Polymer Composites in the Aerospace Industry || The response of aerospace composites to temperature and humidity

The response of aerospacecomposites to temperatureand humidity

12
F.R. Jones, J.P. ForemanUniversity of Sheffield, Sheffield, UK
12.1 Introduction

Polymer matrix composites have matured over recent years with their application asmajor structural components of aircraft such as the wing box on the Airbus A380.The main advantage of fibre composites arises from their high specific strength andstiffness which means that low weight structures can be designed. Energy costshave driven the demand for low weight structural composites. The gradual replace-ment of non-structural components has also provided designers with the confidenceto increase the number of applications. However, composites have a long track recordgoing back to the beginnings of the plastics industry, more than 100 years ago, wherephenolic resinepaper cores were essential to the development of electrical applica-tions, and fabric-reinforced phenolic mouldings became popular for more complexshaped artefacts. However, more critical was the use of ‘fibreglass’ for the hulls ofleisure boats in the 1950s, military ships in the 1970s and pipes for transportingboth potable drinking water (without taint) and corrosive effluents. Here, the durabilityand corrosion resistance of glass fibreereinforced plastic (GFRP) provided the maindriver. The development of carbon fibre composites since the 1960s spurred academicand industrial studies so that full understanding of the micromechanics of failure hasbeen achieved over several decades [1,2].

A composite consists of a reinforcement, which is mainly a high-performance fibreor micro- or nano-sized inorganic particle dispersed in a polymeric matrix. The lattercan be acicular, such as carbon nanotubes. Most recently, hybrid systems have beendeveloped to reduce the volume fraction of the polymer matrix, which has the lowestmodulus of all components. The polymer is also the most sensitive to absorption ofaqueous and organic fluids. The effect of environments is also complicated by themorphology of the matrix, which often involves a phase-separated thermoplastic ina thermoset resin. In some aerospace laminates, heterogeneous rubber or thermoplasticparticles are used in the inter-laminar regions as toughening agents. A further compli-cation is the formation of interphases between the reinforcement and matrix, whichrepresents a region of mixed polymer at the reinforcement interface [3].

This chapter is concernedwith the effect of absorption ofmoisture (and other organics)on the micromechanics of failure of aerospace composite materials. Figure 12.1 shows aschematic temperature and humidity profile of a military aircraft on a typical flight sortie

Polymer Composites in the Aerospace Industry. http://dx.doi.org/10.1016/B978-0-85709-523-7.00012-8Copyright © 2015 Elsevier Ltd. All rights reserved.

http://dx.doi.org/10.1016/B978-0-85709-523-7.00012-8

Skin R.H.

Skin temperature

Take off and climb

0

40

80

120

Time from take-off (h)

Ski

n te

mpe

ratu

re/re

lativ

e hu

mid

ity (°

C/ %

R.H

.)

0 2.5 5.0 7.5 10.0

Cruise at 17 km M-2.1

Cruise @ 7.6 km M-0.7

Terrain follow at sea level

Figure 12.1 Schematic of skin temperature and local relative humidity during a typical flight ofa military jet [4].

336 Polymer Composites in the Aerospace Industry

[4]. On the ground, the relative humidity (RH) is high but decreases on take-off; however,one may encounter flight at sea level where humidity rises. Also, an aircraft’s skin tem-perature depends on speed and elevation. Of note is the rapid rise in skin temperature dur-ing supersonic flight to temperatures in excess of 120 �C, referred to as thermal spiking.To understand the effects of these flight variables, we will discuss Fickian diffusion andthe sensitivity of polymers and fibres to aqueous and nonaqueous environments. We willalso describe the effect of thermal spiking on moisture absorption since this could be amajor source of residual stress in an aerospace composite.

12.2 Moisture absorption

Organic matrix polymers are sensitive to water ingress. Therefore, in humid environ-ments, moisture diffuses into polymer matrix composites to differing degrees. The extentof moisture absorption depends on the molecular structure of the polymer. This aspect isdescribed in Section 12.3. Dissolved water can plasticise the matrix and reduce the glasstransition temperature, Tg, and hence service temperature of the composite significantly.In conjunction with this, hydrolytic degradation can reduce the thermo-mechanical prop-erties further. For aerospace structures in particular, the rate of absorption and its temper-ature dependence is especially critical since life of an aircraft is several decades.

Most structural composites are highly anisotropic, consisting of unidirectionalmaterials where the mechanical properties in the longitudinal (or 0�) direction are fibredominated, while in the transverse (or 90�) direction they are matrix dominated.

The response of aerospace composites to temperature and humidity 337

Therefore, moisture absorption has a larger effect on the transverse properties of atypical composite system. Despite this, the strength of a 0� composite is also affectedby moisture ingress since the reloading of a broken fibre occurs through shear stresstransfer from the ‘interphasal’ matrix. To achieve isotropy, unidirectional plies arestacked at a set of angles such as 0�,�45� and 90� to form a laminate. In this situation,moisture ingress will modify the residual stress state in the individual laminae.

Whereas the thermodynamic properties of the resin and environment will determinethe maximum extent of moisture absorption, diffusion is a kinetic property which istemperature dependent. Normally, the latter is considered to be Fickian in nature.For highly polar polymers, other diffusion models need to be considered, but formost structural composite materials, the Fickian laws are applicable.

12.2.1 Fickian diffusion

Fickian laws were developed for thermal equilibration, but thermal diffusion issignificantly faster than moisture diffusion by a factor of w106. This means thata 12 mm thick composite can take 13 years for moisture equilibration at 350 K,while thermal equilibration would take only 15 s. The consequence of this is thata typical aerospace structure will take many years in service before the structurehas become saturated with water. Thus, engineers need to deal with propertygradients resulting from moisture absorption. However, this is complicated by thevariations in temperature and RH at different storage and holding locations andduring flight. As a result of this, water will tend to move within a structure inresponse to environmental fluctuations. For instance, at these locations with ahigh ambient temperature, moisture will diffuse deeper into the structure as eachmolecule has a higher energy. Correspondingly, during flight missions, the structurewill experience low RH at high altitudes and the moisture will tend to move inthe opposite direction. However, this effect is inhibited by the low rates at lowtemperatures [5,6].

The distribution of water throughout the composite will arise from an averaging of aseries of moisture profiles over time. The solution to Fick’s second law, which definesthe diffusion constant (D) for moisture uptake, is given by Eqn (12.1):

D ¼ pd2

16M2N

M2 �M1

t1=22 � t1=21

!2

(12.1)

whereM1 and M2 are the moisture contents at times t1 and t2; and d is the thickness ofthe specimen. The experimental measurement of D uses coupons where diffusionthrough all the faces occurs so that an edge correction is required for the calculation ofa true one-dimensional (1D) diffusion constant [5e8].

For the prediction of the time required for moisture equilibration of a structure,the diffusion constant, D, for each material is required, but their measurement istime-consuming because the equilibrium moisture content (MN) has to be deter-mined in long-term tests. Therefore, accelerated tests are used. As with other


kinetic processes, D is related to temperature (T) according to the Arrheniusequation,

D ¼ A expð � Ea=RTÞ (12.2)

where Ea is the activation energy for moisture transport and A is the pre-exponentialfactor.

While many resin systems exhibit Fickian diffusion, others show non-Fickianbehaviour. The experimental reabsorption diffusion curve shown in Figure 12.2represents a smooth process which is linear over the first 60% before reaching equi-librium. For the resin system used, Fickian diffusion could only be obtained afterpreconditioning because often the early stage of moisture absorption deviates fromFickian kinetics. This creates difficulties in the calculation of a precise diffusion con-stant. It is thought that the non-Fickian behaviour results from the reorganisation ofthe network as the material becomes plasticised [9e11].

Figure 12.2 also illustrates this point where the moisture concentration in thefirst absorption appears to approach equilibrium, but then after a longer periodthe moisture sorption accelerates. Much work on the thermal spiking phenomenon(see Section 12.6) has confirmed that the relaxation of the network occurs in thepresence of water making more polar sites available which can bind additionalwater molecules.

AbsorptionFickian absorptionDesorptionFickian desorptionRe-absorptionFickian re-absorption

Moi

stur

e co

ncen

tratio

n/ re

sidu

al w

eigh

t (w

t.%)

0

0.5

1.0

1.5

2.0

2.5

√Time (s½)0 1000 30002000

Figure 12.2 Isothermal absorption and re-absorption curves for the bismaleimide epoxy resinNarmco Rigidite 5245 at 90% RH at 45 �C. Calculated theoretical Fickian first absorption anddesorption curves are also shown [4].


Composite materials are typically anisotropic, so the diffusion constants in thelongitudinal and transverse directions can differ. These are related to each otherthrough Eqn (12.3):

D ¼ Dx

2641þ d

l

�Dy

Dx

�12

þ d

b

�Dz

Dx

�12

3752

(12.3)

where Dx, Dy and Dz are the diffusion constants through the thickness, d, along thelength, l, and across the breadth, b.

For a unidirectionally reinforced 0� ply, Dz and Dx are equivalent, both occurring at90� to the fibres. Therefore, Eqn (12.3) becomes:

D ¼ Dx

2641þ d

bþ d

l

�Dy

Dx

�12

3752

(12.4)

It is possible to measure the influence of the fibres on the diffusion because, with apoor interfacial bond between the fibre and the resin, rapid transport will take place atthe interface and can be differentiated from the resin-dominated diffusion at 90� to thefibres. If Dy is greater than Dx, then capillary diffusion at a poor interfacial bond mustbe occurring. For perfectly bonded fibres, Dx will be significantly larger than Dy

because of the relatively higher surface area of resin in the 90� direction. Typicalvalues of the diffusion constant are 10�6 mm2/s for a resin and 10�7 mm2/s for acomposite.

12.2.2 Prediction of moisture content and time dependence

We can predict the distribution of moisture in a laminate using Fickian diffusionkinetics. Figure 12.3 shows the calculated distribution of moisture in a carbon fibrelaminate of thickness 0.94 mm after 3 days at 96% RH and 50 �C. As result of Fickiananalysis, we can observe that it would take 13 years to reach equilibrium if the materialhad a thickness of 12 mm. However, this does not reflect the actual environment sincethe temperature and RH can change.

12.2.3 Moisture distribution in a laminate

Figure 12.4 shows how the moisture profile in a laminate through thickness changesas a result of a thermal excursion. Since the rate of moisture diffusion is controlledby an activation process (Eqn (12.2)) at higher temperatures, the moisture willdesorb from the surface, whereas subsurface moisture will move into structure.Therefore, over several cycles, the moisture profile will even out throughout thethickness.

Figure 12.3 Theoretical moisture distribution within a unidirectional carbon fibre laminate fromNarmco Rigidite 5245C, after 3 days at 50 �C and 96% RH (curved line) and after a rapidthermal excursion to 150 �C (the ‘spiked’ curve) The shaded area indicates the water which hasmoved out of the laminate during the thermal excursion [7].

Figure 12.4 Theoretical moisture distribution within a unidirectional carbon fibre laminate fromNarmco Rigidite 5245C, after 3 days at 50 �C and 96% RH and a rapid thermal excursion to150 �C. The shaded area indicates the water which has moved into the laminate during thethermal excursion [7].


12.2.4 Interfacial stability during moisture absorption

The durability of a composite material to moisture ingress can often be determined bythe maintenance of interfacial integrity. Therefore, it is useful to study diffusion indifferent directions of a unidirectional composite.


Equation (12.4) can be used to calculate the diffusion constants in the direction par-allel, Dx, and, at 90�, Dy and Dz. Rearrangement gives

D12 ¼ ð1þ d=bÞD

12x þ ðd=lÞD

12y (12.5)

Therefore, a plot of Eqn (12.5) enablesDx andDy to be estimated from the diffusionconstants for coupons of differing geometry. In this way, the degradation of the inter-face under aggressive environments in accelerated tests can be identified. However,interfacial degradation to moisture should be avoided by correct selection of fibreand sizing in separate experiments. Often, single-filament tests such as fragmentation,pull-out or micro-droplet can be used to assess the interfacial stability of certaincombinations of fibres and resin. With aerospace carbon fibres, the oxidative fibretreatment coupled with epoxy resin sizes in epoxy matrices provides a ‘good’ interface.It should be remembered that a perfect interface will lead to brittle failure and thatpartial debonding or interphasal yield is required for tough performance [3].

12.3 Moisture sensitivity of matrix resins

Moisture diffuses into polymers to differing degrees, which is a function of themorphology and molecular structure. The following aspects are critical:

1. The polarity of the molecular structure2. The degree of cross-linking3. The degree of crystallinity in the case of a thermoplastic matrix (although certain modified

thermosets can also have a crystalline component)4. The presence of residuals in the material

While (1e3) are clearly variables of the matrix, the presence of residuals (4) is not soobvious since this refers to unreacted hardeners and other impurities. Impurities suchas calcium, sodium or potassium salts arise from inclusions in fillers or as contaminantson the surface of reinforcing fibres. The sodium and potassium content of glass fibres hasbeen steadily reduced over recent years because they were widely acknowledged asresponsible for the water sensitivity of glass fibre composites in aqueous environments.

Table 12.1 gives typical values of the diffusion constant for a series of thermoset-ting matrices where the effect of polarity of the resin on the equilibrium on moistureconcentration is illustrated. Epoxy resins are the network product of the reaction of amultifunctional epoxide monomer with a hardener. One of the important aspects ofcomposite materials which needs to be recognised is that the material is synthesisedat the same time as the component is manufactured. The precise chemistry of the finalnetwork can be uncertain but is a function of the chemical structures of the epoxide andhardener and/or catalyst, which determine the mechanism of cure. In this way, thecured resins can have differing polarities.

For example, an anhydride hardener (with tertiary amine accelerator) will produce a‘polyester’-type structure with low polarity, whereas an amine hardener will form ab-hydroxyl amine network with high polarity. Hydrogen bonding of water moleculesat these sites enhances the concentration of moisture which can be absorbed. However,with catalytic curing agents, polyether structures of relatively low polarity will form.Since it is quite common to use mixed curing agents for composite matrices, the cured

Table 12.1 Typical moisture diffusion coefficients and maximummoisture concentrations of aerospace epoxy matrix resins [8,10]

Resin MN/% D/10L3 mm2/sRelativehumidity/% T/�C

Epoxy: General purpose(DGEBA)

6.9 2.8 96 50

Epoxy: Dow DEN 431 (EpoxyNovolak)

0.48 5.99 11 50

Epoxy aerospace: Narmco 5245 1.96 11.0 96 45

1.38 14.7 75 45

0.8 19.4 46 45

0.54 17.2 31 45

Epoxyecyanate ester: PrimasetPT30/DEN 431

2.96 3.62 96 50

2.19 4.85 75 50

1.44 6.34 46 50

1.04 6.12 31 50

0.70 5.46 16.5 50

0.48 5.99 11 50


network will have a complex structure where its polarity is strongly dependent on thedegree of incorporation of the individual components and the chemical mechanismsemployed. In Table 12.1, it is clearly shown how the choice of resin and curing mech-anism strongly influences the moisture sensitivity.

12.3.1 Epoxy resins

Figure 12.5 shows an example of the cured structure of a thermoset resin using twodifferent epoxide monomers with two different amine hardeners. The epoxies aretriglycidyl p-aminophenol (TGAP) and tetraglycidyl 4,4-diaminodiphenylmethane(TGDDM), which are tri- and tetra-functional, respectively (n¼ 3 and n¼ 4,respectively); and the two hardeners are 4,4-diamino-diphenylsulphone (DDS)and dicyandiamide (DICY).

12.3.2 Advanced matrix resins

As shown in Table 12.2, thermoplastic matrices such as polyethersulphone (PES) andpolyetheretherketone (PEEK) absorb much less moisture than the advancedepoxies. PEEK is an example of a partially crystalline linear polymer with very lowmoisture absorption. Thermoplastics which are relatively nonpolar and absorb low

Table 12.2 Moisture content of carbon fibre composites fromthermosetting and thermoplastic matrices; also shown is the benefitof modifying epoxy resin with thermoset and thermoplasticmodifiers to reduce MN

Resin Type Modifier MN/%Relativehumidity/% T/�C

Fibredux924E*

Epoxythermoset

None 2.44 96 50

924C Epoxy PES 1.72 96 50

927C Epoxy Cyanateester/PI

0.98 96 50

NarmcoRigidite5255C

Epoxy BMI 0.82 96 50

APC 2 PEEKthermoplastic

None 0.02 50 23 (350 h)

APC 2 PEEK None 0.23 100 (Immersion) 100 (360 h)

PMR 15 BMI thermoset None 0.32 96 50

PES, polyethersulphone; PI, polyimide; BMI, bismaleimide.*Calculated from data on cast resins (6.95%).

Figure 12.5 Typical chemical structure of a cured epoxy resin showing the use of two epoxiesand two hardeners. R represents a connection to the 3D network.



concentrations of moisture are also used as ‘flow-control’ additives and/or tougheningagents in advanced epoxy resins and help provide a reduction in moisture sensitivity.

In recent times, commodity thermoplastics have become significant contenders asmatrices for glass fibre composites where environmental resistance is sought; forexample, polypropylene is now available in a pre-preg form for fusion bonding.Since it has low polarity and is also partially crystalline, moisture absorption isvery low.

12.4 Mechanism of moisture retention in aerospaceepoxies

12.4.1 Chemical aspects

The epoxy amine cure mechanism begins with energetically favourable ring openingof the unstable epoxide group by a primary amine, resulting in the formation of ahydroxyl group and secondary amine. This is followed by the less favourable reactionof an epoxide group with a secondary amine or the even less favourable reaction of anepoxide group with a hydroxyl group (etherification).

If each monomer reacted only twice (n¼ 2), a simple linear polymer is created, butthis is clearly not the case in thermosetting resin systems. In theory, each TGAPmonomer can react three times through its epoxide groups and further three timesthrough the resultant hydroxyl groups, bringing its maximum theoretical functionalityto six (n¼ 6). However, in practice, this is highly unlikely to occur partly due to theetherification reaction being energetically less favourable and partly due to thedramatic decrease in mobility each monomer experiences as the number of times ithas reacted increases. When a monomer is unreacted, it is free to migrate throughthe monomerepolymerenetwork system. After it has reacted once, its translationalfreedom has been removed, and each subsequent reaction reduces its mobility furtheruntil it is firmly locked into the 3D network. At this point, its ability to react with otherspecies present in the network is significantly hindered.

Epoxy resins used in composite manufacturing are intentionally prepared with anexcess of epoxy compared to hardener. This is done to ensure that the hardener iscompletely reacted during cure and is not allowed to remain in the network and causeplasticisation. A mixture with an epoxyehardener ratio of 1:1 would not be able tofully cure due to the mobility arguments given in this chapter. The epoxy excess meansthat there will be a significant number of unreacted and relatively polar functionalgroups present in the cured 3D network.

Each time a monomer reacts more than twice (n> 2), it is considered to have chem-ically cross-linked, creating a permanent covalent link between adjacent polymerchains. Multifunctional epoxy resins such as TGAP and TGDDM form networkswith diamines such as DDS and DICY with a significant number of cross-links present.Extensive cross-linking adds to the strength and stiffness of the cured resin but pro-duces a brittle material with little inherent toughness. This can be a problem if theetherification reaction is able to occur to a significant level. A highly cross-linked


3D network will contain enclosed areas where no further reactions are possible,leading to the so-called free volume.

It is clear that the type and ratio of functional groups present in the network willaffect its polarity and degree of cross-linking [12e14] and hence its moisture absorp-tion characteristics. While the original monomers are practically frozen in place as afused network, diluents introduced into the system are capable of diffusing deepinto the network. In the case of water, the additional molecules either merge to formdiscrete micro-phases located in the free volume or hydrogen-bonded to polar func-tional groups on the polymer chain. The ratio of water present in the free volume tohydrogen-bonded water can be estimated using near-infrared spectroscopy (NIR),such as described by Mijovic and Zhang [15]. While free and hydrogen-bonded waterhave different characteristic peaks, as shown in Figure 12.6, it is difficult to obtain aprecise ratio.

The absorbed water has a distinct and often undesirable effect on the thermo-mechanical properties of the resin. Discrete pockets of water will plasticise the networkby dissipating thermo-mechanical energy via the translational degrees of freedomassociated with unattached molecules. Correspondingly, individual water moleculeshydrogen-bond to appropriate sites such as hydroxyl and amino groups in particular.Here, while there is an increase in binding energy due to the hydrogen bond, there isalso increased energy dissipation because each individual water molecule imparts extradegrees of freedom to the system. In the case of small diluents such as water, which

Figure 12.6 NIR difference spectra for diglycidyl ether of bisphenol-A (DGEBA)ediethylenetriamine (DETA) with absorbed moisture levels varying between 0.46% and 2.69%. The insetshows a deconvoluted analysis with peaks at 7085, 6834 and 6541/cm, which correspond to freewater and singly and doubly hydrogen-bonded water, respectively [15].


have a high value of specific degrees of freedom, this energy balance favoursdissipation which accounts for the deleterious effect such molecules have onthermo-mechanical properties. For larger diluents, such as unreacted monomers andoligomers, the situation is less clear as these molecules will have lower value forthe specific degrees of freedom.

Hydrogen bonding of water molecules to the cross-linked network in moisture-affected epoxies is further complicated by the likelihood that the absorbed water canhydrogen-bond more than once, as shown in Figure 12.7. Several analyses of such sys-tems suggest that the proportion of absorbed moisture that is doubly hydrogen-bondedto the network may be as high as 20% [15]. Whether the doubly bonded water prefer-entially acts as a donor and acceptor or as a donor twice is unclear, as is whether it isdoubly bonded to a single polymer chain or between two adjacent chains (creating afurther weaker cross-link). There appears to be no evidence to suggest that water iscapable of hydrogen bonding more than twice per molecule in such systems.

12.4.2 Predictive modelling

There are a number of models which can be used to predict the effect that absorbeddiluents, in particular water, have on a cured polymeric resin. One of the most power-ful of these is Group Interaction Modelling (GIM), a continuum-type model with a setof versatile input parameters based on the number and type of chemical functionalgroups present in the network. This allows the complex chemistry of amine-curedepoxy resins to be catered for whilst retaining the speed afforded by using a set oflinked constitutive equations of state for property prediction.

GIM is particularly suitable for predicting the change in polymer propertiesobserved upon introduction of a diluent into a thermosetting resin. Firstly, the model

Figure 12.7 Variation in hydrogen bonding between absorbed water and polymer backbone.(a) Singly bound water as acceptor; (b) singly bound water as donor; (c) doubly bound water asdonor and acceptor; (d) doubly bound water as donor twice. Hydroxyl groups are shown fromthe polymer backbone, but amino groups are also possible as donor and acceptor.


deals with chemical cross-linking by reducing the degrees of freedom assigned to eachmer unit with a cross-link site. Secondly, the addition of diluents is accounted for bymodifying the input parameters of the system according to the amount and type of ex-tra molecules present. For example, water which is free to migrate through the networkhas a significantly higher number of translational degrees of freedom than a monomerwhich has reacted two or more times and is locked into the 3D network. Additionally,hydrogen bonding is accounted for by increasing the binding energy of the system foreach hydrogen bond. In this way, multiple hydrogen bonding can be catered for in theproperty prediction routines.

The fundamental equation of state central to GIM is a modified LennardeJonespotential function which describes the interaction energy between adjacent polymers,E. This function has powers of 6 and 3 instead of the normal 12 and 6 because volume,V, is proportional to the square of the interchain separation distance, r. In a polymer,the chain length is significantly larger than r and is therefore assumed to be invariant.Ecoh refers to the zero point cohesive energy and V0 is the zero point volume.

E ¼ Ecoh

"�V0

V

�6

� 2

�V0

V

�3#

(12.6)

The potential function shown in Eqn (12.6) is used as the basis for the prediction ofa series of thermo-mechanical properties for the polymer as a function of temperature,strain, strain rate and pressure. Similarly, an equation for the prediction of glass tran-sition temperature of the polymer is derived by assuming the change from glassy torubbery properties is centred on the temperature, Tg, at which the attractive forcebetween adjacent polymers is zero. At this point, large-scale translational motionbecomes possible, resulting in a significant dissipation of stored energy which islost as heat. Equation 12.7 uses GIM principles to predict Tg using a remarkablylimited set of inputs, including degrees of freedom, N and the Debye temperature, q1.

Tg ¼ 0:224 q1 þ 0:0513Ecoh

N(12.7)

Equation (12.7) predicts Tg via a baseline stiffness (the Debye temperature, often550 K for phenyl-containing polymers) and a balance between attractive, bindingenergy (Ecoh) and repulsive, thermal energy (the latter being governed by the degreesof freedom, N). A simplistic approach to achieving a higher Tg would be to increaseEcoh without increasing N too much. Clearly, this is easier said than done, and it isworth repeating that thermoset resins with high Tg are also liable to be brittle. Corre-spondingly, an easy way of reducing Tg is to increase N without a simultaneous sig-nificant increase in Ecoh. This is the exact situation which occurs when smallmolecules infiltrate a thermoset network. Water in particular adds significantly tothe degrees of freedom of the system (translational freedom) but only makes a modestincrease in the cohesive energy (hydrogen bonding). Overall, the Tg is reduced asmuch as 20 �C for each 1% absorbed water present in the system.

Table 12.3 GIM parameters for TGAP cured with DDS

Functional group N Ecoh (J/mol)

CH2 2 4500

Phenyl 3 25,000

N 2 9000

CH(OH) 2 20,800

SO2 2 45,000

O 2 6300


By assigning values ofEcoh andN for the epoxy, hardener andwater, the effect of add-ing increasing amounts of water to a cured epoxy can be modelled using GIM via Eqn(12.7). A typical set of parameters are given in Table 12.3 for TGAP cured with DDS.

Along with the composition of the cured resin (the epoxyehardener ratio) andcorrections to N (for cross-linking or translational freedom) and Ecoh (for hydrogenbonding), the GIM prediction of Tg for dry TGAPeDDS is approximately 298 �C.Each 1% of water added to the system reduces the predicted Tg in line with experimentalvalues. The model predictions and experimental values can be seen in Figure 12.8,where there is good agreement between the two sets of data. Experimental values aremeasured by dynamic mechanical thermal analysis (DMTA) and are presented asbroad ranges as there is often significant scatter in the Tg of moisture-affected samples.

Figure 12.8 Schematic diagram showing the effect of absorbed moisture content on the glasstransition temperature(s) in TGAP-DDS resin. The two broad ranges in experimentalmeasurements illustrate the development of a secondary Tg peak. The GIM calculations using theparameters for singly (upper) and doubly (lower) hydrogen-bonded water predict this effect well.


The clear change from one Tg to two peaks in the DMTA plot on absorption ofwater is common to many epoxy resins, including DGEBA, TGAP and TGDDMwhich represent functionalities of n¼ 2, 3 and 4. While the exact origin of this splittingof the Tg peak is still yet to be understood, the possibility that it is a result of absorbedwater being singly and doubly hydrogen-bonded to the polymer network can beexplored using GIM. The two solid lines in Figure 12.8 are predictions of the changein Tg due to increasing moisture using one and two hydrogen bonds per water mole-cule. Ecoh is increased and N decreased to account for the extra hydrogen bond. Theagreement with experimental values between the two sets of model parameters overa range of moisture contents adds weight to the argument that the peak splitting iscaused by hydrogen bonding to differing levels.

12.5 Anomalous effects

12.5.1 Role of impurities and unreacted resin components

Cured resins behave as semi-impermeable membranes allowing water to diffuse butacting as a barrier to larger molecules so that osmosis can occur. Therefore, if the resincontains water-soluble impurities, thermodynamics will drive the water into the resin.Osmotic pressure (p) is a colligative property and is directly proportional to the molalconcentration of the impurity. As shown in Eqn (12.8):

p ¼ RTc (12.8)

where c is the molal (mol/kg) concentration of solubilised inclusions, T is thetemperature and R is the gas constant.

Since the moisture can plasticise and soften the resin, the pressure which developsat an inclusion will cause a blister to form. In the case of glass fibreereinforcedmaterials, this has been referred to as ‘boat pox’. This is especially prevalent in fresh-water areas as opposed to sea environments because c in Eqn (12.8) will be the differ-ence between the concentrations of impurities in the ‘resin’ and the environment. Inunsaturated polyester resins, the impurities leading to osmosis are likely to be residualacids and anhydrides or glycols from the synthesis of the unsaturated polyester. In thecase of glass fibre composites, some aqueous-based sizings can also cause osmosis atthe interface. Therefore, the selection of fibre ‘finish’ is very important in the manufac-ture of a composite destined for use in aqueous (e.g. chemical plant or marine) envi-ronments. With advanced composite materials, osmosis is less of a problem but canoccur. Dicyandiamide (DICY) is commonly used as a ‘latent’ curing agent for epoxyresins. With a melting point of 160 �C, solid DICY can be dispersed into the resin.Curing will only begin when the DICY melts. Accelerators are available which reducethe cure temperature. However, to achieve a satisfactory cure, without the formation ofencapsulated DICY particles, it is essential to finely disperse them throughout thepre-preg material; this is often referred to as micronisation. If this is not done effi-ciently, enhanced moisture absorption can be observed. In some cases, this can lead


to blistering. Furthermore, the hydrolysis products can provide an alkaline environ-ment which can degrade glass fibres [7,8].

Chen and Birley [16,17] demonstrated that the residual stress state can contribute toabsorption of water providing a thermodynamic mechanism akin to osmosis. Blisterswere shown to occur at the interface between the gel-coat and structural resins, andthey were attributed to localised residual stresses. To avoid blister formation on immer-sion in water or other aqueous environments, the presence of impurities, which can actas osmotic centres, needs to be minimised, and the properties of any gel-coat or surfacefinish should be matched to those of the structural resins.

12.6 Thermal spiking

Moisture absorption is also affected by brief excursions to elevated temperatureswhich can lead to an enhancement of the concentration of absorbed moisture. Thisis referred to as thermal spiking. Thermal spikes as short as a minute can cause a sig-nificant increase in moisture content. A typical moisture absorption curve measuredduring the presence of a series of thermal spikes is given in Figure 12.9. Above amaximum spike in temperature (Figure 12.10), the moisture content falls [9]. This sug-gests that the matrix is responsible, and this is confirmed in Figure 12.11, where it canbe seen that the ‘thermal spiking’ phenomenon is associated with the matrix and is nota consequence of combining the reinforcement with the resin.

Figure 12.9 Illustration of the effect of thermal spiking to 140 �C on the moisture absorption ofa Fibredux 927C unidirectional laminate in 96% RH at 50 �C. The individual points representthe actual moisture content immediately before and after a thermal spike. The continuous line isfor the isothermal control under identical humid conditions.

Figure 12.10 Moisture enhancement after thermal spiking of carbon fibre laminates conditionedat 50 �C and 96% RH with 22 thermal spikes. Narmco Rigidite (C); Fibredux 927D (,);Fibredux 924C (B) [7].

Moi

stur

e co

nten

t (w

t.%)

2.5

5.0

7.5

10.0

Thermal spike temperature (°C)50 100 150 200

Figure 12.11 Moisture enhancement of matrix resins: Unmodified epoxy Fibredux 924E (B);thermoplastic modified epoxy Fibredux 924T (,); Fibredux 924T within a composite(estimated) (C); carbon fibre Fibredux 924C composite (-) [7].


12.7 Thermo-mechanical response of resins

Watermolecules plasticise polar polymers such as epoxy and other advanced resins usedas matrices for fibre composites so that absorption of moisture can cause the glass tran-sition temperature,Tg, to be reduced significantly.This effect is illustrated inFigure 12.12[4] and therefore has an impact on the maximum service temperature available to a spe-cific composite system. A rule of thumb for epoxy resins is that Tg is reduced on average

100°C140°C200°C

As-curedIsothermal/ 50°C

0.4

0.2

0

Temperature (°C)

tan

δ

Temperature (°C)100 200 300

0.4

0.2

0

Tg3Tg3

Tg2

Tg1

Tg2

Tg1

100 200 300

(a) (b)

Figure 12.12 DMTA tan d spectra of a cured, untoughened Fibredux 924 base epoxy resinshowing the reduction of Tg with water absorption and the development of secondary peaks(Tg2, Tg3) on moisture absorption: (a) isothermally and (b) dynamically in the presence ofthermal excursions to 100, 140 and 200 �C [4].


by 20 K for each 1%ofmoisture absorbed [18]. Thus, a resinwith 7% absorbedmoistureat saturation can have its glass transition temperature reduced by 140 K.

A typical cured aerospace epoxy resin is highly cross-linked and may contain adispersed nonpolar thermoplastic or rubber which reduces the equilibrium moisturecontent and limits the reduction in Tg. As shown in Table 12.2, more advanced sys-tems, such as an epoxyecyanate ester blend, absorb lower concentrations of waterat equilibrium. However, the water is not uniformly distributed so that the epoxy phaseis still plasticised to the same extent. Figure 12.12 shows that the different componentepoxies and/or hardeners are plasticised differentially with the development of addi-tional relaxation peaks. We discussed the different degrees of water moleculehydrogen bonding as a function of network polarity in this chapter.

The maximum service temperature will also be affected since the resin modulus willtend to drop at the lowest relaxation temperature. Some modified systems also exhibit alarger reduction in Tg because of modifier hydrolysis, especially during thermal excur-sions. Therefore, an additional safety margin should be added to the highest servicetemperature. Since the Tg of a thermoset is a direct function of the cure or postcure tem-perature, a resin cured at 150 �C may have a long-term maximum useful temperatureof only 100 �C. This represents the design service temperature of the structure,although moisture equilibrium will be reached after several years.

The shear modulus of a rubber is inversely proportional to the average networkchain length according to the Gaussian theory of rubber elasticity [19]. Therefore,examination of the storage modulus ðE0

RÞ above the Tg provides the evidence fornetwork hydrolytic resistance. Figure 12.13 shows that the storage modulus ðE0

RÞ

Figure 12.13 The rubber storage modulus of two cyanate ester resins as a function of moistureabsorption with thermal spiking temperature illustrating the hydrolytical instability of thethermally cured AroCy L10 (C) compared to the stability of an epoxy-cured AroCy L10 (B) [7,19].


of a thermally cured cyanate ester, AroCy L10, falls as a result of thermal spikingat temperatures above 50 �C in humid environments. The rubbery modulus isnot affected by simple plasticisation. Therefore, this resin is clearly susceptible tohydrolysis. The choice of hardener is critical, as shown in Figure 12.13, wherethe thermally co-cured epoxyecyanate ester is shown to be hydrolyticallystable [9,20].

12.8 Effect of moisture on composite performance

12.8.1 Thermal stresses

Aerospace laminates are made from plies of embedded continuous fibres at differentorientations. This is because of the anisotropy of the individual plies. That is, the ther-mal and mechanical properties are strongly dependent of the orientation of the fibres tothe applied load. For example, the expansion coefficient parallel to the fibres (al) issignificantly less than the expansion coefficient transversely to the fibres (at) whilethe modulus of a longitudinal ply (El) is significantly higher than the modulus of atransverse ply (Et). Therefore, the matrix shrinkage at 90� to the fibres, on cooling,is constrained so that a tensile thermal stress will develop in the plies transverselyto the fibres while compressive stresses will be in the longitudinal or fibre direction.The stress-free temperature (T1) is also an important parameter which determinesthe magnitude of the residual thermal stress. It is a function of the matrix glass transi-tion temperature, and/or the cure or postcure temperature.


The tensile thermal strain in the longitudinal (l) direction of the transverse (t) ply(εthtl ) of a 0 /90 /0 laminate at temperature T2 is given by Eqn (12.9):

εthtl ¼ Elbðat � alÞðT1 � T2Þ

Elbþ Etd(12.9)

where b and d are the thicknesses of the 0� ply and semi-90� plies.In a 0�/90�/0� laminate, Eqn (12.9) applies to the inner ply, but the thermal strain in

the transverse direction of outer 0� plies will be slightly different and given by Eqn(12.10). The tensile thermal strain in the transverse (t) direction of the longitudinal(l) ply (εthlt ) is given by

εthlt ¼ Eldðat � alÞðT1 � T2Þ

Etbþ Eld(12.10)

The first failure of a composite, such as a simple 0�/90�/0� laminate usually occursby transverse or matrix cracking of the 90� or transverse ply. Since the residual thermalstress in this direction is tensile, the first ply failure will occur at a lower applied strain[21].

To balance the tensile thermal stresses, the plies will go into compression in thelongitudinal direction. Thus, the thermal strain in the longitudinal (l) direction of thelongitudinal (l) plies (εthll ) is given by

εthll ¼ �Et

Elεthtl (12.11)

Therefore, the 0� plies will fail at a slightly higher strain because of the induction ofa compressive stress into the fibres.

Furthermore, the thermal strain in the transverse (t) direction in the transverse (t) ply(εthtt ) is given by Eqn (12.12):

εthtt ¼ �Et

Elεthlt (12.12)

Under multidirectional stresses such as an impact load, the fibres in the transverseplies will fail at a higher resolved strain.

12.8.2 Thermal cracking

The failure strain of a unidirectional fibre composite in the transverse direction isnormally low because the matrix resins have relatively low failure strains while thelarge difference between the moduli of the components magnifies the strain in thematrix under stress. Thus, in an angle ply laminate, the first failure event occurs inthe transverse ply or plies. Reloading of the transverse ply via shear stress transferat the ply interfaces leads to multiple cracking before the fibres reach their failure


strain. Since the thermal strain in these plies is tensile, the first ply cracking of the lami-nate is less than that of an isolated 90� ply.

Equation (12.13) describes this observation:

εtlu ¼ εtu � εthtl (12.13)

Transverse multiple cracking will be initiated on cooling after cure when εthtl

exceeds the ply failure strain (εtu). This can occur when either the stress-free temper-ature (usually, the glass transition temperature of the matrix) and/or the matrix expan-sion coefficient are high in magnitude. Thermal cracking can occur for similar reasonswhen the properties of the matrix change as a result of a thermal excursion.

Thermal fatigue is the phenomenon whereby the material exhibits damage resultingfrom multiple excursions to differing thermal environments. The stresses will also bemultidirectional, and in a composite laminate the tensile stresses at 90� to the fibres inall plies will vary during the heating cycles. Provided that there is no change in theexpansion coefficients of the two plies, the material will be stable and transverse crackswill not form. However, if at is sensitive to temperature, there is a potential for thermalcracking at 90� to the fibres in all plies. Generally, thermal cycling of dry compositematerials to temperatures lower than the cure temperature does not normally lead tothermal cracking during the cooling cycle.

12.8.3 Effect of moisture absorption

The resin matrix in a composite absorbs water in service. As discussed in this chapter,this process is kinetically slow, so the concentration of water in the resin will vary withtime and location. Thermal cycling will tend to move the water into the structure.Furthermore, the absorption of the water causes the polymer to expand; however, ina composite, the swelling is constrained in an analogous way to that described for ther-mal contraction. Thus, the process of conditioning in moist environments leads to areduction in magnitude of the thermal strain present in the laminate.

The degree of swelling can be described quantitatively in analogous equations. Forexample, the swelling strain in the transverse ply in the longitudinal direction (εstl) isgiven by

εstl ¼

Elbðbt � blÞðM1 �M2ÞElbþ Etd

(12.14)

where M1 and M2 are the initial and final moisture concentrations, respectively.As a result, the thermal strain in a laminate composite will relax (i.e. reduced) on

moisture diffusion. The extent of moisture swelling by resins used for composites canvary significantly, so the effect on the residual stress state differs widely. However,the major problem is not the benefits associated with thermal stress relaxation, but thefact that the plasticisation of the matrix may enhance its expansion coefficient. There-fore, on cooling, a wet resin matrix either from its cure temperature or after a thermalcycle can, according to Eqn (12.9), lead to higher values of εthtl and ε

thlt . Figure 12.14

Moi

stur

e ab

sorp

tion

(%)

Ther

mal

stra

in (%

)

0

0.4

0.8

1.2

0.4

0.5

0.6

√time/thickness, √h/mm0 20 40 60

Figure 12.14 Effect of moisture absorption at 50 �C and 96% RH on the thermal strain in abalanced 0�/90�/0� 927C laminate estimated from the curvature of an unbalanced 0�/90�beam [3]. Control (C); 120 �C spike (,);140 �C spike (B) [7,22].


shows how the transverse thermal strain in the 90� ply is reduced on isothermal moistureabsorption and restrained matrix swelling, but in the presence of a thermal excursion thecombination of cooling from the elevated temperature and higher matrix expansion co-efficient results in a much lower effect of moisture absorption. For some matrices, ther-mal spiking can enhance the residual thermal strain and cause thermal cracking [22].

The higher degree of contraction of a wet cross-linked polymer (which will be con-strained) tends to arise from the appearance of a secondary relaxation peak or peaks inthe expansion coefficient temperature profile. The thermal response will mirror anychanges which occur in the DMTA spectrum. This may have its origin in the numberof different types of polar groups present in the cured polymer matrix and their mech-anism of interaction with water. Moisture absorption will decrease Tg and hence T1, butsince this is not usually exceeded in a thermal excursion its influence on the thermalstrain is limited. That means that during a thermal excursion, even to temperatures belowthe glass transition temperature or postcure temperature of the material, the higher con-strained contraction on subsequent cooling may induce a higher residual thermal strain.

In Figure 12.15 [23,24], the additional contraction, which manifests itself as an in-crease in residual thermal strain on cooling, is represented by the difference between theareas under the two curves. In the next cycle, moisture absorption may lead to a furtherswelling of the material and a reduction in the thermal strain. However, after a subse-quent thermal cycle, the thermal strain will be again enhanced. If the thermal stresseswhich are induced into the material exceed the transverse cracking strain of the individ-ual ply, then thermal cracks will form. Because of the 3D nature of the thermal stresses,

Figure 12.15 Temperature dependence of the transverse thermal expansion coefficients of abismaleimide modified epoxy resinebased carbon fibre composite (Narmco Rigidite 5245C):dry (continuous curve) and wet (dashed curve). The hatching illustrates the differing constrainedshrinkages (and hence thermal strain) for wet and dry laminates [23,24].


thermal cracking occurs in every ply. As shown in Figure 12.16, thermal cracking of theplies will occur differentially because the moisture absorption will not be uniformlydistributed throughout the material. In Figure 12.16, the thermal cracking occurredmore rapidly in the outer ply [25]. Matrix cracking during thermal cycling of PMR15 composites is complicated by the release of volatiles during cure, which becomeoccluded in the matrix. This causes the thermal expansion coefficients of the pliesand the stress free temperature to vary during thermal cycling because these volatilesslowly desorb over a number of themal cycles. As a result the composites exhibit the-mal cracking (or thermal fatigue) on cooling from each thermal excursion [26].

Figure 12.16 Thermal transverse cracking of a 0�/90�/0� laminate from PMR15 during ageingat 390 �C [25].


12.9 Fibre-dominated properties

Fibre choice is complex where cost often outweighs performance. For industrial com-posites, E glass therefore dominates. In aerospace, where cost can be spread over alonger life, often with savings in maintenance and service, carbon and other high-performance fibres are selected. For example, the high-modulus carbon fibres providethe highest dimensional stability of satellite dishes, where performance is critical.

12.9.1 Carbon fibres

Carbonfibres are commonly used in aerospace structures. There are various grades of car-bon fibre ranging in modulus between 220 and 490 GPa for those prepared from polyac-rylonitrile precursors. The in-plane theoretical modulus of graphite is 1000 GPa, so thatthe perfection of orientation of the graphite crystals to the fibre axis is responsible forthe lower practical values. As a result, the stressestrain curve of an individual filamentof high-strength carbon fibre (Efz 220e240 GPa) will show increasing modulus withstrain. From a strength perspective, there are no reports of static fatigue.

For pitch precursors, the refinement and spinning processes lead to larger crystalswith better alignment to the fibre axis and hence a higher modulus.

With respect to the effect of environments, carbon fibres can be largely consideredto be inert since they are stable to higher temperatures than the resin matrices can with-stand. At temperatures above 300 �C, the fibres begin to degrade in oxidising atmo-spheres, but most polymers have a lower maximum service temperature. Theadvanced polyimides and PMR (polymerisation of monomer reactants) systems cansurvive temperatures up to z450 �C, but these are usually in short-term applications[27]. It is also unlikely that composites will be in contact with damaging solvents suchas concentrated oxidising acids such as sulphuric and/or nitric acids.

The fibres are not resistant to the temperatures encountered in metal and ceramicmatrix composite melt processing, and protective coatings or low-temperaturemanufacturing routes are needed.

12.9.2 Advanced polymer fibres

Aerospace composite structures can also employ high-performance polymeric fibressuch as the aramids: Kevlar (DuPont) or Twaron (Teijin Twaron), poly(p-phenyl-ene-2,6-benzobisoxazole) (PBO) (Toyobo) and high-modulus polyethylene (PE)(Dynema, Certran and Spectra). Generally ultra-high-molecular-weight PE(UHMWPE) can be considered to be inert to most environments except that the servicetemperature is limited to <130 �C.

Aramid fibres are aromatic polyamides. Therefore, the presence of the polar andhydrolysable amide group introduces the potential for moisture absorption up to5%. As a result, these fibres exhibit time-dependent loss of strength or static fatiguein aqueous environments [28,29]. The fibres are spun from sulphuric acid solutionso that residuals might be responsible. These fibres are also highly susceptible to hy-drolysis in alkaline environments. As with other polymers, aramid fibres are


susceptible to oxidation, especially in the presence of ultraviolet (UV) light. Fortu-nately, the degradation products help to protect the fibres since they act as a UV screen.However, protective coatings should be used for ropes and cables. Furthermore, thematrix for composites needs to prevent moisture from reaching the fibres and to screenout UV rays. Van Dingenen [30] has examined strength retention under thesedemanding environments. For composite applications, the sizing of aramid fibreshas to provide chemical coupling between the fibre and the matrix for interfacial integ-rity as well as limiting moisture ingress.

PBO fibres may also be thermally sensitive, which has been attributed to acidiccontaminants from the spinning solvent [31].

12.9.3 Glass fibres

The strength of a glass or other brittle material is determined by the presence of Griffithflaws [32]. By drawing glass into fibres, the population of flaws of critical dimension isreduced. Therefore, there will be a statistical distribution of strengths which is commonlydescribed using the Weibull model. However, Bartenev [33] and Metcalfe and Schmitz[34] have shown that fibres have three populations of flaws of differing dimension whichare considered responsible for the length dependence of strength. Thus, commercialE-glass fibres typically have an average strength of 3 GPa with wide distribution as indi-cated by a Weibull modulus ofw4. Protective coatings (called ‘sizes’) are essential forprotection from damage during processing. The size also contains adhesion promoterswhich ensure durable interfaces. This is discussed in detail elsewhere [3,35].

Glass fibres exhibit time-dependent fracture under a static load, which is referred toas static fatigue [36]. Since there is no time dependence of strength in a vacuum, it isconsidered that water is involved in the reduction in strength. Thus, a stress corrosionmechanism in ‘condensed’ water is inferred. The chemical reactions associated withthis process are given in Figure 12.17.

Thus, the sodium ions present in E glass act as a catalyst for the degradation of thesilica network which is propagated by hydroxyl ions. The static fatigue of glass fibresinvolves three stages which operate at differing applied loads, according to the Charlesmechanism [37].

Figure 12.17 The mechanism of hydrolytic degradation of the silica network in E glass andrelated fibres, showing the catalytic role of Naþ and OH� ions.


Stage I is a stress-dominated region at high stresses where crack propagation is fastcompared to the rate of reaction with water (i.e. fibre corrosion at the crack or flaw tip).The rate of diffusion of sodium ions to the surface is rate determining, and the fractureis dominated by the mechanical loading.

Stage II is the stress corrosion region where the rate of corrosion is similar to therate of crack propagation. Therefore, the crack remains sharp and propagates intoweakened glass.

Stage III is the stress-assisted corrosion region where the effect of stress on the failuretime is much less significant because the rate of hydrolysis of the silica network is higherthan the rate of crack growth. The corrosive effect leads to a rounding of the crack tip.This reduces the potential for crack propagation according to the Irwin equation:

smax ¼ 2sa

�x

r

�12

(12.15)

where sa is the applied stress and smax is the stress at a crack tip of depth, x, and radius,r.

A typical bundle of E-glass fibres can have a typical lifetime of 106e107 min at anapplied strain of z0.5%. This provides a lifetime of 2e20 years. Therefore, providedthe fibres are protected from both mechanical damage and moisture diffusion, a com-posite could be expected to have a minimum lifetime, at 0.5% strain in the fibres, ofapproximately 20 years. From Eqn (12.1), one-sided exposure to water would extendthe life of a 10 mm thick composite to z120 years and two-sided diffusionz30 years. Loss of interfacial integrity will lead to a more rapid diffusion of the envi-ronment and will accelerate the degradation.

12.9.3.1 Thermal effects

The strength of E-glass composites is strongly affected by excursions to temperaturesabove 300 �C. Feih et al. [38,39] have studied these effects in the context of durabilityon exposure to fire. Figure 12.18 shows how the strength is retained up to approxi-mately 300 �C while at 350 �C a strong temperature dependence has been observed.

The most likely mechanism involves the reversible hydrolysis of the silica networkwhich is shown in Figures 12.19. Below w300 �C, surface silanol groups will tend tocondense to form siloxane bonds, whereas above this temperature hydrolysis domi-nates [35,38].

The change from strength stability below 300 �C can be attributed to the thermody-namics of the above reaction. As with any chemical reaction, the rate of hydrolysis is afunction of the presence of catalysts. Since the alkali metal glass modifiers(sodiumepotassium) catalyse hydrolysis, alkali metal-free glasses are expected toshow more resistance to high temperatures.

Recently, Owens Corning has replaced its E-glass fibres with boron-free formula-tions (Advantex). These fibres have similar mechanical properties to E glass but arebelieved to exhibit similar durability to ECR glass fibres which were developed forcorrosion-resistant applications.

0

20

40

60

80

100

120

Time (s)

Bun

dle

stre

ngth

(%)

650°C550°C

450°C

350°C

250°C

150°C

0 2000 4000 6000 8000

Figure 12.18 The effect of time at temperature on E-glass fibre bundle strength [35,38].

Figure 12.19 Mechanism of reversible hydrolysis of silica network in E glass and related fibres.Below 300 �C, the reaction moves to the right; but at z350 �C, hydrolytic degradationdominates [35].


12.9.3.2 Environmental stress corrosion cracking (ESCC) of glassfibres

E-glass fibres have a reduced lifetime under load in the presence of moisture, but this iseven more severe in a corrosive environment [40e42]. The rate of strength degrada-tion in non-loaded E-glass fibres is maximised at a pH close to zero (i.e. in diluteaqueous mineral acids). Only in highly alkaline environments (pHz 13) is a similarreduction in strength observed. Under load, ESCC of E-glass filaments also has amaximum degradation rate in an environment of pHz 0.2 (0.5 M H2SO4) [41].ESCC occurs in acidic environments because the network modifiers (Ca, Na, Al, Kand Fe) are leached from E glass to leave the silica network largely intact. Therefore,for the fibre to fracture, a crack needs to propagate through the weakened glass sheafwhich retains a covalently bonded structure.

In alkaline environments, hydrolysis of the silica network occurs sequentially withpH as shown in Figure 12.20 [43].

Thermodynamically, step 1 with equilibrium constant (K1) occurs at a pHz 10,and step 2 (K2) at pHz 12.5.

Figure 12.20 Hydrolysis and ionisation of a silica network at differing pH [43].


Therefore, at high pH, found with aqueous alkalis, a rapid hydrolysis of the silicanetwork occurs. Any cracks can be blunted through corrosive mechanisms or ion pre-cipitation. Therefore, weakening of the fibres is more important in alkaline environ-ments, and brittle ESCC fractures are not normally observed (except when stressconcentrations occur as a result of a crystallisation pressure at the boundary betweenparts of the composite with a degraded and non-degraded interface) [44].

ECR and boron-free glass fibres have improved durability over E glass so thatESCC of glass fibre composites is less likely when modern glasses are employed.The design of glass fibre composites for resistance to stress corrosion cracking hasbeen discussed in detail elsewhere [7,8].

12.10 Nonaqueous environments

Organic solvents are used in paints, in paint strippers and as degreasing agents, so theresistance of organic polymer-based composites to these materials should be dis-cussed. Cross-linked thermoset resins or crystalline thermoplastics do not dissolvein common organic solvents. However, the amorphous regions of the thermoplasticsand the cured thermosets used as composite matrices will swell after contact. The sol-vent will tend to dissolve (i.e. diffuse) into the polymer in an analogous mechanismto moisture diffusion but at a higher rate. The degree of swelling will be maximisedin contact fluids with a similar solubility parameter as the polymer matrix. Thesolubility parameters of the matrix and the solvent can be estimated using simpleadditive principles [45]. The concept of solubility parameters is simply describedby Eqn (12.16):

dsolvent ¼ dpolymer (12.16)

where dsolvent and dpolymer are the solubility parameters.The maximum plasticisation of the matrix will occur under these conditions which

will lead to a high potential to creep under off-axis loads. The residual stress state willalso be significantly modified, as discussed in this chapter.

Schulte [46,47] has demonstrated how different organic solvents, such as hydraulicfluid encountered in the aerospace structures, lead to a reduction in the secant modulusof �45� glass fibre laminate under flexural fatigue and the number of cycles to failure.The matrix in this case was a polyether imide (PEI) which is plasticised by ingress ofthe fluid. A reduction in the matrix modulus means that the shear strength of the matrixwill also be reduced with the consequence that the failure mechanism in flexure willchange from matrix-fracture to delamination.


Matrix swelling also contributes to the loss of durability in the 0� composite whereplasticisation causes buckling in the compressive face of the coupon [46,47]. This occursbecause in compression, the plasticised matrix is unable to support the reinforcement.

The reader should be reminded that PEI is an amorphous polymer and is much moresusceptible to solvents than partially crystalline polymers such as PEEK, which areused for structural composites.

Weatherhead [48] and Jones [7] provide a survey of the durability of resins to arange of environments.

12.11 Composite unidirectional properties

From the discussion in this chapter, it can be seen that degradation of the fibres willlead to a reduction of the 0� strength of a ply or composite. However, in order to reachthe fibres, diffusion of the environment through the matrix is anticipated. As describedin this chapter, this can be a slow process unless degradation of the interface hasoccurred which promotes ingress of the environment. Assuming that interfacial failurehas not occurred, then the interaction of the environment with the matrix is important.If chemical degradation occurs, it is probable that the matrix will be embrittled and pre-mature failure can occur, initiated by either matrix cracking or fibre breaks propagatinginto the degraded resin. Environmental stress corrosion cracking of glass fibre compos-ites often arises from low-strain fibre breaks propagating into the matrix. Good designoften uses ductile barrier resins in combination with the structural resin.

12.11.1 Tensile strength

Foreman and Behzadi [49e52] modelled the effect of matrix ductility on the unidirec-tional tensile strength of a carbon fibre composite. Figure 12.21 shows how ductility ofan epoxy resin leads to higher 0� strength. Interestingly, the best prediction is achievedwhen the viscoelasticity of the matrix is included in the model. Absorbed moisture andother solvents will increase the ductility of the matrix. The absorption of water or anorganic solvent can be expected to plasticise the resin matrix. The reduction of the Tgand/or generation of lower temperature relaxation peaks will tend to reduce the yieldstrength of the matrix. As a result, ingress of small quantities of water or other envi-ronments will provide stress redistribution mechanisms near a fibre break. The tensilestrength of the composite might increase. The complication arises from changes in theresidual stress state which will tend to negate the thermal compressive stress in thefibres and reduce the strain to first fibre fracture.

Furthermore, the absorption of water by the interphasal polymer can reduce its yieldstrength below the interfacial bond strength. Thus, the ‘apparent’ interfacial shear strengthwill be reduced, and a yield front, rather than a debond,will propagate along thefibre inter-face modifying the stress transfer micromechanics at a fibre break. A consequence is thatthe stress concentrations in adjacentfibres to the fibre breakwill be reduced, and the prob-ability of the formation of a flaw of critical dimensions is also reduced. The number ofinteracting fibre breaks associated with a flaw of critical dimensions will increase.

1.60

1.50

1.40

1.30

1.20

1.10

1.00

0.900.1 1 10 100 1000 10000

Log composite length (mm)

Com

posi

te fa

ilure

stra

in (%

)

Experimental failure strainfor carbon UD composites

Elastic matrixCurtis’ method

Viscoelastic matrix

Figure 12.21 Plot of the relationship between the predicted unidirectional (UD) carbon fibrecomposite failure strain and composite length. The data are presented for a viscoelasticTGAPeDDS matrix and a purely elastic matrix. The dotted line represents a typicalexperimental failure strain for a UD carbon fibre composite [49].


12.11.2 Compressive strength

In compression, the mechanism of failure changes from fibre fracture to fibre buckling.Here, the modulus of the matrix is a critical factor in supporting the fibre and by pre-vention of kinking (of the fibres) in a shear band. Soutis [53] has discussed the model-ling of these types of failure using Eqns (12.17) and (12.18):

s ¼sy

"1þ

�sTy

sy

�2

tan2 l

#12

f0 þ f(12.17)

where sy and sTy are the in-plane shear and transverse yield stresses of the composite,respectively. f0 is the assumed fibre misalignment angle in the kink band, f is theadditional fibre rotation in the kink band under a remote compressive stress s and l isthe band orientation angle.

sðgÞ ¼ sult

�1� Exp

�� G12g

sult

��(12.18)

where G12 is the elastic shear modulus, g is the shear strain and sult is the shearstrength.

Figure 12.22 shows how the compressive modulus of a TGAPeDDS epoxy resin isreduced by moisture absorption. At the highest strain rate, the compressive modulus


fell from 4.4 to 3.6 GPa on absorption of 7% water. It is also noticed that initialmoisture absorption is not so deleterious. This is because the unoccupied volume(sometimes referred to as ‘free volume’) will be filled first before plasticisation occurs.

The reduction in compressive modulus will have an effect on the compressive yieldstrength, as shown in Figure 12.23. The trend is different in that it is a continuous

190

180

170

160

150

140

130

120

110

100

900 1 2 3 4 5 6 7 8 9

Moisture conc. (%)

Com

pres

sive

yie

ld s

treng

th (M

Pa)

MY0510-A--1.67×10–4 s–1

1.67×10–3 s–1

1.67×10–2 s–1

Figure 12.23 The effect of water on the compressive yield strength of TGAP epoxy resin curedwith DDS. The experiments were conducted at different strain rates at 22 �C [54].

2800

3000

3200

3400

3600

3800

4000

4200

4400

4600

MY0510-A--1.67×10–4 s–1

1.67×10–3 s–1

1.67×10–2 s–1

Moisture conc. (%)

Com

pres

sive

mod

ulus

(MP

a)

0 1 2 3 4 5 6 7 8

Figure 12.22 The effect of water absorption on the compression modulus of TGAP epoxy resincured with DDS. The experiments were conducted at different three strain rates at 22 �C [54].

Table 12.4 Measured compressive strength properties of t800/924c unidirectional (0�) laminates at a range oftemperatures at 95% relative humidity [53]

Test temperature(�C)

Compressivestrength MPaExpt

Compressivestrength MPaprediction

Young’sModulus* (GPa)

Shear strength(MPa)

Shear yieldstress (MPa)

Shear modulusx

(GPa)

20-dry 1415 (1411) 160 110 40 6.0

20-wet 1060 (1040) — (89) (29.5) (5.4)

50-dry 1230 (1235) 155 105 35 5.8

50-wet 930 (917) — (78) (26) (5.4)

80-dry 1137 (1129) 149 98 32 5.4

80-wet 828 (829) — (69) (23) (4.9)

100-dry 973 (953) 136 90 28 4.9

100-wet 654 (653) — (54) (18.5) (4.5)

Moisture content, MN¼ 1.42%. Assumed initial fibre misalignment, f0¼ 1.75�; kink band inclination angle, l¼ 15�; the value of f0 is not affected by the environmental test conditions.( ) estimates from Eqns (12.17) and (12.18) and the measured unidirectional compressive or shear strength properties.*Secant axial modulus measured at 0.25% axial strain.xSecant shear modulus measured at 0.5% shear strain.

366Polym

erCom

positesin

theAerospace

Industry


decrease. This is most likely a result of the load state near the yield point when thewater molecules become mobilised. The compressive properties of unidirectionalcarbon fibre composites have been studied by Soutis [53], where the reduction in shearyield strength and compressive modulus causes fibre kinking to occur at lower stresses.Table 12.4 shows the measured and predicted compressive strengths of 0� carbon fibre(T800) Fiberdux 924 epoxy composites in the presence of absorbed moisture. Thisshows that absorbed moisture and/or possible organic cleaning solvents cause the dete-rioration of the compressive properties of composite materials.

The compressive properties of a composite under all loading conditions are stronglyaffected by moisture absorption because of the reduction in shear properties of thematrix polymer. The design of artefacts with polymer matrix composites needs toreflect the limitations of these materials in compression, especially in service whereenvironmental conditioning is likely.

12.12 Conclusions

This chapter reviews the state of knowledge regarding moisture absorption by aero-space composites in service. The discussion centres on the generic use of epoxy resinsas matrices for polymer matrix composites. Some reference to other matrices is alsomade but we have limited this in order for the main principles to be understood. Mois-ture ingresses into these resins relatively slowly so that the effects on the mechanicalproperties are complex. The moisture mainly influences the thermo-mechanical prop-erties of the matrix and so we concentrate on how thermal strains, tensile and compres-sive properties of the composites are affected.

In order to achieve these objectives, we discussed diffusion of moisture into thesematerials using Fickian kinetics and recognised that for composites based on glass orcarbon fibres, the changes in matrix properties are critical. Using Group InteractionModelling, the mechanisms responsible for these changes are confirmed. Of relevanceto this book is the point that for many aerospace epoxy resins, moisture absorption canlead to additional relaxation peaks in their thermo-mechanical response. As a result,the service temperature of polymer matrix composites will be reduced. Thus, thecompressive properties are strongly affected because the matrix carries the shearstresses which develop. Fortunately, the diffusivity is such that sufficient water isonly absorbed over many years.

One means of reducing the time scale of moisture absorption is to subject the struc-ture to rapid thermal excursions. Because the original structure of the cured epoxyresin is in a non-equilibrium state, thermal spiking can lead to more rapid moistureabsorption. Relaxation of the matrix structure will occur as a result of these moistureand thermal events. Other mechanisms which lead to enhanced moisture concentra-tions are discussed in the context of matrix resin and fibre selection.

The effect of moisture absorption on composite performance has been discussed interms of changes in thermal strain in cross-ply laminates and the introduction of


transverse cracks during thermal excursions. The nature of the failure of unidirectionalplies under tensile and compressive loads is also discussed in detail.

The conclusions of this chapter should be kept in mind when considering the failuremechanisms described in other parts of this book when selecting the matrix resins andreinforcing fibres for the design of durable composite structures.

References

[1] Nicolais L, Borzacchiello A, Lee SM, editors. Wiley encyclopedia of composites. 2nd ed.Hoboken, New Jersey: John Wiley & Sons; 2012.

[2] Middleton D. Composite materials in aircraft structures. Harlow: Longman; 1990.[3] Jones FR. The interphase in fibre composite materials. Wiley Encycl Compos 2012;3:

1396e402.[4] Hough JA. Enhanced moisture absorption in advanced composites [Ph.D. thesis].

University of Sheffield; 1997.[5] Springer GS. Environmental effects on composite materials, vol. 1. Westport, USA:

Technomic; 1981.[6] Shen CH, Springer GS. J Compos Mater 1976;10:2e20.[7] Jones FR. Durability of reinforced plastics in liquid environments. In: Pritchard G, editor.

Reinforced plastics durability. Cambridge: Woodhead; 1999. pp. 70e110.[8] Jones FR. The effects of aggressive environments on long-term behaviour. In: Harris B,

editor. Fatigue in composites. Cambridge: Woodhead; 2003. pp. 117e46.[9] Xiang ZD, Jones FR. Compos Sci Technol 1997;57:451.[10] Karad SK, Jones FR, Attwood D. Polymer 2002;43:5209e18.[11] Karad SK, Jones FR, Attwood D. Polymer 2002;42:5643e9.[12] Foreman JP, Jones FR, Porter D, Behzadi S. Polymer 2008;49(25):5588e95.[13] Foreman JP, Porter D, Behzadi S, Jones FR. Compos Part A 2010;41(9):1072e6.[14] Foreman JP, Porter D, Behzadi S, Travis KP, Jones FR. J Mater Sci 2006;41(20):

6631e8.[15] Mijovic J, Zhang H. Macromolecules 2003;36(4):1279e88.[16] Chen F, Birley AW. Plast Rub Compos Proc Appl 1991;15:161.[17] Chen F, Birley AW. Plast Rub Compos Proc Appl 1991;15:169.[18] Wright WW. Composites 1981;12:201.[19] Karad SK, Attwood D, Jones FR. Compos A 2002;33A:1665e75.[20] Treloar LRG. The Physics of rubber elasticity. 3rd ed. Oxford: Clarendon Press; 1975.[21] Bailey JE, Curtis PT, Parvizi A. Proc R Soc Lond 1979;A366:599.[22] Calota E, Jones FR. Unpublished data; 1997.[23] Jones FR. Handbook of polymer-fibre composites. Harlow: Longman; 1994. 379e391.[24] Jacobs PM, Jones FR. In: Tsai SW, Springer G, editors. Composite design, manu-

facture and application (ICCM VIII), vol. 2. Corina, CA, USA: SAMPE; 1991. Ch 16paper G. 8 pages.

[25] Xiang ZD, Jones FR. Effect of isothermal ageing on thermomechanical stability ofcarbon fibre reinforced PMF-15 resin matrix composites. In: Found MS, editor. Exper-imental techniques and design in composites materials. Sheffield Academic Press; 1995.pp. 406e20.

[26] Simpson PM, Jones FR. Composites 1991;22:105.[27] Xiang ZD, Jones FR. Comp Sci Technol 1993;47:209.

http://refhub.elsevier.com/B978-0-85709-523-7.00012-8/ref0010


















































[28] Howard A, Parratt NJ. Life prediction for aromatic polyamide reinforcements. In:Harrington WC, Strife J, Dhingra AK, editors. ICCM 5, Warrendale, PA, USA;1985. pp. 277e92.

[29] Horio M, Kaneda T, Ishikana S, Shimamura K. J Fibre Sci Technol Jpn 1984;40:1285.[30] van Dingegen JLJ. Gel spun high performance polyethylene fibres. In: Hearle JWS,

editor. High performance fibres. Cambridge: Woodhead; 2001. Ch 3, pp. 62e92.[31] Pegoretti A, Traina M. Liquid crystalline organic fibres and their mechanical behav-

iour. In: Bunsell AR, editor. Handbook of tensile properties of textile and technicalfibres, vol. 12. Cambridge: Woodhead; 2009. pp. 354e436.

[32] Griffiths AA. Phil Trans R Soc Lond 1920;A221:163.[33] Bartenev GM. The structure and mechanical properties of inorganic glasses. Groningen:

Walters-Noordhoff; 1970.[34] Metcalfe AG, Schmitz GK. Glass Technol 1972;13:5.[35] Huff NT, Jones FR. The structure and properties of glass fibres. In: Eichhorn SJ, et al.,

editors. Handbook of textile fibre structure. Natural, regenerated, inorganic and specialistfibres, vol. 2. Cambridge: Woodhead; 2009. pp. 307e52. 9.

[36] Thomas WF. Phys Chem Glass 1960;1:4e18.[37] Charles RJ. J Appl Phys 1958;29:1549.[38] Feih S, Mouritz AP, Mathys Z, Gibson AG. J Comp Mater 2007;41:2387e410.[39] Feih S, Manatpon K, Mathys Z, Gibson AG, Mouritz AP. J Mater Sci 2009;44:392e400.[40] Sekine H, Beaumont PWR. Compos Sci Technol 1998;58:1665e89.[41] Cockram DR. Glass Technol 1981;22:211e4.[42] Jones FR, Rock JW. J Mater Sci Lett 1983;2:415e8.[43] Fox PG. Mechanisms of environmentally sensitive cracking in glasses. In: Proc. congr.

on mechanisms of environmental stress cracking. London: Metals Soc; 1977.pp. 268e82.

[44] Sheard PA. Transverse and environmental cracking of glass fibre reinforced plastics[Ph. D. thesis]. University of Surrey; 1986.

[45] van Krevelen DW, Te Nijenhuis K. Properties of polymers. Elsevier; 2009. 4. s.l.[46] Schulte KL. Cyclic mechanical loading. In: Pritchard G, editor. Reinforced plastics

durability, vol. 5. Cambridge: Woodhead; 1999. pp. 151e85.[47] Schulte K, Mulkers A, Berg HD, Schoke H. Environmental influence on the fatigue

behaviour of amorphous glass/thermoplastic matrix composites. In: Degallaix S,Bathias C, Fongeres R, editors. International conf. on fatigue of composites; 1997.pp. 339e46.

[48] Weatherhead RG. FRP technology, fibre reinforced resin systems. London AppliedScience Publishers; 1980.

[49] Behzadi S, Curtis PT, Jones FR. Compos Sci Technol 2009;69:2421.[50] Foreman JP, Behzadi S, Porter D, Curtis PT, Jones FR. J Mater Sci 2008;43(20):

6642e50.[51] Foreman JP, Behzadi S, Tsampas SA, Porter D, Curtis PT, Jones FR. Plast Rubber

Compos 2009;38(2e4):67e71.[52] Foreman JP, Behzadi S, Porter D, Jones FR. Philos Mag 2010;90:4227e44.[53] Soutis C. Plast Rubber Compos 2009;38:55e60.[54] Behzadi S. Role of matrix yield in composite performance [Ph. D. thesis]. UK:

University of Sheffield; 2006.





























































Documents

Polymer Composites in the Aerospace Industry || The response of aerospace composites to temperature and humidity