Polymer Degradation and Stability 92 (2007) 596e604www.elsevier.com/locate/polydegstab

The degradation of new thermally degradable thermosets obtained bycationic curing of mixtures of DGEBA and 6,6-dimethyl

(4,8-dioxaspiro[2.5]octane-5,7-dione)

Lidia Gonzalez a, Xavier Ramis b, Josep Maria Salla b, Ana Mantecon a,*, Angels Serra a

a Departament de Quımica Analıtica i Quımica Organica, Universitat Rovira i Virgili. C/Marcel$lı Domingo s/n, 43007 Tarragona, Spainb Laboratori de Termodinamica, ETSEIB. Universitat Politecnica de Catalunya, Av. Diagonal 647, 08028 Barcelona, Spain

Received 7 September 2006; received in revised form 14 December 2006; accepted 1 January 2007

Available online 13 January 2007

Abstract

The thermal degradation of thermosetting materials prepared by cationic copolymerization of mixtures of different proportions of diglyci-dylether of bisphenol A (DGEBA) with 6,6-dimethyl (4,8-dioxaspiro[2.5]octane-5,7-dione) (MCP) initiated by ytterbium or lanthanum triflateor using a conventional initiator, BF3$MEA was investigated. To study the thermal degradation, several techniques were used such as thermog-ravimetry (TGA), infrared spectroscopy (FTIR) and calorimetry (DSC) and the volatiles evolved during degradation were identified by massspectrometry. The materials prepared possess the characteristics of thermally degradable thermosets, due to the presence of ester groups inthe polymer chain, which are broken at the beginning of degradation. The degradability increased when lanthanide triflates were used in thecuring, especially the ytterbium salt and when the proportion of MCP in the material increased.� 2007 Elsevier Ltd. All rights reserved.

Keywords: Epoxy resins; Degradation; Thermosets; Cationic polymerizations; Reworkability

1. Introduction

The widespread use of epoxy materials in technologicalapplications is due to their excellent adhesive and thermo-mechanical properties, as well as their easy processing. Oncecured these resins form highly cross-linked networks witha high thermal stability, which can be a disadvantage froman ecological point of view. For example, microelectronicapplications need degradable thermosets, because if a compo-nent of an electronic device is defective the entire circuit boardmust be thrown away. Because the electronic device can bereused, some authors [1] applied the term ‘‘reworkable’’ tothe epoxy thermoset. However, this term does not mean thatthe polymeric material can be reused or recycled. The conceptof reworkable thermoset can be defined as the ability to

* Corresponding author.

E-mail address: anamaria.mantecon@urv.cat (A. Mantecon).

0141-3910/$ - see front matter � 2007 Elsevier Ltd. All rights reserved.

doi:10.1016/j.polymdegradstab.2007.01.007

breakdown under controlled conditions in order to remove itfrom the substrate. The application of these degradation con-ditions to these materials diminishes the crosslinking densityand therefore their mechanical properties under controlledcircumstances. Rework enables the straightforward repair, re-placement or recycling of electronic devices assembled withsuch materials. These thermosets have applications as under-fills, which fill the gap between an integrated circuit chipand the substrate and encapsulate the solder interconnects.These products are adhesives, which reduce the strain in thesolder joints by transferring some of the strain energy intothe underfill layer. The underfill provides enhancement ofsolder fatigue life and also corrosion protection to the inte-grated circuit chip, resulting in 10- to 100-fold improvementin fatigue life as compared to an unencapsulated package [2].

Several authors designed new molecular structures, which oncuring introduce breakable groups into the three-dimensionalnetwork. Thus, Sastri and Tesoro [3] introduced cleavable disul-fide linkages; Buchwalter and Kosbar [4] put hydrolysable ketal

597L. Gonzalez et al. / Polymer Degradation and Stability 92 (2007) 596e604

groups; Malik et al. [5] incorporated sterically hindered urealinkages; Shirai and coworkers [6] sulfonate ester moietiesand recently Liu and Hsieh [7] formed thermally reversiblebicyclic structures from a DielseAlder reaction between malei-mide and furan functionalized polymers. However, the mostextended and easiest approach to prepare reworkable epoxythermosets introduces thermally labile secondary and tertiaryester groups. Among the research groups that follow this routeshould be considered the work reported by Ober [8,1] Shirai[9,10] and Wong [11,12].

Most of the work done in the field of thermally reworkableepoxy thermosets is based on cycloaliphatic structures whichcontain ester groups. Cycloaliphatic epoxides should becross-linked with anhydrides or catalytic systems becausethey do not react with nucleophilic agents such as amines.Moreover, they usually lead to high crosslinking densities,which reduce their mechanical characteristics. Thus, thesereworkable systems lack of versatility and the mechanicalproperties should be improved for several applications.

In previous papers we reported a new route of introducingester or carbonate groups in thermosetting materials by copoly-merizing commercial epoxy resins with lactones [13e15] andcyclic carbonates [16,17] by a ring-opening mechanism. Thisstrategy presents some advantages, such as the possible modu-lation of the properties of the final materials by changing thecharacteristics of the epoxy resin and the co-monomer and theirfeed ratio. Moreover, the type of initiator used can also changethe characteristics of the final materials. Following this routewe could introduce thermally breakable groups into the net-work, to reduce the shrinkage on curing and therefore the inter-nal stress generated by the shrinkage. We could also delay thegelation point until greater conversions, which reduces theshrinkage after gelation. In this way the mechanical propertiescould be increased.

In the present work we report a thermal degradation study ofa series of new materials obtained by cationic copolymeriza-tion of DGEBA epoxy resin with 6,6-dimethyl (4,8-dioxaspir-o[2.5]octane-5,7-dione) (MCP) initiated by ytterbium orlanthanum triflates or using a conventional catalyst, BF3$MEA.In previous work [18] we studied this curing and could provethat the global shrinkage on curing decreases on adding MCPto the epoxy resin and that the proportion of ester groups inthe material depends on the feed ratio of co-monomers andthe initiator used. The aim of this work is to understand thethermal degradation chemistry, the kinetics, and their influenceon the reworkability of the proposed materials.

2. Experimental

2.1. Materials

Diglycidylether of bisphenol A (DGEBA) EPIKOTE RESIN827 from Shell Chemicals (epoxy equiv.¼ 182.08 g/eq) wasused as received.

6,6-Dimethyl (4,8-dioxaspiro[2.5]octane-5,7-dione) (MCP)(Aldrich) was used as received.

Lanthanum (III) and ytterbium (III) trifluoromethanesulfo-nates and borontrifluoride monoethylamine (BF3$MEA)(Aldrich) were used without purification.

2.2. Preparation of the materials

The samples were prepared by mixing the selected initiatorin the corresponding amount of MCP and adding the requiredproportion of DGEBA with manual stirring. The preparedmixtures were kept at �18 �C before use.

The mixture was cross-linked by pouring it into a silanizedglass mould and then cured in an oven at 140 �C for 3 h,followed by post-curing at 160 �C for 2 h (at 170 �C forBF3$MEA). The samples were then allowed to cool at roomtemperature. It was confirmed by DSC that they were com-pletely cured.

2.3. Characterization and measurements

Calorimetric studies were carried out on a Mettler DSC-821e thermal analyzer in covered Al pans under N2 at10 �C/min. The calorimeter was calibrated using an indiumstandard (heat flow calibration) and an indiumeleadezincstandard (temperature calibration). The samples weighedapproximately 7 mg.

The glass transition temperatures (Tgs) of the cured andsemi-degraded materials were calculated as the temperatureof the half-way point of the jump in heat capacity when thematerial changed from the glassy to rubbery state.

The FTIR spectra were recorded with a Jasco FTIR-680PLUS spectrophotometer with a resolution of 4 cm�1 inthe absorbance mode, equipped with an attenuated-total-reflection accessory with thermal control and a diamond crys-tal (golden gate heated single-reflection diamond ATR,SpecaceTeknokroma). The conversions of the reactive groupswere determined from the normalized changes in absorbanceby the LamberteBeer law, as we explained previously [13].

Thermogravimetric analyses (TGAs) were carried out witha Mettler TGA/SDTA 851e thermobalance. Cured sampleswith an approximate mass of 7 mg were degraded between 30and 600 �C at a heating rate of 10 �C/min in N2 (100 ml/min)measured under normal conditions. Non-isothermal thermo-gravimetric tests were carried out at rates of 2, 5, 10 and20 �C/min to evaluate the kinetic parameters. Isothermal stud-ies at selected temperatures were also carried out.

The degradation of the materials was performed in nitrogenatmosphere in a Carbolite tubular oven at 200 and then at250 �C and at selected times several fractions of volatile mat-ter evolved were trapped in a flask refrigerated by liquid nitro-gen. The fractions collected were diluted in acetonitrile andthen studied by gas chromatography (Hewlett-Packard 5890equipped with a capillary column Ultra 2) coupled to a Hew-lett-Packard 5989A quadrupole detector equipped with a dou-ble source ionization by electronic impact and chemicalionization to identify the structures formed.

The residues produced maintaining the materials in anoven at 200 and 225 �C temperatures at different times were

598 L. Gonzalez et al. / Polymer Degradation and Stability 92 (2007) 596e604

investigated by FTIR-ATR spectra and Tgs were determined inorder to follow the degradation of the materials by theseprocedures.

2.4. Kinetic analysis

Integral non-isothermal kinetic analysis was used to deter-mine the kinetic triplet (A e pre-exponential factor, E eactivation energy and g(a) e integral function of degree ofconversion).

The degree of conversion as the mass loss is defined as:

a¼ m0�m

m0�mN

ð1Þ

where m is the mass corresponding to temperature T, m0 is theinitial mass and mN is the mass of the substance at the end ofthe experiment.

If we accept that the dependence of the rate constant on thetemperature follows the Arrhenius equation, non-isothermalkinetic analysis may start with the kinetic equation:

da

dt¼ b

da

dT¼ Aexp

�� E

RT

�f ðaÞ ð2Þ

where b is the heating rate, da/dt is the rate of conversion, R isthe universal gas constant, T is the temperature and f(a) is thedifferential conversion function.

By using the CoatseRedfern [19] approximation to resolvethe so-called temperature integral and considering that 2RT/E<< 1 the KissingereAkahiraeSunose equation (KAS)may be written as:

lnb

T2¼ ln

�AR

gðaÞE

�� E

RTð3Þ

For each conversion degree, the linear representation ofln [b/T2] versus T�1 enables E and ln [AR/g(a)E] to be deter-mined from the slope and the ordinate in the origin. If thereaction model, g(a), is known, for each conversion the corre-sponding pre-exponential factor can be calculated for everyactivation energy.

Integration of rate equation in isothermal conditions givesthe isoconversional expression:

ln t ¼ ln

�gðaÞ

A

�þ E

RTð4Þ

where t is cure time.It can be observed how the isothermal constant ln [g(a)/A]

(Eq. (4)) is directly related by R/E for every value of a to theconstant ln [AR/g(a)E] of the non-isothermal adjustment(Eq. (3)). Thus, taking the non-isothermal data ln [AR/g(a)E]and E from Eq. (3), we can determine the isothermal parame-ters of Eq. (3) and simulate isothermal curing without knowingg(a).

In this study, we used the reduced master curves procedureof Criado [20] and the CoatseRedfern method to assign a reac-tion model to the systems studied [21]. Different kinetic

models have been studied: diffusion (D1, D2, D3 and D4),AvramieErofeev (A2, A3 and A4), power law, phase-bound-ary-controlled reaction (R2 and R3), autocatalytic (nþm¼ 2and 3) and order n (n¼ 1, 1.5, 2 and 3). The rate constant,k, was calculated with E and A determined at conversion of0.5, using the Arrhenius equation.

3. Results and discussion

The reaction of DGEBA and MCP in cationic conditionsleads to the formation of poly(ethereester) structures throughthe formation of intermediate spiroorthoesters (SOEs).Scheme 1 shows the chemical structure of the three-dimen-sional polymer formed. In previous work [18] we have studiedthis curing by FTIR experiments using ytterbium and lantha-num triflates and BF3$MEA as initiators and confirmed theformation of an intermediate spiroorthoester. We also provedthat MCP does not homopolymerise in the conditions used.From these experiments we could determine that ytterbiumtriflate leads to the highest proportion of ester linkages inthe network and the boron complex to the lowest. We alsodetected that some proportion of SOE remained unreacted inthe materials and therefore the proportion of ester groups isin some cases lower than expected. As we can see in thescheme, the use of MCP as co-monomer leads to the formationof tertiary ester groups in the polymeric chain, but the excessof epoxy groups in the formulations contributes to a higher

O CH2 CH CH2

O

O

O

O

O CH3

CH3

O

OCH3

CH3O

OOO CH2

O

OOO

OO

O

CH3

H3C

+

SOE

Lewis acid

Lewis acid

DGEBA

MCP

Scheme 1.

599L. Gonzalez et al. / Polymer Degradation and Stability 92 (2007) 596e604

amount of poly(ether) units. Thus, the higher the proportion ofMCP in the mixture the higher is the proportion of ester groupin the material for a selected initiator.

Ester groups are thermally cleavable by a b-eliminationmechanism [22], which leads to the formation of acid andvinyl ether groups as chain ends; such is represented inScheme 2. Tertiary esters are more easily thermally degradedthan primary or secondary ones. Each time an ester group iscleaved, the local crosslink density decreases and therefore,the higher the proportion of tertiary esters the more degradablecharacter of the modified epoxy resin should be.

Fig. 1 shows the TGA and DTG curves of the materialsobtained from DGEBA/MCP 2:1 (mol/mol) mixtures usingthe corresponding amount of the three initiators tested. Table 1collects the thermal data obtained for all the materials studied.As we can see, the thermal stability of BF3$MEA initiated ma-terials is much higher than when we used lanthanide triflates.

O

OOO

OO

O

H2C

H3C

H

O

OOO

OOH

O CH3

CH2+

Scheme 2.

Yb(OTf)3

La(OTf)3

BF3·MEA

Yb(OTf)3

La(OTf)3

BF3·MEA

20

40

60

80

100

50 150 250 350 450 550

Weig

ht (%

)

50 150 250 350 450 550Temperature (°C)

Fig. 1. Thermogravimetric and DTG curves at 10 �C/min in N2 atmosphere of

several thermosetting materials obtained from DGEBA/MCP 2:1 formulation

initiated by lanthanide triflates and BF3$MEA.

This could be attributed to a lower proportion of ester groups orto a higher crosslinking density. However, from the TGA andthe DTG curves for DGEBA homopolymer samples (Fig. 2) andthe values collected in Table 1 (entries 1, 5 and 9), we can con-clude that the presence of lanthanide triflates in the materialsinitiated the degradation at lower temperatures. Thus, it seemsthat lanthanide triflates play a catalytic role in degradation.Moreover, the higher the Lewis acidity (ytterbium is moreacidic than lanthanum) the greater is this effect. If we lookat the plot of the derivatives in Figs. 1 and 2, we can see dif-ferent shapes for all the samples tested. Thus, the material ob-tained with the ytterbium catalyst has a broad peak for both theco- and homopolymer, which indicates that all the degradation

Table 1

Calorimetric and thermogravimetric data of all systems studied

Entry Formulationa Tgb

(�C)

Tc

(�C)

Tmaxd

(�C)

Char

yielde

(%)

1 DGEBA/Yb 0.006 150 287 345 20

2 DGEBA/MCP/Yb 3:1:0.021 132 241 335 19

3 DGEBA/MCP/Yb 2:1:0.015 124 230 332 19

4 DGEBA/MCP/Yb 1:1:0.009 107 214 319 18

5 DGEBA/La 0.006 145 303 354 19

6 DGEBA/MCP/La 3:1:0.021 127 250 350 18

7 DGEBA/MCP/La 2:1:0.015 121 233 348 17

8 DGEBA/MCP/La 1:1:0.009 100 216 344 17

9 DGEBA/BF3$MEA 0.096 165 332 433 19

10 DGEBA/MCP/BF3$MEA 3:1:0.336 160 296 429 19

11 DGEBA/MCP/BF3$MEA 2:1:0.240 154 288 425 18

a The composition of the formulations are given in molar ratios.b Glass transition temperatures of the materials cured at 140 �C for 3 h and

at 160 �C for 2 h (at 170 �C for BF3$MEA).c Temperature of a 2% of weight loss calculated by thermogravimetry.d Temperature of the maximum degradation rate calculated by

thermogravimetry.e Char yield at 600 �C.

DGEBA / Yb(OTf)3

DGEBA / BF3·MEA

DGEBA / La(OTf)3

DGEBA / BF3·MEA

DGEBA / La(OTf)3

DGEBA / Yb(OTf)3

20

40

60

80

100

50 150 250 350 450 550

Weig

ht (%

)

50 150 250 350 450 550Temperature (°C)

Fig. 2. Thermogravimetric and DTG curves at 10 �C/min in N2 atmosphere of

several thermosetting materials obtained from DGEBA cured with lanthanide

triflates and BF3$MEA.

600 L. Gonzalez et al. / Polymer Degradation and Stability 92 (2007) 596e604

processes overlap. The material obtained with lanthanumtriflate leads to degradation in two stages for DGEBA aloneand to a three peak curve for the copolymer. Thus, it seemsthat the weak broad maximum at 250 �C should correspondto the decomposition of chains in which the presence of estergroups is higher. Therefore, lanthanum and ytterbium triflatesact differently in the thermal degradation, because of theirdifferent capabilities to catalyze the individual degradativeprocesses. Thus, by several b-elimination mechanisms littlefragments could be originated giving place to a weight lossat these temperatures. BF3$MEA leads to the appearance oftwo peaks for both types of materials, more definite for thecopolymeric material and at slightly lower temperature. How-ever, there are not big differences in the initial degradationtemperatures for the ytterbium and lanthanum triflates initiatedmaterials (see entries 3 and 7 in Table 1) but much more dif-ferences with the boron initiated material (entry 11). It must besaid that in order to get reworkable materials the most signif-icant parameter is the initial degradation temperature. Thus,the use of lanthanide triflates as initiators increases the rework-ability of the thermosets.

As expected, there is also an increase in the degradability ofthe materials on increasing the proportion of MCP in the reac-tive mixture. Fig. 3 shows the thermogravimetric curves andtheir derivatives for materials initiated with lanthanum triflate.On increasing the proportion of ester groups the first peak inthe DTG curves increases steadily. Thus, it can be confirmedthat this peak is due to a weight loss in which ester groupsare broken. The second peak, with the highest rate of decom-position decreases slightly in magnitude and temperature onincreasing the proportion of ester groups in the material andtherefore it should be related to the breakage of ether groupin the network.

The values collected in Table 1 for all the initiators showthat the maximum variations on the degradation parameterson increasing the proportion of MCP in the sample occurs in

DGEBA / MCP 3:1DGEBA

DGEBA / MCP 1:1DGEBA / MCP 2:1

20

40

60

80

100

50 150 250 350 450 550

Weig

ht (%

)

50 150 250 350 450 550Temperature (ºC)

Fig. 3. Thermogravimetric and DTG curves at 10 �C/min in N2 atmosphere of

several thermosetting materials obtained from different formulations of

DGEBA/MCP cured with La(OTf)3.

the initial degradation temperature and only slight differencescan be noted in Tmax because this temperature is related to thecleavage of ether groups. Moreover, the introduction of estergroups significantly increases the degradability with referenceto the pure DGEBA. Thus, in the proportion DGEBA/MCP 3:1the initial degradation temperature is reduced by about 40e50 �C with reference to the pure DGEBA, whereas an addi-tional increase of MCP produces a lower decrease. It mustbe noted that big differences in the thermal degradability ofthe materials are obtained with lanthanide triflates and MCPin comparison to conventional DGEBA/BF3$MEA system(entry 9) which reduces both the initial and the maximumrate degradation temperatures. Lower initial degradation tem-peratures are not desirable because degradation could occur oncuring or during use. With reference to the residues after heat-ing until 600 �C there are not big differences among the mate-rials, and only a slight decrease was observed on increasingthe proportion of MCP.

The degradability was also tested by isothermal thermog-ravimetry. Fig. 4 shows the plot of weight loss versus timeat 275 �C. We can see that ytterbium triflate at longer timesraises the weight loss with reference to the lanthanum salt,although in the initial steps of degradation there are only littledifferences, although lanthanum salt increases the initialdegradability. Moreover, the effect of increasing the propor-tion of MCP in the sample is higher for materials initiatedby lanthanum than by ytterbium salts.

Thermogravimetric analysis has been widely used to esti-mate the kinetic parameters of degradation processes, suchas activation energy (Ea), kinetic model, Arrhenius pre-exponential factor (A) and rate constant (k) [23,24]. The effectof the initiators and the formulations used to prepare the sam-ples were dynamically studied by TGA. For each material thetests carried out at different rates provided the weight loss andthe rate of weight loss depending on the temperature. Thesedata were standardised by means of Eq. (1) to find the degreeof conversion with the temperature at a rate of 10 �C/min.

70

80

90

100

0 10 20 30

76%

79%

86%

81%Weig

ht (%

)

time (min)

2:1-La

3:1-La

2:1-Yb

3:1-Yb

Fig. 4. Isothermal thermogravimetric curves at 275 �C for 25 min in N2 atmo-

sphere of several thermosetting materials obtained from formulations DGEBA/

MCP 3:1 and 2:1 cured with Yb(OTf)3 and La(OTf)3.

601L. Gonzalez et al. / Polymer Degradation and Stability 92 (2007) 596e604

To find the kinetic parameters associated with the set ofthermogravimetic curves obtained for different materials, weapplied the isoconversional KAS equation, Eq. (3), at a conver-sion of 0.5. We used the reduced master curves procedure ofCriado [20] and the CoatseRedfern method [19] to assign areaction model to the degradative processes studied. By thisprocedure we calculated the apparent activation energies asso-ciated with the simultaneous degradation processes and thepre-exponential factor for the n¼ 3 kinetic model. The factthat the shape of the curves have more than one peak for lan-thanum triflate and BF3$MEA cured materials (Fig. 1) makethe comparison of the kinetics of degradation difficult. Thus,the kinetic study was only done for the ytterbium triflate curedmaterials, which shows a unimodal degradation curve. Theresults are summarized in Table 2. The activation energy ata conversion of 0.5 and the pre-exponential factor do notshow a clear tendency when the proportion of MCP was var-ied. Because of the compensation effect between the activationenergy and the frequency factor it would be more accurate toanalyze the degradation rate using the rate constant calculatedby the Arrhenius equation. Thus, the values of k confirma quicker degradation of the material when a higher proportionof MCP was added in the modification of DGEBA.

By applying the isoconversional procedure, Eqs. (3) and(4), we can predict the time necessary to reach a determineddegree of weight loss at a selected temperature. Fig. 5 showsthe plot of weight loss against time at 275 �C for the differentformulations cured with lanthanide triflates. As a generaltrend, the materials cured by ytterbium salts show a much bet-ter reworkability. Shorter times are needed for the degradationof the materials obtained from DGEBA/MCP 1:1 formulationfor both initiators. Whereas at weight losses of 25% the differ-ences in time to reach this conversion in all the samples are notmuch (in a range of 130 min) when the weight loss is about50% these differences increase notably (about 450 min).This behaviour can be related to the initial degradation of estergroups, which should lead to more volatile fragments.

Fig. 6 shows the dependence of the activation energy on thedegree of conversion; calculated using the isoconversionalmethod, Eq. (3) [25], for the materials prepared using ytter-bium triflate as initiator. As we can see, at lower degrees of

Table 2

Kinetic parameters calculated from the thermogravimetry of the materials

obtained using ytterbium triflate as initiator

Entry Formulationa Eab

(kJ/mol)

ln Ac

(s�1)

k300�Cd� 103

(s�1)

1 DGEBA/Yb 0.006 155.4 23.43 0.1016

2 DGEBA/MCP/Yb 3:1:0.021 164.5 25.78 0.1574

3 DGEBA/MCP/Yb 2:1:0.015 163.3 25.66 0.1778

4 DGEBA/MCP/Yb 1:1:0.009 168.5 27.49 0.3750

a The composition of the formulations are given in molar ratios.b Values of activation energies were evaluated by the isoconversional inte-

gral method (Eq. (3)) applied at conversion of 0.5.c The values of pre-exponential factor for kinetic model with n¼ 3 kinetic

model g(a)¼ 2�1[�1þ (1� a)�2].d The values of rate constants at 300 �C were calculated using the Arrhenius

equation ln k ¼ ln A� ðE=RTÞ.

conversion the lowest energy is the one of the material witha higher proportion of MCP. This can be attributed to thefact that degradation may begin by the cleavage of estergroups. However, at conversions about 0.5 there is no cleartendency because mainly the ether groups in the network arebroken. Fig. 7 shows the dependence of activation energy onthe degree of conversion for the materials obtained fromDGEBA/MCP 3:1 formulation cured with the three initiators.The materials prepared with lanthanide triflates show a similarevolution, but during the whole process the material with theytterbium salt degrades with lower activation energy. The ma-terial obtained with the boron complex has a completely dif-ferent behaviour, because the initial value is much higher,then decreases until a 30% of conversion and finally increasessteadily until the end of the degradation. This behaviour can berelated to the two maximum observed in the DTG curves.

15

30

45

60

0 100 200 300 400 500time (min)

Weig

ht lo

ss (%

)

3:1-Yb2:1-Yb

DGEBA-Yb

1:1-Yb

DGEBA-La3:1-La2:1-La1:1-La

Fig. 5. Predicted evolution of weight loss (%) against time at 275 �C of several

thermosetting materials obtained from DGEBA/MCP formulations cured with

ytterbium and lanthanum triflate.

Conversion (%)

25

75

125

175

225

275

0 25 50 75 100

Ea (kJ / m

ol)

DGEBA

3:1

2:1

1:1

Fig. 6. Dependence of activation energy on the degree of conversion obtained

by applying isoconversional analysis to thermogravimetric curves of several

thermosetting materials prepared from DGEBA, and DGEBA/MCP formula-

tions cured with ytterbium triflate.

602 L. Gonzalez et al. / Polymer Degradation and Stability 92 (2007) 596e604

Several authors [1] studied the change in the glass transitiontemperature (Tg) of the reworkable thermosets as a function ofthe thermal treatment, because there is a relationship betweenthis parameter and the crosslink density, as Tg varies inverselywith Mc and affects the mechanical properties of the material.In contrast to using TGA, where only the loss of volatiles ismeasured, the determination of the Tg by DSC during degrada-tion reveals a more direct information on the behaviour of thenetwork. Initial Tg determined from the materials cured in iso-thermal conditions are collected in Table 1. Fig. 8 shows thevariation of Tg of the materials initiated by ytterbium triflateat a temperature of 200 and 225 �C. As we can see, an increasein the proportion of MCP in the sample leads to a notablereduction of the Tg of the material by applying the same ther-mal treatment. The diminution of the Tg is more evident at thebeginning and increases noteworthy with the proportion ofester in the network, which is in accordance with the cleavage

50

100

150

200

250

300

0 20 40 60 80 100Conversion (%)

Ea (kJ / m

ol)

Yb(OTf)3

La(OTf)3

BF3·MEA

Fig. 7. Dependence of activation energy on the degree of conversion obtained

applying isoconversional analysis to thermogravimetric curves of several ther-

mosetting materials obtained from DGEBA/MCP 3:1 (mol/mol) formulation

cured with lanthanide triflates and BF3$MEA.

85

95

105

115

125

135

0 250 500 750 1000 1250

time (min)

Tg

(°C

)

3:1 (225ºC)2:1 (225ºC)3:1 (200ºC)2:1 (200ºC)

Fig. 8. Evolution of Tg against time for the materials obtained from DGEBA/

MCP 3:1 and 2:1 formulations cured with ytterbium triflate after degradation

at 200 and 225 �C in an oven.

of ester groups that reduce the crosslinking density but doesnot produce a significant weight loss. On increasing the tem-perature from 200 to 225 �C the effect is much more evident.

In order to follow the chemical processes involved in thedegradation we recorded the FTIR spectra of the materialsmaintained at 200 and 225 �C for 20 h and integrated the esterabsorption at 1734 cm�1 and the reference band at 1605 cm�1

of the DGEBA aromatic ring. Fig. 9 shows the variation of thenormalized absorption of the ester group against time at 200and 225 �C for two different formulations initiated by ytter-bium triflate. In this way, we can monitor the disappearanceof ester groups and therefore the cleavage of the main chain.As we can see, on increasing the temperature from 200 to225 �C there is a rapid variation in the evolution of estergroups. Thus, 30 min at 225 �C leads to a loss of 39% of estergroups of the sample 2:1 and 25% of the sample 3:1, whereasat 200 �C only a 19% and a 2%, respectively, disappear. Thisindicates that a temperature of 225 �C is much better than200 �C to degrade the sample and to rework the materials.In fact, it is necessary to heat for 480 min at 200 �C to cleavehalf of the ester groups in the formulation 2:1 and 540 min forthe formulation 3:1.

In order to know the identity of volatile fragments producedin the degradation of the thermosets, we put the material ob-tained from a DGEBA/MCP (2:1) mixture cured with ytter-bium triflate into an oven at 200 �C for 7 h and then at250 �C for 3 and 9 h. The volatiles evolved were collectedin three fractions trapping them in liquid nitrogen and thenstudied by gas chromatography coupled with mass spectrome-try. Several products could be identified, which gave someinformation about the degradative processes. In the first frac-tion, obtained after 7 h at 200 �C, we could identify phenoland isopropyl phenol, which are formed by the breakage ofether linkages of the units introduced by DGEBA and cyclo-propane carboxylic acid which can be formed by a doubleb-elimination of the diester group introduced by MCP anddecarboxylation such as the one represented in Scheme 3.Thus, at temperatures of 200 �C not only ester groups are

0.5

0.7

0.9

1.1

1.3

1.5

1.7

0 300 600 900 1200

A1734/A

1605

time (min)

3:1 (200°C)2:1 (200°C)3:1 (225°C)2:1 (225°C)

Fig. 9. Evolution by FTIR spectroscopy of the A1734/A1605 ratio against time

for the degradation in an oven in N2 atmosphere at 200 and 225 �C of the ma-

terials obtained from 3:1 and 2:1 formulations cured with Yb(OTf)3.

603L. Gonzalez et al. / Polymer Degradation and Stability 92 (2007) 596e604

broken but also ether groups and CeC linkages. In the frac-tions collected at 250 �C, the second and the third after 3and 9 h, respectively, only the former contained cyclopropanecarboxylic acid, which means that ester groups are broken

O

O

OO

OOH H H

O

OO

OH

O

OO

OH

H COOHCO2

Scheme 3.

much easier than ether or carbonecarbon linkages. Moreover,in this fraction and in the third we could identify other prod-ucts represented in Scheme 4.

In a previous article about the preparation of this type ofthermosetting materials we proved that the increase of MCPin the materials reduced the global shrinkage, especiallywhen we used ytterbium triflate as initiator [18]. In the presentstudy we confirm that a higher proportion of MCP in the ma-terial and the use of ytterbium triflate are also advantageous inorder to reach the highest reworkability of these thermosets.Thus, lanthanide triflates are better than the conventionalcationic BF3$MEA initiator in order to reduce the shrinkageand to increase the degradability of this type of thermosets.Moreover, the addition of MCP to the samples does not worsethe Tg values of the materials due to the structure of the co-monomer with a tertiary ester and a cyclopropyl group whichrestrict the internal mobility of the polymeric chain. Thus,

FRACTION 1 FRACTION 2

FRACTION 3

C OH

O

Cyclopropanecarboxylicacid (m / z 86)

O

OH

3-(4-isopropylphenoxy)prop-2-en-1-ol (m / z 192)

O

CH3

1-isopropyl-4-methoxybenzene (m / z 150)

OH

4-ethylphenol(m / z 122)

OH

4-isopropyl phenol(m / z 136)

OH

Phenol (m / z 94)

C OH

O

Cyclopropanecarboxylicacid (m / z 86)

OH

4-isopropyl phenol(m / z 136)

benzene(m / z 78)

O

CH3

1-isopropyl-4-methoxybenzene (m / z 150)

OH

4-ethylphenol(m / z 122)

isobutane(m / z 44)

OH

4-isopropyl phenol(m / z 136)

Scheme 4.

604 L. Gonzalez et al. / Polymer Degradation and Stability 92 (2007) 596e604

it seems that the strategy of modification of DGEBA bycopolymerization with MCP allows the enhancement of someproperties of the thermosetting materials at the same time.

4. Conclusions

The copolymerization of DGEBA and MCP increases thethermal degradability of epoxy resins through the cleavageof ester groups introduced into the polymer chain. The higherthe proportion of MCP in the reactive mixture the moredegradable the materials are.

Lanthanide triflates lead to more degradable materials thanthe conventional BF3$MEA cationic initiator.

From the volatiles evolved during degradation and theevolution followed by FTIR spectroscopy of the residues aninitial b-elimination of the tertiary ester groups can beconfirmed.

Acknowledgements

The authors from the Universitat Politecnica de Catalunyawould like to thank CICYT and FEDER (MAT2004-04165-C02-02) for their financial support. The authors from theRovira i Virgili University would like to thank the CICYT(Comision Interministerial de Ciencia y Tecnologıa) andFEDER (Fondo Europeo de Desarrollo Regional) (MAT2005-01806).

References

[1] Chen JS, Ober CK, Poliks MD. Polymer 2002;43:131e9.

[2] Wang L, Li H, Wong CP. IEEE Trans Electron Packag Manuf

2001;24:115e22.

[3] Sastri VR, Tesoro GC. J Appl Polym Sci 1990;39:1425e37;

Sastri VR, Tesoro GC. J Appl Polym Sci 1990;39:1439e57.

[4] Buchwalter SL, Kosbar LL. J Polym Sci Part A Polym Chem

1996;34:249e60.

[5] Malik J, Goldslager K, Clarson SJ. J Appl Polym Sci 2002;85:

856e64.

[6] Shin Y, Kawaue A, Okamura H, Shirai M. React Funct Polym

2004;61:293e302.

[7] Liu YL, Hsieh CY. J Polym Sci Part A Polym Chem 2006;44:

905e13.

[8] Yang S, Chen JS, Corner H, Breiner T, Ober CK, Poliks MD. Chem

Mater 1998;10:1475e82.

[9] Shirai M, Morishita S, Okamura H, Tsunnoka M. Chem Mater

2002;14:334e40.

[10] Shirai M, Kawaue A, Okamura H, Tsunnoka M. Chem Mater

2003;15:4075e81.

[11] Wang L, Li H, Wong CP. J Polym Sci Part A Polym Chem

2000;38:3771e82.

[12] Wong CP, Wang L, Shi S-H. Mat Res Innovat 1999;2:232e47.

[13] Mas C, Ramis X, Salla JM, Mantecon A, Serra A. J Polym Sci Part A

Polym Chem 2003;41:2794e808.

[14] Gimenez R, Fernandez-Francos X, Salla JM, Serra A, Mantecon A,

Ramis X. J Polym Sci Part A Polym Chem 2005;46:10637e47.

[15] Mas C, Ramis X, Salla JM, Mantecon A, Serra A. J Polym Sci Part A

Polym Chem 2006;44:1711e21.

[16] Cervellera R, Ramis X, Salla JM, Mantecon A, Serra A. J Polym Sci

Part A Polym Chem 2005;43:5799e813.

[17] Cervellera R, Ramis X, Salla JM, Serra A, Mantecon A. J Polym Sci Part

A Polym Chem 2006;44:4546e58.

[18] Gonzalez L, Ramis X, Salla JM, Mantecon A, Serra A. J Polym Sci Part

A Polym Chem 2006;44:6869e79.

[19] Coats AW, Redfern JP. Nature 1964;201:68e9.

[20] Criado JM. Thermochim Acta 1978;24:186e9.

[21] Ramis X, Salla JM, Cadenato A, Morancho JM. J Therm Anal Cal

2003;72:707e18.

[22] Sivasamy P, Palaniandavar M, Vijayakumar CT, Ledered K. Polym

Degrad Stab 1992;38:15e21.

[23] Morancho JM, Salla JM, Ramis X, Cadenato A. Thermochim Acta

2004;419:181e7.

[24] Liu YL, Wei WL, Chen YJ, Wu CS, Tsai MH. Polym Degrad Stab

2004;86:135e45.

[25] Ramis X, Cadenato A, Morancho JM, Salla JM. Polymer 2003;44:

2067e79.