Comparison between the thermal and microwave curing of bismaleimide resin

Comparison Between the Thermal and Microwave Curingof Bismaleimide Resin

Ismail Zainol,1 Richard Day,2 Frank Heatley3

1Advanced Materials Research Centre, SIRIM Berhad, Lot 34, Jalan Hi-Tech 2/3, Kulim Hi-Tech Park, 09000 Kulim,Kedah, Malaysia2Manchester Materials Science Centre, UMIST, Grosvenor Street, Manchester M1 7HS, United Kingdom3Department of Chemistry, University of Manchester, Oxford Road, Manchester M13 9PL, United Kingdom

Received 16 July 2002; accepted 2 April 2003

ABSTRACT: Microwave energy was used to cureMatrimid 5292A bismaleimide resin. The degree of cure atdifferent cure times was determined with differentialscanning calorimetry. A comparison was made with thesame resin cured in a conventional oven. The resin curedfaster in the microwave oven than in the conventionaloven. The glass-transition temperature (Tg) depended onthe degree of cure. The samples cured with a conventionaloven showed slightly higher Tg’s than the microwave-

cured samples at higher degrees of cure. At low degrees ofcure, the two heating methods yielded materials withsimilar Tg’s. Fourier transform infrared spectroscopy andsolid-state cross-polarity/magic-angle-spinning 13C NMRspectroscopy were used to investigate the cured struc-tures. There was no difference in the chemical reactionstaking place during the microwave cure and the thermalcure. © 2003 Wiley Periodicals, Inc. J Appl Polym Sci 90:2764 –2774, 2003

INTRODUCTION

The microwave processing of thermosetting materialsoffers significant advantages over conventional heat-ing, such as reduced cure cycles, more uniform curing,and improved physical and mechanical properties.1

Microwave heating takes place through dipolar inter-actions between molecules and an alternating electricfield rather than through thermal conduction or con-vection as in conventional heating. Microwave radia-tion penetrates materials, and the heat is generatedinternally. Studies on epoxy resin systems have shownthat microwave processing significantly reduces thecure cycle without sacrificing the mechanical proper-ties.2,3

Bismaleimide (BMI) resins are presently the bestcandidates to replace epoxy resins in high-perfor-mance composites because of their high temperaturestability and resistance to hot and wet environments.4

BMI resins contain reactive double bonds (CAC) inthe maleimide groups that can polymerize with orwithout the presence of a catalyst. Various curingreactions for BMI resin formulations have been sug-gested.5 The main reaction proposed is homopolymer-ization via a free-radical mechanism.6 Conventionalcuring processes involve heating in an oven, auto-clave, or compression mold, in which heat transfer to

the material is accomplished through conduction orconvection. Just as for other thermosetting materials,the low thermal conductivity of these resins leads tothermal gradients inside the resin, and the curingcycle becomes longer.

Initial studies of the microwave processing of BMIsystems have been reported.7,8 Researchers8 foundthat the reaction kinetics in conventional and micro-wave ovens were identical under isothermal condi-tions. They concluded that no microwave effect couldbe observed. The term microwave effect implies thepossibility of altering the rate and path of a chemicalreaction to produce a different morphology and dif-ferent physical and mechanical properties. The afore-mentioned researchers used in situ fiber-optic Fouriertransform near-infrared spectroscopy to measure theextent of the reaction. Liptak et al.7 however, foundmore rapid curing when microwave heating was usedthan when a conventional oven was used. They usedextraction methods to determine the gel content in thecured resin. The apparent contradiction between thesetwo reports is probably due to the different techniquesused to cure and characterize the samples. Neithershows in detail the cure chemistry in microwave andconventional ovens. It is important to study the curingpath because the final structure and mechanical prop-erties of a network depend on the reaction path.

The objective of this study was to compare the curechemistry of a BMI resin when conventional and mi-crowave heating were used. The cure reactions werefollowed with spectroscopic techniques. The effect of

Correspondence to: I. Zainol ([email protected]).Contract grant sponsor: SIRIM Berhad.

Journal of Applied Polymer Science, Vol. 90, 2764–2774 (2003)© 2003 Wiley Periodicals, Inc.

the sample size on the reaction rate in conventionaland microwave ovens was also studied.

EXPERIMENTAL

Material

The BMI resin used was Matrimid 5292A (Ciba–Ge-igy). The chemical structure is shown in Figure 1.9 Thematerial was used as received.

Curing procedures

Microwave processing

The sample was cured in a 60-mm-diameter, 10-mm-high Pyrex glass dish. A Moulinex Quickchef 1000QEmicrowave oven with a glass turntable and adjustablepower settings was used. Initially, each sample wasexposed to microwave radiation by being placed inthe middle of the microwave cavity at the full-powersetting (635 W). The samples were heated for differentperiods of time ranging from 6 to 11 min. Two differ-ent sample weights, 2.5 and 5.0 g, were used in thisstudy. Part of each sample boiled after 10–12 min ofheating. The curing procedure was then optimized toobtain a high degree of cure with few bubbles. Thiswas achieved through the control of the microwavepower setting and the placement of the sample wherethe microwave field was uniform.

The uniformity of the microwave power distribu-tion within the cavity was determined by the heatingof an epoxy slab and the observation of the variationsin color from point to point. The microwave field wasfound to be uniform at the edge of the glass turntable.The optimized cure cycle for the microwave cure wasas follows: first, each sample was heated at an averagepower of 473 W for 20 min, and then it was postcuredat full power (635 W) for 3 min

Conventional oven processing

Conventional thermal curing was carried out in acirculating air oven (model 9-300, Townson and Mer-cer) at various times ranging from 15 min to 24 h. Thesamples were placed in the oven at room temperaturebefore the oven was turned on and set to the requiredtemperature. This step allowed the samples to heat upgradually and prevented exotherms and bubbles. Theheating rates of the oven were as follows: from roomtemperature (25°C) to 190°C at 9°C min�1 and from

190 to 210°C and from 210 to 250°C at 4°C min�1. Forthe study of the effect of the sample size, two differentsample weights, 3.0 and 5.0 g, were cured isothermallyat 190°C. For the study of the effect of the cure tem-perature, samples weighing 5 g were cured at 170, 180,or 190°C. Fully cured samples were prepared with themanufacturer’s recommended procedure:10 2 h at190°C, 5 h at 210°C, and 5 h at 250°C.

Characterization techniques

Differential scanning calorimetry (DSC)

The thermal properties of the resins were investigatedwith a DuPont 900 instrument. The DSC cell was firstcalibrated with high-purity indium. The samples wereheated from 25 to 400°C at a heating rate of 10°Cmin�1 in flowing nitrogen at 40 mL min�1. The panswere closed but not hermetically sealed, and the sam-ple weights were 8–12 mg. The samples were takenfrom the middle of the glass mold.

Glass-transition temperature (Tg)

DSC was used to determine Tg for samples with de-grees of cure (�DSC) below 0.7. A Polymer Laboratoriesdynamic mechanical thermal analysis (DMTA) instru-ment was used to determine Tg’s for samples with�DSC � 0.7. For the latter measurements, the sampleswere cut into rectangular bars (1.3 mm � 6.2 mm� 12.0 mm) and clamped rigidly from the fixed clampto the driven clamp in a single-cantilever mode. Thesample was heated from 10 to 300°C at a heating rateof 5°C min�1 and a frequency of 1 Hz. Tg was taken tobe the maximum in tan �.

Structural analysis

The chemical structures of the cured samples wereinvestigated with Fourier transform infrared (FTIR)spectroscopy and solid-state 13C NMR spectroscopy.FTIR spectra were obtained with a PerkinElmer 2000spectrometer. Each sample was pressed into a discwith dried potassium bromide (KBr) powder at a con-centration of 1 wt %. Each spectrum was scanned from4000 to 450 cm�1 32 times at a 2-cm�1 resolution, andthe spectra were coadded. Solid-state cross-polarity/magic-angle-spinning (CP-MAS) 13C NMR spectrawere run on a Varian Associates Unity spectrometeroperating at 75.4 MHz and equipped with a DotyScientific 7-mm CP-MAS probe. Powdered sampleswere packed into 7-mm zirconia or alumina rotorsfitted with Kel-F end caps. The contact time was 2.5ms, and the recycle time was 3 s. Approximately 500cycles were time-averaged.

Figure 1 Chemical structure of Matrimid 5292A.

THERMAL AND MICROWAVE CURING OF BISMALEIMIDE 2765

Sol–gel analysis

Residual monomer and other low-molecular-weightnoncrosslinked materials were studied with sol–gelanalysis. The analysis was carried out by the soakingof powdered samples in dichloromethane in a hotwater bath at 50°C for 2 weeks. The insoluble portionwas filtered with a glass-fiber filter, dried, andweighed.

RESULTS AND DISCUSSION

DSC analysis of the resin

A DSC curve obtained from uncured Matrimid 5292Ais shown in Figure 2(a). The DSC curve shows a strongand sharp endothermic peak at 157°C that is immedi-ately followed by an exothermic peak. The endother-mic peak is due to the melting of crystalline regions inthe resin.11 The exothermic region is due to the curingreaction during the heating process and has a maxi-mum at about 202°C. The area under the exothermicpeak [or the total enthalpy of the curing reaction for anuncured sample (�Hunc)], between 160 and 280°C, is183.3 J g�1. There is no Tg in the DSC curve for theuncured resin, and this can be attributed to the factthat the resin is a crystalline material.

�DSC has been calculated with the following equa-tion:12

�DSC � 1 ��Ht

�Hunc(1)

where �Ht is the residual enthalpy of the reaction of asample cured for time t.

Figure 2(a) shows the DSC curve of a partially curedsample (�DSC � 0.18). Tg can be clearly observed at55°C. An exothermic peak appears between 90 and150°C as a result of cold crystallization during the DSCscan. Wang et al.13 reported a similar observationwhen they studied another semicrystalline polyimide.This peak gradually decreases as curing proceeds andis not present in samples with �DSC � 0.45. The endo-thermic peak shifts from 157 (for the uncured resin) to166°C.

Curing rate

Figure 3(a) shows �DSC for 2.5- and 5.0-g samples aftermicrowave curing as a function of the curing time. Thevalue of �DSC increases with the curing time. The 5.0-gsamples cure rapidly between 7 and 8 min, the valueof �DSC increasing from 0.40 to 0.80 over this timeinterval. The value of �DSC then increases gradually to0.98 after an 11-min cure time. The 2.5-g sample showsa slower curing rate than the 5.0-g sample, and thissuggests that the heating rate is affected by the sampleweight. The reason may be that the amount of micro-

Figure 2 Typical DSC curves of Matrimid 5292A: (a) un-cured and (b) partially cured (�DSC � 0.18).

Figure 3 Effect of the sample weight on �DSC as a functionof time for the microwave curing of Matrimid 5292A aftercuring optimization.

2766 ZAINOL, DAY, AND HEATLEY

wave radiation absorbed depends on the dielectricloss of a sample. The larger the sample volume is, thegreater the absorption will be of microwave energy ofthe sample in comparison with the cavity, which has afixed loss. Another reason is the slower rate of heatloss from the surface of a thick sample in comparisonwith a thin sample, particularly for BMI resins, whichhave very poor thermal conductivity.14 Similar effectswere observed by Demeuse and Johnson15 when theystudied the effect of the sample volume on the heatingrate with variable-frequency microwave heating. Theyfound that the thicker isocyanate/epoxy samplesheated faster than the thinner samples. Theory alsoshows that the temperature at the center of a volu-metrically heated slab will depend on the square of thethickness.

The value of �DSC as a function of time after curecycle optimization is shown in Figure 3(b). Sampleswith high �DSC values (0.98) and few bubbles havebeen obtained. Some of each sample boils after 10–12min of heating at full power because of the hot-spoteffect. Nonuniform heating is common in multimodemicrowave cavities because of the variation of themicrowave radiation with the position and becausethe outer layer cools more easily. Nightingale16 usedthe same microwave oven for her studies and, using asheet of partially cured epoxy resin, showed that thecenter part of the cavity was the area at which themicrowave energy was most concentrated.

The value of �DSC as a function of time at 170, 180,and 190°C is presented in Figure 4(a). The rate of cureis highest at 190°C and lowest at 170°C, as expected.Matrimid 5292A resin contains only one type of mono-mer and thus undergoes homopolymerization afterheat treatment.17 The reactive species involved in thereaction are maleimide groups (CAC), and they reactwith one another through free-radical polymeriza-tion.18 Generally, the curing process for the resin canbe divided into four stages. The first stage is the melt-ing process. In the microwave oven, the resin melts inonly 6 min, whereas 15 min is needed in the conven-tional oven. As shown in the DSC curve [Fig. 2(a)], theas-received resin melts at 155°C.

The second stage is polymerization, which is shownas an exothermic peak in the DSC curve [Fig. 2(a)]. Theresin starts to polymerize immediately after melting.The polymerization process involves free-radical ad-dition reactions. According to Brown and Sandreczki,6

in the initial step, biradicals are formed as a result ofthe thermal homolysis of the double bond in the ma-leimide groups. The biradicals then react with othermaleimide groups to form propagation radicals. Thepropagation radicals are very reactive and will reactwith other maleimide units to form bigger molecules.The reaction rate depends on the mobility of these freeradicals. The mobility of reactive sites is highly influ-enced by the viscosity of the medium and the degree

of cure.11 In microwave heating, a 5.0-g sampleachieves a degree of cure of 0.6 after 7 min, whereas inthe conventional oven, the resin does not reach thisdegree of cure until 15 min.

The third stage shows a slower curing rate than thesecond stage for both methods. The viscosity of thesystem increases very rapidly as the degree of cureincreases. As a result, the cure reaction is slow becausethe mobility of the reactive sites is restricted.19 In thisstage, the microwave-cured samples achieve an �DSCvalue of 0.96 in 11 min, whereas in the conventionaloven, it is only 0.87 after 420 min. The cure reactions inthis stage depend on the temperature: the higher thetemperature, the higher the degree of cure at a giventime. Thus, a sample cured with conventional heatingat 190°C shows a higher degree of cure than one curedat 170°C. This is expected because at higher tempera-tures, the reactive sites are more mobile than at lowertemperatures, thus giving higher degrees of cure.

In the final stage, the cure reactions totally stop eventhough the heating is continued. This is because thenetwork has become highly crosslinked; the mobilityof reactive sites is totally restricted, and they aretrapped in the matrix.20 As shown in Figure 4(a), in theconventionally cured material, no further reaction canbe observed even when the curing time is prolonged

Figure 4 (a) Effect of temperature on �DSC as a function oftime for 5.0 g of Matrimid 5292A cured in a conventionaloven at different temperatures and (b) the effect of thesample weight at 190°C.


to 24 h. The temperature at this stage is known as thevitrification temperature.

Figure 4(b) shows the effect of the sample weight onthe degree of cure as a function of time when samplesare heated conventionally at 190°C. This figure showsa slower reaction rate in a 5.0-g sample than in a 3.0-gsample in the initial stages up to a 120-min heatingtime. This is because a longer time is required to melta bigger sample (5.0 g) in the first stage on account ofslow heat transfer from the oven into the material. Thecuring rate is higher after 120 min in the second stagefor a 5.0-g sample than for 3.0-g sample. In this stage,the curing rate is affected by the exothermic heat fromthe polymerization process. The magnitude of heat

released is proportional to the size of the sample. Moreheat is produced in a 5.0-g sample than in a 3.0-gsample, and so the reaction rate is higher in a 5.0-gsample because the heat is lost from this more slowly.

Tg

Figure 5 shows Tg as a function of �DSC for bothmicrowave-cured and conventionally cured samples.There is only a small increase in Tg from 48 to 57°C asthe degree of cure increases from 0 to 0.7. This isprobably because the molecular segments begin topolymerize19 and there is little crosslinking. Tg of apolymer is very much influenced by its molecularweight and increases proportionally with increasingmolecular weight.21 There is no significant differencein Tg at this stage for the microwave-cured and con-ventionally cured samples, and this indicates that themorphology of samples at this stage is similar withboth curing methods. This suggests that the cure re-action is the same in this region. After the degree ofcure reaches 0.70, DMTA must be used because theDSC technique is not sufficiently sensitive to detect Tg.A typical DMTA curve is shown in Figure 6. Tg risesrapidly when �DSC is greater than 0.70 in both curingmethods. The increase in Tg suggests that the mole-cules become more crosslinked and that the move-ment of the polymer chains is restricted as the curingprogresses.11 Microwave-cured samples show lowerTg’s than conventionally cured samples for values of�DSC between 0.7 and 0.9. The difference in Tg may be

Figure 5 Tg as a function of �DSC, as determined by DSCand DMTA, for 5.0 g of Matrimid 5292A cured by a conven-tional oven (OV) at 190°C or a microwave oven (MW) at a635-W power setting.

Figure 6 Typical DMTA curves for microwave-cured Matrimid 5292A (�DSC � 0.93).


attributed to the variation of the crosslinking networkformed at higher degrees of cure. Alternatively, thedifferent values may be a result of different coolingrates. When the degree of cure approaches 100%, how-ever, both cure methods show similar Tg’s. Tg isknown to be strongly dependent on the crosslink den-sity.20 To estimate the structural variation of samplescured in conventional and microwave ovens, the mo-lecular weight between crosslinks (Mc) can be calcu-lated from the corresponding Tg according to the fol-lowing empirical equation:22

Mc �3.9�104

Tg � Tgo(2)

where Tgo is the glass-transition temperature of thenoncrosslinked polymer. Tgo is 48°C, according to aDSC curve of a resolidified sample. From this equa-tion, Mc has been calculated for conventionally curedand microwave-cured samples for �DSC � 0.7; theresults are presented in Figure 7. The values of Mc forthe microwave-cured samples are generally higherthan those for the conventionally cured samples, butthe values merge as the degree of cure approaches100%. These results show that samples cured in a

conventional oven have a higher crosslink densitythan the microwave-cured samples in the region �DSC� 0.7–0.9, but the fully cured samples are similar. It isvery important to confirm the cure reactions becausethe two heating methods give different values of Tg

and Mc below �DSC � 1, and this gives the impressionthat the reaction scheme might be different. This hasbeen done by the determination of the structures ofsamples as the curing progresses with spectroscopictechniques.

Structure of the cured resin

BMI resins contain a reactive double bond(HOCACOH) in the maleimide groups that can un-dergo homopolymerization via a free-radical additionreaction.5 The proposed reaction scheme is shown inFigure 8.6 The reactions show that the maleimidegroup (I) is converted into the succinimide group (III)as a final product via an intermediate (II). FTIR11 andNMR23 spectroscopy have been used to follow thisreaction in conventional and the microwave ovens.

FTIR spectroscopy

Figure 9 shows the overlaid FTIR spectra of uncuredsamples and conventionally cured and microwave-cured samples at different degrees of cure. The detailsof the uncured spectrum interpretation have been re-ported.24 Several spectral changes can be observedafter polymerization:

• The disappearance of the peak at 3100 cm�1 char-acteristic of the COH stretch in maleimide groups(HOCACOH). These groups have been con-verted into succinimide groups (HOCOCOH),the COH stretch of which shows as a broad peakcentered at 2915 cm�1. This peak overlaps withthat of the COH stretch in methylene groups.

• The disappearance of the peak centered at 1150cm�1 (CONOC, maleimide) and the growth of apeak centered at 1185 cm�1 (CONOC, succinim-

Figure 7 Calculated Mc values as a function of �DSC (�0.7)for samples cured with a conventional oven (OV) or a mi-crowave oven (MW).

Figure 8 Homopolymerization reaction scheme of Matrimid 5292A.


ide). These peaks strongly overlap, but the dou-blet structure is perceptible.

• The disappearance of the peak at 690 cm�1 char-acteristic of maleimide groups (CACOH).

The peak at 1709 cm�1 (maleimide CAO) is broad-ened as a result of overlap with a slightly differentpeak due to the formation of succinimide CAO.

The peak changes at 3100, 1150, and 690 cm�1 havebeen used by several workers to follow the cure reac-tion in BMI polymerization.23,24 Here, the peak heightchanges at 3100 and 1185 cm�1 have been used tofollow the cure reaction. The peak at 3100 cm�1 hasbeen chosen as it is well resolved from other peaksand suitable for quantitative analysis. It is difficult toanalyze this peak in the spectrum of the uncured resinbecause it is split on account of the high crystallinity ofthe resin. This problem was overcome by the remelt-ing of the resin at a temperature low enough to avoidcuring before the spectrum was obtained. The forma-tion of succinimide molecules was followed by anexamination of the peak changes at 1185 cm�1. Eventhough this peak overlaps with the peak at 1150 cm�1,

this is acceptable for the purpose of comparative anal-ysis. The peaks at 3038 and 831 cm�1, characteristic ofCOH in benzene, are unchanged throughout the pro-cess. The peak at 3038 cm�1 has been chosen as aninternal reference.

Figure 10(a,b) shows the peak height changes at3100 and 1185 cm�1 for conventionally cured andmicrowave-cured samples as a function of �DSC. Thepeak heights have been normalized to the peak heightat 3038 cm�1 (COH, benzene ring) to eliminate vari-ations in the sample thickness and concentration. Bothfigures show that as the curing progresses, the peakdecrease at 3100 cm�1 is proportional to the peakincrease at 1185 cm�1. This suggests that a similarreaction occurs in both curing methods. Close obser-vation shows that the patterns of the peak changes at3100 and 1185 cm�1 are slightly different for the twocure methods. The peak height at 3100 cm�1 is slightlylower in a conventionally cured sample than in amicrowave-cured sample at the same degree of cure.This shows that the concentration of the maleimidegroups is higher in the microwave-cured sample. Thisis probably due to residual reactive groups trapped in

Figure 9 FTIR spectra of Matrimid 5292A: (a) uncured, (b,d) conventionally cured (�DSC � 0.3 or 1.0, respectively), and (c,e)microwave-cured (�DSC � 0.35 or 0.98, respectively).


the microwave-cured sample, which cannot be de-tected by DSC.

The peak changes at 1185 cm�1 are significantlydifferent in the two curing methods. Initially (�DSC� 0.2), conventionally cured samples show a lowerpeak height than microwave-cured samples, but theheights are higher when �DSC is greater than 0.2. Thepeak height different is maximum when �DSC is be-tween 0.4 and 0.7. This indicates that lower succinim-ide groups are present in the microwave-cured sam-ples with this cure state. This suggests that a largeramount of intermediate species (structure II in Fig. 8)remains in the matrix of the microwave-cured sam-ples. Fast microwave heating results in immediatechanges of maleimide moieties into radicals (II); how-ever, because of the comparatively low temperature,this species does not immediately transform into suc-cinimide molecules. As the curing proceeds, the tem-perature increases, and more succinimide groups areformed. This is shown in Figure 10(b): the peak heightat 1185 cm�1 increases rapidly when �DSC is greaterthan 0.7. Figure 10(a) also shows that microwave-cured samples have a slightly higher number of resid-ual maleimide groups than conventionally cured sam-ples. This is due to more reactive groups being

trapped in the network during microwave heating asa result of fast curing at an earlier stage. The details ofthe chemical structure have also been studied withNMR.

FTIR spectroscopy has been used to study the extentof the reaction.25 The peak height ratios from Figure10(a,b) have been used to calculate the degree of curewith the following formula:

�FTIR � 1 �Rt

Runc(3)

where Runc and Rt are the peak height ratios at 3100cm�1 of an uncured sample and a sample cured fortime t, respectively.

NMR analysis

Solid-state 13C-NMR is a technique complementary toinfrared spectroscopy in the determination of thechemical structures of materials. Figure 11(a) showsthe NMR spectra of a sample of the uncured resinobtained with solid-state 13C-NMR. Resonance signalsbetween 30 and 40 ppm are characteristic of CH2groups of the methylene bridge between the two aro-matic rings.23 Between 112 and 144 ppm, the signalcan be attributed to unsaturated carbons in aromaticrings and in alkene groups in the maleimide ring. Thesignal between 160 and 180 ppm is due to the carbonof the carbonyl groups of the two maleimide rings.NMR spectra from conventionally cured samples with�DSC � 0.56 and �DSC � 1.0 are shown in Figure11(b,c), respectively, and spectra of microwave-curedsamples with �DSC � 0.50 and �DSC � 0.95 are shownin Figures 11(d,e), respectively. Two peaks at 30–50and 170 ppm appear as the curing progresses. The30–50 ppm peak is attributed to the sp3 carbon ofsuccinimide groups (OCHOCHO), and the peak at170 ppm is characteristic of carbonyl groups from thesuccinimide environment23 (Fig. 8). The intensity ofthese peaks increases as the curing proceeds in bothcure methods. Basically, both conventionally curedand microwave-cured samples show similar NMRspectra, and this suggests that the same curing reac-tions take place in each case. The NMR results are inagreement with the FTIR data, which show that thechemical structures of both conventionally cured andmicrowave-cured samples are the same. The differ-ence in the peak intensity in the NMR spectra is due tothe residual monomer present in the polymer matrix.The peak area at 160 ppm (CAO maleimide) is largerin the microwave-cured sample than in the conven-tionally cured sample. This may be because more re-sidual monomer is trapped in the polymer matrix inthe microwave-cured sample.

Figure 10 Peak height changes as a function of �DSC for IRpeaks at 3100 (CAC, maleimide) and 1181 cm�1 (CONOC,succinimide): (a) conventionally cured samples (5.0 g at190°C) and (b) microwave-cured samples (5.0 g at a 635-Wpower setting).


In addition to structural identification, NMR has alsobeen used to calculate the extent of the reaction througha comparison of the peak areas at 166 (CAO maleimide)and 172 ppm (CAO succinimide) with a componentanalysis technique; an example of this analysis is shown

in Figure 12. The degree of cure from NMR (�NMR) canbe calculated with the following equation:

�NMR �A172

A172 � A166(4)

Figure 11 Solid-state CP-MAS 13C NMR spectra of Matrimid 5292A: (a) uncured, (b,c) conventionally cured (�DSC � 0.56 or0.98, respectively), and (d,e) microwave-cured (�DSC � 0.50 or 0.95, respectively). Asterisks denotes sidebands from chemicalshielding anisotropy.


where A166 is the peak area at 166 ppm and A172 is thepeak area at 172 ppm.

Comparison of �DSC, �FTIR (the degree of curefrom FTIR), and �NMR

The degrees of cure calculated with eqs. (2)–(4) aresummarized in Table I. The results show that thedegrees of cure by the DSC, FTIR, and NMR tech-niques are in poor agreement with one another. TheDSC analysis gives the highest degree of cure,whereas the FTIR analysis shows the lowest. The dis-crepancies in the results are probably due to differentmethods being used to measure the concentrations ofunreacted groups. For example, in DSC, a sample isheated in a small oven or DSC cell, and the area underthe exothermic peak (�H), characteristic of the poly-merization process, is measured. The magnitude of�H relies on the ability of DSC to record all the heatreleased from the remaining reactive groups of themonomer during polymerization. This assumes thatall unreacted groups are totally reacted after the DSCscan. However, some of the reactive species actuallyremain trapped and are unable to react. As a result,the measured �H value becomes smaller than theactual value, and this gives a larger degree of curewhen applied to eq. (1). Prime26 reported that thesensitivity of DSC decreases as the degree of cureincreases.

In FTIR analysis, the concentration of HOCAC ofmaleimide groups is directly measured in the polymermatrix on the basis of Beer’s law. The peak height at3100 cm�1 is proportional to the concentration of ma-leimide groups (HOCAC). The �FTIR results are lowerthan �DSC because more HOCAC groups are detectedby FTIR than by DSC. FTIR spectra show the presenceof HOCAC groups even for a sample with �DSC � 1.This is probably due to residual monomers or end

groups of polymer chains. In FTIR analysis, however,there are several sources of error. Scattered infraredradiation makes the direct application of Beer’s lawinaccurate. Another source of error is the difficulty ofchoosing the correct base line, and this affects theaccuracy of the data.

In the solid-state 13C-NMR technique, the actualnumber of maleimide and succinimide carbonylgroups (CAO) is measured directly, similarly to FTIR,but it does not depend on Beer’s law. Even thoughthese peaks overlap one another, the peak area iscalculated separately with the component analysistechnique. The results are expected to be more accu-rate than those of FTIR. They show that the degree ofcure calculated with NMR is closer to that found withthe DSC technique. As shown in the NMR spectra inFigure 11(c), the amount of CAO, maleimide is muchgreater in the microwave-cured sample. It is proposedthat this carbonyl group is probably from a residualmonomer trapped in the matrix or end groups of thepolymeric chain. To investigate this assumption, wehave conducted a sol–gel analysis.

Sol–gel analysis

The results of a sol–gel analysis for conventionallycured and microwave-cured samples are shown inFigure 13. The conventionally cured samples weigh5.0 g and are cured at 190°C, whereas the microwave-cured samples weigh 5.0 g and are cured at the full-power setting (635 W). Interestingly, the results showa linear relation between the insoluble percentage and�DSC. Microwave-cured samples display slightlyhigher amounts of the soluble monomer than the con-ventionally cured samples, and this means that moreresidual monomer is trapped in microwave-curedsamples. This result agrees with NMR results but doesnot support the Tg results at a lower value of �DSC(�0.7). This is probably because the amount of theresidual monomer present in the polymer matrix doesnot significantly affect the Tg value at lower degrees ofcure for both curing methods. When �DSC is greaterthan 0.7, the microwave-cured samples show slightlygreater amounts of soluble material. This supports the

TABLE IComparison of the Degree of Cure Determined

by DSC, FTIR, and NMR

Conventional oven(5.0 g at 190°C)

Microwave oven(5.0 g at 635 W)

�DSC �FTIR �NMR �DSC �FTIR �NMR

0.56 0.47 0.50 0.330.66 0.39 0.65 0.370.81 0.47 0.85 0.500.98 0.86 0.95 0.771.0 0.66 0.98 0.67

Figure 12 Component analysis of the NMR spectrum.


argument that the lower Tg value in microwave-curedsamples is due to more residual monomer and a lowercrosslink density. It has been reported20 that fast cur-ing when microwave heating is used can result inunreacted monomer being trapped in the matrix of anepoxy system. However, when �DSC approaches 1.0,the percentage of insoluble material in samples curedby both heating methods becomes similar. This sug-gests that the trapped monomer eventually reacts asthe heating proceeds in the microwave oven. Thisexplains why the Tg values for both methods convergeat 100% cure.

CONCLUSIONS

Microwave processing significantly reduces the curecycle for Matrimid 5292A. Spectroscopy has revealedthat the cure chemistry of both conventional oven andmicrowave processing is basically the same. Rapidcuring in a microwave oven does not occur becausethe reactions that take place have been altered. This isprobably due to volumetric heating that takes placewhen microwave heating is used. It is important tocontrol the microwave power, as it affects the sample

temperature and subsequently the morphology of thefinal sample. Through control of the microwavepower, samples with high degrees of cure and fewbubbles can be obtained. Samples cured with a con-ventional oven have a higher crosslink density thanmicrowave-cured samples up to �DSC � 0.95, but Mc

converges at 100% cure. The fast curing that occursduring microwave heating leads to some monomersbeing initially trapped in the matrix, but these reactlater in the cure cycle.

References

1. Mijovic, J.; Wijaya, J. Macromolecules 1990, 23, 3671.2. Boey, F. Y. C.; Lee, T. H. Radiat Phys Chem 1991, 38, 419.3. Wei, J.; Hawley, M. C. Int SAMPE Symp 1994, 39, 45.4. Stenzenberger, H. D.; Romer, W.; Herzog, M.; Konig, P. Int

SAMPE Symp 1988, 33, 1546.5. Mijovic, J.; Andjelic, S. Macromolecules 1996, 29, 239.6. Brown, I. M.; Sandreczki, T. C. Macromolecules 1990, 23, 94.7. Liptak, S. C.; Wilkinson, S. P.; Hedrick, J. C.; Ward, T. C.;

McGrath, J. E. Am Chem Soc Symp Ser 1991, 475, 364.8. Mijovic, J.; Corso, W. V.; Nicolais, L.; d’Ambrosio, G. Polym

Adv Technol 1988, 9, 231.9. Barrett, K. A.; Chaudhari, M. A.; Lee, B. H. Int SAMPE Symp

1988, 33, 398.10. Driscoll, S. B.; Walton, T. C. SAMPE J 1987, 23(4), 9.11. Hummel, D. O.; Heinen, K. U.; Stenzenberger, H. D.; Siesler, H.

J Appl Polym Sci 1974, 18, 2015.12. Jordan, C.; Galy, J.; Pascault, J. P. J Appl Polym Sci 1992, 46, 859.13. Wang, Y. D.; Cakmak, M.; Harris, F. W. J Appl Polym Sci 1995,

56, 837.14. Zainol, I. Ph.D. Thesis, UMIST, 2001.15. Demause, M. T.; Johnson, A. C. Mater Res Soc Symp Proc 1994,

347, 723.16. Nightingale, C. Ph.D. Thesis, UMIST, 2000.17. Stenzenberger, H. D. Appl Polym Symp 1973, 22, 77.18. Stenzenberger, H. D. In Polyimides; Wilson, D.; Stenzenberger,

H. D.; Hergenrother, P. M., Eds.; Blackie: London, 1990; p 79.19. Enns, J. B.; Gillham, J. K. J Appl Polym Sci 1983, 28, 2567.20. Marand, E.; Baker, K. R.; Graybeal, J. D. Macromolecules 1992,

25, 2243.21. Billmeyer, F. W. Textbook of Polymer Science, 2nd ed.; Wiley:

New York, 1971; p 256.22. Nielsen, L. E. J Macromol Sci C 1969, 3, 69.23. Loustalot, M. F. G.; Da Cunha, L. Polymer 1997, 39, 1833.24. Di Giulio, C.; Gautier, M.; Jasse, B. J Appl Polym Sci 1984, 29,

1771.25. Tung, C. M. Polym Prepr (Am Chem Soc Div Polym Chem)

1987, 28(1), 7.26. Prime, R. B. In Thermal Characterization of Polymeric Materials;

Turi, E. A., Ed.; Academic: New York, 1981; p 486.

Figure 13 Sol–gel analysis results for samples cured with aconventional oven (OV) or a microwave oven (MW) as afunction of �DSC.


Documents

Comparison between the thermal and microwave curing of bismaleimide resin