticles on cure are

almost exclusively attributed to the filler material, despite a

lack of controls to delineate processing effects. To date,

changes in nanocomposite Tg or cure behavior are reported

to be due to one of six filler-centric mechanisms:

(1) restriction of molecular motion via non-covalent sur-

face interactions [20, 37–46]; (2) plasticization by soft

particle coating [47–49]; (3) cross-linking with the nano-

particles [20, 39, 48–54]; (4) chemical ratio interference

[55, 56]; (5) aggregation [50, 57–59]; and (6) non-specific

decreases in cross-linking [47–49]. If these mechanisms are

explained in terms of our basic understanding of cure,

Table 1, we find that a majority of conclusions drawn fall

under ‘‘molecular motion increases/decreases.’’ The cate-

gory listed as ‘‘non-specific decrease in crosslinking’’ is not

included in Table 1. It was used as a general explanation

where Tg decreases were attributed to the nanoparticle

without reference to a specific mechanism, and was only

seen in solvent-blended studies. In general, there is almost

no attempt to determine if the dispersion methods used in

processing had an effect. In this paper, we explore possible

experimental errors caused either by the presence of

residual solvent or incomplete cure.

Table 1 Attributed nanocomposite effect mechanisms grouped

according to influences described by established cure research

Known effects from

established cure research

Attributed effects in nanocomposites

literature

Effective resin to hardener

ratio

Chemical ratio interference from

reactive nanoparticle surfaces

The presence of residual

solvents

Mostly neglected

System degree of cure

Matrix molecular motion

increase/decrease

Non-covalent surface interactions with

nanoparticle.

Cross-linking to nanoparticles.

Plasticization by soft particle coating

nanoparticle aggregation.

Thermal analysis of epoxy-based nanocomposites

123

Solvent effect in the nanocomposite literature

To explore the hypothesis that solvent effects may be

incorrectly attributed to the nanoparticle addition in the

nanocomposite literature, we compared Tg data from

mechanically mixed studies (no possible solvent effect)

[21, 41–43, 48, 49, 58] with solvent-blended studies (sol-

vent effects possible) [20, 45–47, 52, 54–56, 59–64],

(Fig. 2). As can be seen, there is a large difference in both

the number and severity of negative Tg trends between the

two groups; 7 out of 14 solvent-blended studies reported

Tg? reductions with a maximum drop of 50 �C [47, 52, 55,

59–62] compared to 2 out of 8 in the mechanically mixed

group with no more than a 15 �C decrease [48, 58].

Comparing the sample preparation procedures of the

solvent-blended studies with Tg decreases [47, 55, 60, 61]

to those with clear increases [20, 45, 46, 54, 63, 64], we

find Tg decreases are generally associated with static degas

at lower temperatures. At the same time, Tg increases are

associated with agitation at higher temperatures, as both

groups applied vacuum in their work. The correlation

between degassing procedure and outcome can be taken as

further evidence for a solvent effect, particularly as one

author in the decreased Tg group noted that significant

solvent remained after degassing [47].

The absence of solvent controls in any of the reviewed

papers does not allow us to delineate solvent from other

effects. It is fair to infer that solvent-blended studies that

correlate their Tg decreases with aggregation [55, 59, 62,

64] or chemical ratio interference [39, 55, 56] may be

incorrectly attributing the cause due to overlooked solvent.

Even those studies with positive trends may have the

magnitude of the nanoparticle effect masked by the pre-

sence of residual solvent.

40

30

20

10

0

–10

–20

–30

–40

–50

40

30

20

10

0

–10

–20

–30

–40

–50

Nor

mal

ized

Tg∞

/°C

Nor

mal

ized

Tg∞

/°C

0 0.01 0.02 0.03 0.04 0.05 0.06

0 0.01 0.02 0.03 0.04 0.05 0.06

Filler volume fraction

Filler volume fraction

Ag – Liang et al, poly. Sci., 2011 (Untreated) [47]Ag – Chan et al, Appl poly Sci 2011 [43]Gr – Martin–Gallego et al, Polymer 2011 [42]Gr – Hu et al., Comp Sci and Tech 2010 [48]

Ag – Liang, poly. Sci., 2011 (Flexible surface) [47]Ag – Suriati et al., J Mater Sci: Mater Electron 2011 [57]Gr – S. Ganguli et al. Carbon 2008 [40]BN – Voo et al., J Plastic Film Sheeting 2011 [41]

BNNTs – C Zhi et al. Pure Appl. Chem 2010 [39]

Ag – Ren et al, Matls Chem and Phys 2012 [63]Gr – Corcione et al, poly Eng Sci 2012 [61]Gr – Rafiee et al, Small 2010 [44]Gr – Zamam et al, Adv Funct Mater 2012 [62]

Gr – Zaman et al, Polymer 2011 [45]Gr – Y Guo, Ind. Eng. Chem. Res. 2011 [46]BN – J. Yu et al., Polymer 2012 [20]

Gr – C Bao et al., J.Mater. Chem. 2011 [60]Gr – Kim et al, J. Mater. Chem. 2011 [59]Gr – Monti et al., Composites: Part A 2013 [58]Gr – Bortz et al., Macromolecules 2012 [51]Gr – Liu et al., Poly–Plast Tech Eng 2012 [54]Gr – Shen et al., Polymer 2013 [55]BNNT – Huang et al., Adv Funct Mater 2012 [53]

(a)

(b)

Fig. 2 Final Tg data from the

open literature, normalized to

provided final Tg of neat epoxy,

groups by a studies that did not

use solvent (mechanically

mixed) and b solvent-blended

studies

G. Rivers et al.

123

Incomplete cure

In the solvent-blended group, several authors [20, 47, 56, 59–62]

report Tg values near or above their maximum cure temperature

with only two of these reports verifying that cross-linking was

not prematurely terminated by vitrification [59, 62]. It was also

noted that decreased reaction rates are correlated with increased

filler loadings for both silver [18, 37] and graphene [39, 43, 50].

Therefore, the addition of nanofillers may lead to incomplete

cure that varies with nanoparticle content. Under these cir-

cumstances, a nanoparticle that is expected to increase final Tg

by restricting molecular motion within the matrix will actually

lower Tg, due to the sensitivity of instantaneous Tg to the degree

of cure [31].

Discernible nanoparticle effects

Despite the potential interferences from solvent and

incomplete cure, conclusions can be drawn regarding the

existence of nanoparticle effects caused by aggregation, non-

stoichiometric chemistry, and nanoparticle surface interac-

tions. In general, these conclusions are consistent with our

basic understanding of the established cure literature.

Aggregation of nanofillers has been correlated with a

reduction in Tg? [41, 55, 58, 59, 62, 64] and no significant

cure rate change [57]. Although in many cases the exact

magnitude of the effect may be obscured by solvent, the

balance of evidence indicates that aggregation does cause

reductions in Tg. The reduction has been shown to occur in

the absence of solvent [58] and as a change in Tg trend

associated with the onset of aggregation at high filler

loadings [64]. We speculate that the mechanism responsi-

ble for reducing Tg is related to the aggregates acting as a

flexible second phase. Similarly, the lack of cure rate

impact might be due to a significant fraction of the nano-

particle surface becoming occluded within the aggregates,

including reactive surface groups.

Non-stoichiometric chemistry is known to lead to reduc-

tions in Tg? [27, 30]. Unfortunately, all reports relevant to this

review that attributed this effect to the nanoparticles occur

within the solvent-blended group [55, 56, 60]. In two of these

cases, we have identified non-stoichiometric chemistry due to

solvent as a likely cause of the observed effect. An alcohol was

used, and a significant reaction between the solvent and the

matrix is expected based on either initial solvent concentration

[60] or conditions of the degas cycle [56].

For well-dispersed nanoparticles with epoxy-reactive

surface chemistry, where the reactants added by the fillers

did not significantly impact the epoxy stoichiometry, both

final Tg and cure rate were reported to increase. This is

applicable to nanoparticles with short functional groups

that coat the surface plentifully [9, 15, 19, 24, 38, 41, 46,

49, 50, 52, 53], or nanoparticles sparsely coated with

longer polymer chains capable of reacting with the epoxy

matrix. This was associated with an increase in cure

enthalpy when normalized to the mass of resin and hard-

ener [20, 39, 51, 54]. The acceleration of cure rate was

attributed to the increased concentration of reactive groups,

dependent on the type of reactive group decorating the

nanoparticle [25, 26, 38, 39, 50–53]. In cases where the

altered stoichiometry did not degrade the epoxy cross-link

density appreciably, Tg? was increased by cross-linking

the matrix to the rigid nanoparticles, which reduced

molecular mobility only after the particle–matrix cross-

links were formed [20, 38, 41, 46, 49, 52, 54, 56].

Nanoparticles sparsely coated in non-reactive polymers

that weakly interacted with the epoxy matrix through

hydrogen bonding or Van der Waals force increased Tg? of

the final nanocomposites [63], while dense coatings pro-

duce a polymer brush effect, introducing a flexible parti-

cle–matrix interface and lowering Tg? [48].

Well-dispersed nanoparticles with bare non-reactive

surfaces were reported to produce an increase in Tg?, and a

reduction in cure rates. Both of these effects were attrib-

uted to steric hindrance of the epoxy chains by physical

interactions with the nanoparticle surfaces during curing

and in fully cured epoxy, by either high surface energy and

a reduction in void free-volume [37], or p–p stacking

interactions between nanoparticle p-bond networks and

epoxy resin aromatic groups [21, 45, 46, 49, 59].

A proposed methodology for accurate data

interpretation

We hereby propose a simple methodology and demonstrate

how it can be used to avoid potential misinterpretation of

80 90 100 110 120 130 140 150

Temperature/°C

Rev

C p/

J g–1

K–1

Arb

itrar

y of

fset

113.67 °C

113.56 °C

122.22 °C

1.2% IPA 60mass% silver1.1% IPA 0mass% silverNeat epoxy control

Fig. 3 Typical heat capacity results from modulated DSC scans of

fully cured composites and control group samples, demonstrating Tg?

Thermal analysis of epoxy-based nanocomposites

123

thermal analysis results. In this approach, a set of solvent

controls representative of the residual solvent content of

the composite is used. The method has focused on char-

acterizing Tg because it is the most frequently reported

property in the literature for epoxy nanocomposites con-

taining silver, graphene, or BN. Although Tg character-

ization by DMA is valid, we have characterized Tg by

DSC, because this affords the opportunity to confirm

complete cure within the same experiment.

Materials and methods

The composite system consisted of DEGBA epoxy resin

(DER331 Dow Chemical) filled with silver microflake

(Sigma Aldrich 327077). They were solvent blended using

either acetone (Sigma Aldrich 270725) or isopropanol

(IPA, Sigma Aldrich 34863) and hardened by triethylene-

tetramine (TETA, DEH 24 Dow Chemical). A limited

set of composite compositions ranging between 0 and

140

130

120

110

100

90

80T

g ∞/°

C

140

130

120

110

100

90

80

Tg ∞

/°C

140

130

120

110

100

90

80

Tg ∞

/°C

140

130

120

110

100

90

80

Tg ∞

/°C

Neatepoxycontrol

1.2 mass%IPA

60 mass%Ag

Neatepoxycontrol

1.0 mass%Acetone

60 mass%Ag

1.3 mass%Acetone

70 mass%Ag

Neatepoxycontrol

1.1 mass%IPA

0 mass%Ag

1.2 mass%IPA

60 mass%Ag

Neatepoxycontrol

0.8 mass%Acetone0 mass%

Ag

1.0 mass%Acetone

60 mass%Ag

1.3 mass%Acetone

70 mass%Ag

2.5 mass%Acetone0 mass%

Ag

(a) (b)

(c) (d)

Fig. 4 a, b Tg? displayed with

neat epoxy, but not solvent

controls. c, d Tg? with solvent

controls, capturing solvent

effect. The filler used is silver

microflake. Note: The data in

(a) and (b) for silver-filled

samples are replotted in (c) and

(d). Solvent mass% neglects the

filler mass

550

500

450

400

350

3000 0.5 1 1.5 2 2.5 3

Solvent Concentration in matrix/mass%

Nor

mal

ized

ent

halp

y/J

g–1

Neat epoxy (collected)

Acetone diluted, silver filled

Acetone diluted, 0mass% silver

IPA diluted, 0 mass% silver

IPA diluted, silver filled

0mass%Ag

0mass%Ag

0mass%Ag

0mass%Ag

60mass%Ag

60mass%Ag

70mass%Ag

Fig. 5 Normalized enthalpy of

reaction; average of three

replicates excepting neat epoxy

which is an average of 6

replicates. Error bars display

standard deviation. Note:

Solvent mass% neglects the

filler mass

G. Rivers et al.

123

70 mass% silver, with additions of up to 1.2 mass% IPA

and 2.5 mass% acetone, was studied. Note that solvent

mass% is presented as the matrix mass%, which neglects

the filler mass. The solvent content of the silver-filled

composites is the equilibrium content after the degassing to

constant mass. For the IPA-blended composite, a solvent

control group was diluted to match as closely as possible

the calculated residual solvent contents. For the acetone-

blended composites, two solvent control groups without

silver were produced, bracketing the composite residual

solvent contents.

The epoxy resin was diluted with 10 mass% solvent.

Weighed quantities of silver filler were incorporated by

vortex mixing for 5 min (Corning LSE) and dispersed

using an ultrasonic mixer for 1 h (Fisher-Scientific FS20).

Samples were degassed under *28 in Hg vacuum with

agitation at room temperature. Samples were periodically

weighed, and degassing continuing until the mass stopped

changing. After degassing, TETA hardener was added at a

stoichiometric 13:100 hardener-to-resin ratio by mass, and

the material was vortex mixed an additional 5 min.

Differential scanning calorimetry

DSC samples were encapsulated in hermetically sealed

aluminum pans and cured during a 3 �C/min ramp to

175 �C in a TA4600 differential scanning calorimeter. The

area under the exothermal peak was taken as the reaction

enthalpy. Samples were cooled and a second heating scan

performed to 200 �C, using an underlying rate of 3 �C/min

with modulation of ±0.477 �C every 60 s. No residual cure

signal was observed, confirming that the samples were

fully cured during the first scan. Tg? was determined from

the reversing heat capacity (Rev Cp) signal of the second

scan, using the methodology described by ASTM D7426-

08. Typical second scan Cp data are shown in Fig. 3.

Results and discussion

The final Tg results for the tested composites are shown in

Fig. 4. When compared with neat epoxy, but not solvent

control groups, Fig. 4a, b, it would appear that the addition

of silver microflakes has a strong negative impact on Tg?.

As bare silver microflakes are chemically inert toward

epoxy, the only potential filler influence on Tg would be

due to the physical interaction. However, physical inter-

action is expected to be negligible because the microflakes

have a low specific surface area (1.16 m2/g, Sigma

Aldrich). At ideal dispersion, we calculate that for our

compositions (60–70 mass% silver) a maximum of 5 % of

the matrix volume is within the *15 nm [29] physical

interaction region adjacent to the particle surfaces. There-

fore, the silver microflakes used are not expected to have a

measurable effect on the Tg. When composite Tg? results

are presented alongside those of solvent control groups

without silver, Fig. 4c, d, the correct interpretation

becomes apparent: The reduction in Tg? is almost wholly

due to the presence of solvent.

The enthalpy results for the tested composites are shown

in Fig. 5. These results have been normalized using the

epoxy system mass (resin ? hardener). From the normal-

ized enthalpy data, it is clear that the silver fill has no sta-

tistical effect on the cross-linking density, and that if any

effect exists, it is primarily due to the solvent. Although

there is high scatter for the 1.3 % acetone, 70 mass% silver

composite sample, the trends suggest that solvent contents

beyond 1.3 % can result in decreased cross-linking density.

This is consistent with the previous results reported by Loos

et al. [35]. At low solvent contents, there is no statistically

significant difference in enthalpy, indicating that the

observed Tg? decreases are most likely due to plasticization.

Concluding remarks

The literature on nanocomposite cure studies to date has

almost always attributed all observed cure effects in epoxy-

based silver, graphene, and BN nanocomposites to the

nanoparticles with little consideration of incomplete cure

or residual solvent as potential sources of error. Some

authors are aware that incomplete cure can affect results,

but no attempts were made to correlate results with the final

degree of cure. In general, residual solvent effects are

poorly addressed and unaccounted for in data analysis. As

far as the authors are aware, studies in the literature have

not included solvent control groups. This limits our ability

to draw conclusions, although enough data without solvent

effects were available to enable our analysis.

The current literature on silver, graphene, and BN

nanocomposites is consistent with established knowledge

regarding Tg and cure rate. Chemically inert, well-dis-

persed, and rigid nanoparticles slowed cure kinetics but

increased final Tg, while those with flexible surface coat-

ings lowered the final Tg of the composite. Nanoparticles

with reactive surface chemistry had the potential to raise Tg

and accelerate cure depending on the functional groups

available and the bonding between nanoparticle and

matrix. Aggregated nanoparticles, regardless of surface

chemistry, reduced final Tg and displayed an insignificant

effect on cure rates.

Using solvent control groups to differentiate solvent

effects from filler effects, we conducted our own DSC

experiments. When Tg results are analyzed accounting for

the presence of solvent, the microparticles had no statisti-

cally significant effect on Tg. Solvent effects were statis-

tically significant. At relatively low residual solvent

Thermal analysis of epoxy-based nanocomposites

123

contents, Tg? decreases due to the solvent plasticization

were observed. At higher solvent contents, Tg? decreases

can be attributed to both plasticization and decreased cross-

linking. When solvent controls are not included in the

analysis, the erroneous apparent microparticle effect is very

similar to reported effects for nanoparticles in the

literature.

Given the high likelihood for misinterpretation, any

study using solvent blending should quantify the residual

solvent present in their nanocomposites before cure, and

separate its effect from that of the nanoparticles using

solvent-only control groups. For Tg analysis, complete cure

should be confirmed and methods reported, particularly if

final Tg is near or above the cure temperature. In nano-

composite studies using reactive nanoparticles, there is the

concern that the sufficient addition of nanoparticles may

alter the chemical ratio of the matrix, leading to reduced

cross-linking density and Tg. Although it is not possible to

draw a conclusion from the available literature, this con-

cern warrants further study.

Acknowledgements The financial support of the Natural Sciences

and Engineering Research Council of Canada (NSERC) Strategic

Grant program for this research study is gratefully acknowledged by

the authors.

References

1. Tjong SC. Structural and mechanical properties of polymer

nanocomposites. Mater Sci Eng R. 2006;53:73–197.

2. Crosby AJ, Lee J-Y. Polymer nanocomposites: the ‘‘Nano’’ effect

on mechanical properties. Polym Rev. 2007;47:217–29.

3. Jordan J, Jacob KI, Tannenbaum R, Sharaf MA, Jasiuk I.

Experimental trends in polymer nanocomposites—a review.

Mater Sci Eng A. 2005;393:1–11.

4. Sahoo NG, Rana S, Cho JW, Li L, Chan SH. Polymer nano-

composites based on functionalized carbon nanotubes. Prog in

Polym Sci. 2010;35:837–67.

5. Rao CNR, Sood AK, Subrahmanyam KS, Govindaraj A.

Graphene: the new two dimensional nanomaterial. Angew Chem

Int Ed. 2009;48:7752–77.

6. Das RN, Lauffer JM, Knadle K, Vincent M, Poliks MD, Mar-

kovich VR. Nano and micro materials in a Pb-Free world. IEEE

Electron Compon and Tech Conf. 2011;1228–33.

7. Winey KI, Vaia RA. Polymer nanocomposites. MRS Bull.

2007;32:314–9.

8. Tao AR, Habas S, Yang P. Shape control of colloidal metal

nanocrystals. Small. 2008;4(3):310–25.

9. Kuilla T, Bhadra S, Yao D, Kim NH, Bose S, Lee JH. Recent

advances in graphene based polymer composites. Prog Polym

Sci. 2010;35:1350–7.

10. Song W-L, Wang P, Cao L, Anderson A, Meziani MJ, Farr AJ,

Sun Y-P. Polymer/boron nitride nanocomposite materials for

superior thermal transport performance. Angew Chem Int Ed.

2012;51:6498–501.

11. Zhi C, Bando Y, Tang C, Golberg D. Engineering of electronic

structure of boron-nitride nanotubes by covalent functionaliza-

tion. Physical Rev B. 2006;74:153413.

12. Lazar M. Let’s review: physics, the physical setting. 3rd ed.

United States: Barrons; 2007. p. 217.

13. ‘‘The Nobel Prize in Physics 2010—Advanced Information’’.

Nobelprize.org. 14 Jun 2013 http://www.nobelprize.org/nobel_

prizes/physics/laureates/2010/advanced.html.

14. Incropera FP, DeWitt DP. Fundamentals of heat and mass

transfer. 5th ed. Hoboken: Wiley; 2002. p. 907.

15. Singh V, Joung D, Zhai L, Das S, Khondaker SI, Seal S.

Graphene based materials: past, present and future. Prog Mater

Sci. 2011;56:1178–271.

16. Golberg D, Bando Y, Huang Y, Terao T, Mitome M, Tang C, Zhi

C. Boron nitride nanotubes and nanosheets. ACS Nano.

2010;4(6):2979–93.

17. Amoli BM, Gumfekar S, Hu A, Zhou N, Zhao B. Thiocarboxy-

late functionalization of silver nanoparticles: effect of chain

length on the electrical conductivity of nanoparticles and their

polymer composites. J Mater Chem. 2012;22:20048–56.

18. Jiang H, Moon KS, Li Y, Wong CP. Surface functionalized silver

nanoparticles for ultrahigh conductive polymer composites.

Chem Mater. 2006;18:2969.

19. Kim H, Abdala AA, Macosko CW. Graphene/polymer nano-

composites. Macromolecules. 2010;43:6515–30.

20. Yu J, Huang X, Wu C, Wu X, Wang G, Jiang P. Interfacial

modification of boron nitride nanoplatelets for epoxy composites

with improved thermal properties. Polymer. 2012;53:471–80.

21. Zhi C, Bando Y, Tang C, Golberg D. Boron nitride nanotubes.

Mater Sci and Eng R. 2010;70:92–111.

22. Solanki JN, Murthy ZVP. Controlled size silver nanoparticles

synthesis with water-in-oil microemulsion method: a topical

review. Ind Eng Chem Res. 2011;50:12311–23.

23. Miller SG, Bauer JL, Maryanski MJ, Heimann PJ, Barlow JP,

Gosau J-M, Allred RE. Characterization of epoxy functionalized

graphite nanoparticles and the physical properties of epoxy

matrix nanocomposites. Compos Sci Tech. 2010;70:1120–5.

24. Potts JR, Dreyer DR, Bielawski CW, Ruoff RS. Graphene-based

polymer nanocomposites. Polymer. 2011;52:5–25.

25. Prime RB. Thermosets. In: Turi EA. Thermal characterization of

polymeric materials. New York: Academic Press; 1981.

p. 548–53.

26. Odom RE. RK-TR-63-20: Kinetics of epoxide-carboxylic reac-

tion. army missile research development and engineering labo-

ratory, redstone arsenal Al, Propulsion Directorate; 1963.

27. Ngai KL. The glass transition and the glassy state. In: Mark J,

Ngai KL, Graessley W, Mandelkern L, Samulksi E, Koenig J,

Wignall G, editors. Physical properties of polymers. 3rd ed.

Cambridge: Cambridge Press; 2003. p. 72–146.

28. Wunderlich B. The basis of thermal analysis. In: Turi EA, editor.

Thermal characterization of polymeric materials. New York:

Academic; 1981. p. 169–74.

29. Utracki LA. Polymer alloys and blends: thermodynamics and

rheology. New York: Hanser Publishers; 1990. p. 93–105.

30. Mark J. The rubber elastic state. In: Mark J, Ngai KL, Graessley

W, Mandelkern L, Samulksi E, Koenig J, Wignall G, editors.

Physical properties of polymers. 3rd ed. Cambridge: Cambridge

Press; 2003. p. 3–71.

31. DiBenedetto AT. Prediction of the glass transition temperature of

polymers: a model based on the principle of corresponding states.

J Poly Sci B. 1987;26:1949–69.

32. Menczel JD, Judovits L, Prime RB, Bair HE, Reading M, Swier

S. Differential scanning calorimetry (DSC). In: Menczel JD,

Prime RB, editors. Thermal analysis of polymers fundamentals.

Hoboken: Wiley; 2009. p. 138–44.

33. Lau K-T, Lu M, Lam C-K, Sheng F-L, Li H-L. Thermal and

mechanical properties of single-walled carbon nanotube bundle-

reinforced epoxy nanocomposites: the role of solvent for nano-

tube dispersion. Compos Sci and Tech. 2005;65:719–25.

G. Rivers et al.

123

34. Mondragon I, Bucknal CB. Effects of residual dichloromethane

solvent on the cure of epoxy resin. Plast Rubber Compos Process

Appl. 1994;21(5):275–81.

35. Loos MR, Coelho LAF, Pezzin SH, Amico SC. The effect of

acetone addition on the properties of epoxy. Polym Sci Tech.

2008;18(1):76–80.

36. Hong S-G, Wu C-S. DSC and FTIR analysis of curing behaviors

of epoxy/DICY/solvent open systems. Thermochimica Acta.

1998;316:167–75.

37. Yagci Y, Sangermano M, Rizza G. A visible light photochemical

route to silver–epoxy nanocomposites by simultaneous poly-

merization–reduction approach. Polymer. 2008;49:5195–8.

38. Park S, Kim DS. Preparation and physical properties of an epoxy

nanocomposite with amine-functionalized graphenes. Polym Eng

Sci. 2012. doi:10.1002/pen.23368.

39. Teng CC, Ma CCM, Lu CH, Yang SY, Lee SH, Hsiao MC, Yen

MY, Chiou KC, Lee TM. Thermal conductivity and structure of

non-covalent functionalized graphene/epoxy composites. Carbon.

2011;49:5107–16.

40. Zhi CY, Bando Y, Terao T, Tang C, Golberg D. Dielectric and

thermal properties of epoxy/boron nitride nanotube composites.

Pure Appl Chem. 2010;82(11):2175–83.

41. Ganguli S, Roy AK, Anderson DP. Improved thermal conduc-

tivity for chemically functionalized exfoliated graphite/epoxy

composites. Carbon. 2008;46:806–17.

42. Voo R, Mariatti M, Sim LC. Properties of epoxy nanocomposite

thin films prepared by spin coating technique. J Plast Film Sheet.

2011. doi:10.1177/8756087911419745.

43. Martin-Gallego M, Verdejo R, Lopez-Manchado MA, Sanger-

mano M. Epoxy-graphene UV-cured nanocomposites. Polymer.

2011;52:4664–9.

44. Chan KL, Mariatti M, Lockman Z, Sim LC. Effects of the size and

filler loading on the properties of copper- and silver-nanoparticle-

filled epoxy composites. J Appl Polym Sci. 2011;121:3145–52.

45. Rafiee MA, Rafiee J, Srivastava I, Wang Z, Song H, Yu ZZ,

Koratkar N. Fracture and fatigue in graphene nanocomposites.

Small. 2010;6(2):179–83.

46. Zaman I, Phan TT, Kuan HC, Meng Q, La LTB, Luong L, Youssf

O, Maa J. Epoxy/graphene platelets nanocomposites with two

levels of interface strength. Polymer. 2011;52(7):1603–11.

47. Guo Y, Bao C, Song L, Yuan B, Hu Y. In situ polymerization of

graphene, graphite oxide, and functionalized graphite oxide into

epoxy resin and comparison study of on-the-flame behavior. Ind

Eng Chem Res. 2011;50:7772–83.

48. Liang H, Yu D. Mechanical and thermal properties of (Ag/C

nanocable)/epoxy resin composites. Polym Sci Ser B. 2011;

53(11):601–5.

49. Hu Y, Shen J, Li N, Ma H, Shi M, Yan B, Huang W, Wang W,

Ye M. Comparison of the thermal properties between composites

reinforced by raw and amino-functionalized carbon materials.

Comp Sci and Tech. 2010;70:2176–82.

50. Guo B, Wan J, Lei Y, Jia D. Curing behaviour of epoxy resin/

graphite composites containing ionic liquid. J Phys D. 2009;42:

145307–14.

51. Yung KC, Liem H. Enhanced thermal conductivity of boron

nitride epoxy-matrix composite through multi-modal particle size

mixing. J Appl Polym Sci. 2007;106:3587–91.

52. Bortz DR, Heras EG, Martin-Gullon I. Impressive fatigue life and

fracture toughness improvements in graphene oxide/epoxy com-

posites. Macromolecules. 2012;45:238–45.

53. Qiu SL, Wand CS, Wang YT, Liu CG, Chen XY, Xie HF, Huang

YA, Cheng RS. Effects of graphene oxides on the cure behaviors

of a tetrafunctional epoxy resin. eXPRESS Polym Lett.

2011;5(9):809–18.

54. Huang H, Zhi C, Jiang P, Golberg D. Bando Y, Tanaka T. Adv

Funct Mater: Polyhedral oligosilsesquioxane-modified boron

nitride nanotube based epoxy nanocomposites: an ideal dielectric

material with high thermal conductivity; 2012.

55. Liu Q, Zhou X, Fan X, Zhu C, Yao X, Liu Z. Mechanical and

Thermal Properties of Epoxy Resin Nanocomposites Reinforced

with Graphene Oxide. Polym -Plast Tech Eng. 2012;51:251–6.

56. Shen XJ, Pei XQ, Fu SY, Friedrich K. Significantly modified

tribological performance of epoxy nanocomposites at very low

graphene oxide content. Polymer. 2013;54:1234–42.

57. Cui HW, Fan Q, Li DS, Tang X. Formulation and characteriza-

tion of electrically conductive adhesives for electronic packaging.

J Adhes. 2013;89:19–36.

58. Suriati G, Mariatti M, Azizan A. Effects of filler shape and size

on the properties of silver filled epoxy composite for electronic

applications. J Mater Sci. 2011;22:56–63.

59. Monti M, Rallini M, Puglia D, Peponi L, Torre L, Kenny JM.

Morphology and electrical properties of graphene–epoxy nano-

composites obtained by different solvent assisted processing

methods. Compos A. 2013;46:166–72.

60. Kim K-S, Jeon I-Y, Ahn S-N, Kwon Y-D, Baek J-B. Edge-

functionalized graphene-like platelets as a co-curing agent and a

nanoscale additive to epoxy resin. J Mater Chem. 2011;21:7337.

61. Bao C, Guo Y, Song L, Kan Y, Qian X, Hu Y. In situ preparation

of functionalized graphene oxide/epoxy nanocomposites with

effective reinforcements. J Mater Chem. 2011;21:13290.

62. Corcione E, Freuli F, Maffezzoli A. The aspect ratio of epoxy

matrix nanocomposites reinforced with graphene stacks. Polym

Eng Sci. 2012. doi:10.1002/pen.

63. Zaman I, Kuan HC, Meng Q, Michelmore A, Kawashima N, Pitt

T, Zhang L, Gouda S, Luong L, Ma J. A facile approach to

chemically modified graphene and its polymer nanocomposites.

Adv Funct Mater. 2012;22:2735–43.

64. Ren H, Tang S, Syed JA, Meng X. Incorporation of silver

nanoparticles coated with mercaptosuccinic acid/poly(ethylene

glycol) copolymer into epoxy for enhancement of dielectric

properties. Mater Chem and Phys. 2012;137:673–80.

Thermal analysis of epoxy-based nanocomposites

123