Characterization of epoxy-asphalt binders by differential scanning calorimetry

(cid:1) Curing kinetic parameters and glass transition changes were determined for epoxy polymers and epoxy-asphalt binders under different calorimetric conditions. (cid:1) The glass transition temperature was shifting to higher temperatures as cure progresses and crosslinks were formed, with the neat epoxy polymers to crosslink faster than the epoxy-asphalt binders. (cid:1) The determination of optimum curing conditions based on calorimetric measurements can assist the future material formulations and asphalt processing technologies to avoid the uncontrolled curing of epoxy-asphalt binders. as a reference. The kinetic parameters and the T g change of various crosslinking epoxy-based binders was assessed for the different formulations. According to the results, the T g is shifting to higher temperatures as cure progresses and crosslinks are formed, with the neat epoxy-based polymer to crosslink faster than epoxy-asphalt binders. Difference on the crosslinking performance between the two epoxy polymers was noticed as well. This research provided valuable insight into the chemical thermodynamics of crosslinking epoxy-asphalt that can help the future material designers to control reaction-induced phenomena, such as the phase separation.


Introduction
Nowadays, the global demand for more durable pavement materials designed and developed to withstand the continuously increasing traffic loads and the extreme temperature fluctuations due to the climate change leads to the use of modifiers of different chemistries and various origins. Asphalt binder is a petroleum derivative product from refineries and its composition is influ-enced from the crude oil source and the refining techniques applied to it [1][2][3]. The incorporation of different modifiers in these binders is a common practice but empirically driven in most of the cases. Therefore, fundamental understanding of the kinetics and thermodynamics of modified asphalt binders and how these affect the materials properties and subsequently the pavements performance is of high importance, especially today in which new modification technologies, such as epoxy-based polymers, are proposed.
Epoxy-based polymers have been accepted as premium solution to develop durable asphaltic materials, specifically in countries where the demand to reduce the maintenance costs of asphalt pavements is high [4][5][6][7]. Particularly, epoxy-asphalt has been utilized mainly as surfacing solution on airfields and steel bridge decks with the San Mateo-Hayward bridge (California, the United States) to be the most notable example since this epoxyasphalt surfacing was in service for more than 50 years without maintenance [8]. Nevertheless, the main concern that prohibits the use of epoxy-based modifiers in asphalt is the uncontrolled irreversible curing of epoxy-asphalt [9,10]. Contractors could face high risks of construction failures, because of the relatively slow curing of epoxy-asphalt in case of allowing too early traffic after construction [6].
In this study, the use of differential scanning calorimetry is discussed by performing analyses to asphalt binders modified with various types of epoxy formulations to understand the relationship between the processing (e.g., during in-plant production and infield construction of asphalt pavements) and the properties of studied materials. The determination of optimum cure conditions based on calorimetric measurements will assist the future material formulations and asphalt processing technologies to avoid the uncontrolled over-curing of epoxy-asphalt binders and thus the negative effect of reduced material workability during construction. The procedure to predict the reaction kinetics of crosslinking epoxy-asphalt binders is discussed as well and the calorimetric measurements are introduced later on providing the influential parameters of different epoxy-asphalt binder formulations.

Background
The differential scanning calorimetry (DSC) is the most widely used instrument to determine the enthalpy related transitions of asphalt binders [11,12] and of polymeric materials [13]. DSC monitors the endothermic or exothermic heat flow of a sample under a controlled temperature program, considering that the heat generation and its rate is proportional to the reaction rate when crosslinking polymers are studied [14] and to thermal events, such as glass transition [15]. Especially, all polymeric materials are either in liquid or rubbery state at high temperatures and cooling of these materials demonstrates the divergency between crystalline and amorphous state. In the transition region from an amorphous to a crystalline state, material properties typical of glass are exhibited and stepwise increase of modulus, specific volume, and thus of thermal expansion coefficient, occur. Glass transition temperature (T g ) is the temperature that corresponds to this transition. DSC analyses have been used successfully to evaluate the T g of asphalt binders of different origins as well as the effect of various asphaltic fractions [16][17][18], modifiers [19][20][21][22] and rejuvenators [23,24] on their glass temperature region. The T g can be used to interpret thermal-related defects in asphalt pavements, such as the thermal cracking. For instance, binders with high T g accumulate less thermal stress build-up under a given thermal history and thus are more resistant to low temperature cracking. Based on past research, the T g of asphalt binders of different performance grades and from different crude oil resources was shown to extend over the range between À60 and 0°C [11] and between À50 and À10°C [12] exhibiting different melting peaks. Further, the T g depends on the degree of aging of asphalt binders [19]. An aged asphalt binder shows wider glass temperature range mainly because the new microstructure arrangement leads to weakening of association between the asphaltic molecules. A link of asphaltenes increase has been found with the T g increase [19]. In polymermodified binders, the compatibility of a certain polymeric formulation to asphalt binder is affected negatively when a binder has high asphaltene content [25]. Improvement of polymer-asphalt binder miscibility is reached when the aromaticity of maltenes decreased to certain values [25].
In epoxy-asphalt binders, a T g range from À8 to 50°C for fully cured epoxy-asphalt binders was reported in [22]. It has been found as well that the bisphenol A diglycidyl ether (DGEBA), main compound of epoxy resins, is immiscible with asphalt binders [25][26][27]. Nevertheless, improved miscibility between epoxy-based polymers and asphalt binder is possible by applying specific formulations. Certain epoxy formulations may result miscible polymeric systems with asphalt binders and thus it is of high importance to evaluate their impact on the curing (crosslinking) behaviour of epoxy-asphalt binders.

Polymerization reaction kinetics
Αs a crosslinking (polymerizing) material turns irreversibly into networked microstructures, like epoxy-asphalt binder, and thus it cures, heat releases proportionally to the consumption rate of reactive elements as where C A and C B are the concentrations of the two reactive components of epoxy-asphalt binders before reaction, C AB is the concentration of the newly produced material after curing, K is the rate of reaction and DH T is the exothermic heat of chemical transition. The exothermic heat of reaction DH up to time t can be calculated by where dQ/dt is the rate of exothermic heat generation. Non-isothermal measurements of reaction heat of binders as a function of time and temperature yield information about the degree of curing. The conversion data is derived from the measured exothermic heating rates by dividing the rates by the total exothermic heat. Particularly, the total area under the calorimetric scans, based on the extrapolated base line at the end of reaction, was used to calculate the total heat of reaction (DH T ).
The degree of cure or conversion x is given as where DH(t) is the partial area under a calorimetric measurement up to time t. The total exothermic heat DH T up to t T is defined as where t T is the time required to complete the reaction and DH T is the amount of heat generated during the measurement until the completion of cure. The rate of cure is a parameter proportional to the rate of heat release defined as where dQ/dt is the rate of heat generation and is provided by the measurements.
The determination of reaction kinetics is crucial to develop a prediction scheme for physico-mechanical properties of crosslinking materials [9,10]. The kinetic model allows to predict the cure degree and the reaction rate as a function of time and temperature of crosslinking binders as where K(T) is a temperature-dependent parameter, which is described by an Arrhenius equation, and f(x) is a function of conversion.
In the n th order model, no diffusion controlled factor need to be applied and thus the reaction rate is defined as KðTÞ where K 0 is the pre-exponential kinetic factor, E a is the activation energy or the energy barrier to be overtaken to begin the reaction, R is the universal gas constant and n is the reaction order upon the crosslinking. The crosslinked, or fully cured, molecules do not flow easily when the conditioning temperature of epoxy modifier becomes the full cure temperature and, thus, their mobility lowers. Hence, the Eq. (8) is not always valid for modelling the epoxy-based polymers. Nevertheless, in the case of epoxy-asphalt, it is a sufficient function to describe the reactions of these materials. During the non-isothermal calorimetric analysis in DSC, temperature varies linearly with time as where T is the current temperature, T(0) is the initial temperature, t is time, and b is the calorimetric scanning rate. By combining Eq. (9) with Eq. (11) yields Eq. (12), which represents the theoretical shape of a calorimetric scan, as Eq. (12) can be reduced to a linear form by taking the natural logarithm of both sides of the equation as A multi-linear regression is performed using ln[b(dx/dt)], À1/RT, and ln(1 À x) as variables derived from DSC data to solve for K 0 , E a , and n.
When the degree of curing reaches a high level, a quantitative measurement of the residual heat is very difficult due to the small exotherm. In that case, measuring the T g is a more sensitive way to determine the degree of cure. It is well accepted that for many crosslinking polymers there is a unique relationship between the T g and the degree of cure [14]. Therefore, the evolution of T g is examined in this research as well.

Materials and preparation
Epoxy-based polymers for asphalt binders are normally binary reactive systems able to form co-polymerized networks when epoxy monomers react chemically with specially terminated prepolymers under certain energy conditions. In this research, two polymers were selected and their components used as received without any further purification. The two studied epoxy-based polymers, named KD-HDP and KD-HYP, were supplied by Kindaikasei Corp, Japan, and both polymers were formulated by two liquid components; (i) the epoxy and (ii) the hardener. Especially, the component used as hardener (i.e., B-KD) was an acrylonitrile-butadiene pre-polymer for both KD-HDP and KD-HYP, and the mix of epoxy monomers of KD-HDP and KD-HYP were named A-HDP and A-HYP, respectively.
The two components of KD-HDP were oven-heated for 1 h at 50°C, and then mixed together at weight ratio 61:39 (A-HDP:B-KD), according to the supplier. In similar way, the two components of KD-HYP polymer (weight ratio 50:50 of A-HYP and B-KD) were heated for 1 h at 30°C in the oven, and immediately afterwards mixed together. In both cases, the two components were mixed together for approximately 20 s to minimize the chemical reactions during mixing. The freshly prepared mixes were immediately placed in a refrigerator at À10°C to freeze the reactions. The uncured mixes were directly used for calorimetric measurements.
For the formulation of an epoxy-asphalt binder with equivalent amount of epoxy (20% wt resin of component A) for all studied cases, the uncured KD-HDP and KD-HYP polymers were mixed with an already pre-heated asphalt binder of 70/100 pengrade for 60 s at 170°C. The weight ratio of epoxy-asphalt binders (epoxy:hardener:asphalt) based on the before mentioned chemistries were 20:13:67 and 20:20:60, respectively.
To compare the material behaviour of newly formed epoxyasphalt binders, extra samples prepared using a commercial available epoxy-asphalt binder (EA ref ) formulated with equivalent amount of epoxy resin. This product formulated from two liquid components as well; (i) the component A, which includes epichlorohydrin-bisphenol-A (DGEBA), and (ii) component B, which consists of a mix of a 70 pengrade petroleum asphalt binder with heavy naphthenic distillates and extracts. Nevertheless, the exact incorporating chemistry of component B is proprietary protected by the supplier and thus its chemical formulation is unknown. All the samples were prepared by mixing component A and B at weight ratio of 20:80. Unless otherwise stated, the two components were oven-heated separately for 1 h, to 85 and 110°C, respectively, and after that mixed together for 20 s.
The compositions of studied formulations in this research are collected in Table 1.

Calorimetric measurements
The standard and modulated calorimetric measurements were performed on Perkin-Elmer DSC 4000. Enthalpy calibration in the instrument was conducted using indium standard and dry nitrogen was used as purge gas. After the formulation of samples, they were sealed in aluminium pans with sample weights between 15 and Table 1 Name and composition of studied materials. The total heat flow (dQ/dt) as function of enthalpic (constant pressure) heat capacity of material (C p ), the rate of temperature change (dT/dt) and a kinetic component, which is provided as a function of time (t) and temperature (T), is defined as To study the reaction kinetics of previously described binders, the non-isothermal behaviour of materials in an inert nitrogen environment was investigated using heating rates of 5, 10, 20 and 40°C/min in a temperature range from 0 to 300°C. Moreover, non-isothermal measurements performed to determine the T g evolution at different times over curing of studied materials (two replicates per material). Specifically, all samples were ovenconditioned at 130°C for different time intervals; from 0 to 3 h with time step of 0.5 h and scanned from 0 to 300°C with 40°C/ min heating rate.
In addition to standard non-isothermal mode of DSC, modulated calorimetric analyses were conducted for the same purpose for all newly formed systems. In modulated studies, samples after placing them in aluminium pans were cooled to À50°C for 5 min and then were heated to 200°C at a rate of 5°C/min. The modulated temperature signal was 1°C at a modulation period of 120 s. After reaching the goal temperature, the samples were equilibrated for 5 min at 150°C. A sample was modulated tested for each formulated binder. Fig. 1 shows the standard non-isothermal calorimetric scans of EA ref for a temperature range from 30 to 300°C. As shown in these scans, a single exothermic peak is demonstrated for all applied heating rates, something which happens for the other studied materials as well. Based on these scans, the total exothermic reaction heat (DH T ) of different binders was determined by integration of DSC non-isothermal signal at different heating rates. The shape of exothermic heat was almost the same for all binders independently on the implementing epoxy formulation. The kinetics parameters of occurring reactions can be calculated based on the theory discussed earlier and thus the effect of different formulations in curing of neat epoxy polymers and epoxy-asphalt binders is determined (see Fig. 2). The kinetics data of curing of studied materials are listed in Table 2, which includes also the calorimetric information obtained from both standard and modulated measurements. The later will be discussed in the next sub-section.

Standard calorimetric measurements
The proportion of epoxy (i.e., 20% wt) and asphalt binder together with hardener (i.e., 80% wt) is fixed for all modified binders, but only the formulation of epoxy polymer was changed. It was expected that on changing the epoxy formulation, significant differences observed in the kinetic parameters of binders (i.e., E a , K 0 , n). As shown in Fig. 2, the activation energy (E a ), which is the energy barrier to be overtaken to initiate the curing reaction [9], is higher for EA ref than EA-HDP and EA-HYP. Thus, the reference epoxy-asphalt binder demands higher applied energy levels to form crosslinks and to be cured. However, once this binder overtakes this energy barrier, it cures faster than the other binders. This observation is contradictory with previous studies on the topic shown marginally lower activation energy values for epoxyasphalt binder (46-49 kJ/mol [27] and 65 kJ/mol [28]) than of their neat epoxy polymers (50 kJ/mol [27] and 78 kJ/mol [28]). In [27,28], the curing kinetic parameters have been determined using the same formulas discussed earlier in the paper. In general, the contact opportunities between the reactive components of epoxy polymers are reduced by adding the asphalt binder and the cure reactions are inhibited leading to higher energy needs [9].
In addition to kinetic parameters, the T g changes due to the cure of different crosslinking binders is examined. The T g is shifting to higher temperatures as cure progresses and crosslinks are formed, and thus a higher T g corresponds to a greater degree of cure. In other words, the T g values rise with material crosslinking until achieving an ultimate value corresponding to the ultimate conversion at a given temperature. Fig. 3 shows how the T g changes as cure progresses by the standard non-isothermal scanning of EA ref at different time intervals. Further, the evolution of T g over curing time (in the oven) of different epoxy-based polymers (i.e., KD-HDP and KD-HYP) is shown in Fig. 4(a). Similarly, the T g increase due to the cure of epoxy-asphalt binders (i.e., EA-HDP, EA-HYP and EA ref ) is demonstrated in Fig. 4(b) as well. The T g studies after nonisothermal curing demonstrates significant difference between the epoxy-based polymers and modified binders. By comparing Fig. 4(a) and (b), it is obvious that the epoxy cure faster than epoxy-asphalt binders showing that the asphalt binder prohibits the interaction between the reactive components of polymers and thus slows down the crosslinking process. Moreover, difference on the extent of the T g is observed when different polymers are cured. During cure, the KD-HYP demonstrates higher extent of the T g than KD-HDP (see Fig. 4(a)). Overall, the T g was very sensitive to small changes in the degree of cure by performing stan- dard calorimetric measurements in crosslinking epoxy-asphalt binder and thus it can be used as alternative method to monitor the cure progress.

Modulated calorimetric measurements
As mentioned before, the modulated calorimetric measurements offer higher precision in the determination of the T g by isolating scans from other thermal-driven events as crystallization and melting. In case of reacting systems as crosslinking epoxyasphalt binders, the modulated function can offer kinetic curves as shown in Fig. 5 reflecting the exothermic heat flow unaffected by the heat flow effect of change in C p . In this study, the underlying heating rate was 5°C/min with 120 sec modulation period and 1°C modulation amplitude. Especially, modulated calorimetric scans track the difference in heat flow between a sample and an inert reference pan as function of time and temperature, as in standard calorimetric scans. However, a sinusoidal modulation (oscillation) is applied on a linear heating (or cooling) ramp as shown in   The specific heat capacity signal was calculated by the software based on Discrete Fourier Transformation where the measured amplitudes of the sample temperature and heat flow modulation compared to a reference sinusoidal wave of the same frequency (see black line in Fig. 5). The raw data and the enthalpic calculation for neat epoxy polymers and newly developed epoxy-asphalt binders are provided in Fig. 6.
Regarding the scan of EA ref in Fig. 5, it is obvious the thermal history effect in the heat flow data since irregular peaks appeared. These irregular peaks affect the precise positioning of the baseline tangents and thus the individual T g values of epoxy and asphalt binder cannot be given. This issue is difficult to be resolved for the uncured binders because they should be pre-heated at high temperatures and then cooled down with modulation before any measurement to remove any extraneous thermal effect. This pretreatment before cure will result crosslinks in the studied binders and thus the accuracy of kinetics will be lost. Therefore, it is hard to determine the exact evolution of the T g during curing for binder of thermosetting nature by performing modulated measurements. Therefore, the unmodulated measurements of thermosetting binders through curing studies are more reliable than of modulated ones. This explains the change of the T g position as the cure progresses and thus a broader glass transition region appeared. In Figs. 5 and 6, the broad glass transition region is shown having T g values closer to those of fully cured binders. Finally, the total exothermic heat flow determined from the two performed calorimetric methods demonstrated similar values. For example, the total heat released from the crosslinking EA ref was À50,04 and À51,58 J/gr for the modulated and standard analysis (heating rate of 20°C/min), respectively.

Summary and future challenges
In this study, two different formulations of epoxy-based polymers were used as modifiers for asphalt binder by having a com- mercial available epoxy-asphalt binder as reference, and their thermal behaviour under standard and modulated calorimetric measurements was observed. These measurements performed in a DSC provided insight information about the processing parameters, such as T g , by considering the release of heat during the calorimetric measurements proportional to the progress of chemical reactions. Kinetic parameters (i.e., the activation energy, reaction order and rate) and the T g change were determined for the epoxy polymers and epoxy-asphalt binders under different calorimetric conditions. Difference on the crosslinking performance between the two epoxy polymers was noticed. Particularly, the T g is shifting to higher temperatures as cure progresses and crosslinks are formed, with the neat epoxy polymers to crosslink faster than epoxyasphalt binders. The asphalt binder prohibits the interaction between the two reacting epoxy components and thus the reduc-tion of the T g evolution rate obtained. The effect of applied energy on crosslinking of studied materials was more pronounced on the neat epoxy compared to the asphalt binders modified with those polymers.
The insight information provided by the standard measurements in DSC were compared those obtained from the calorimetric measurements under modulated temperatures as well. As known the modulated calorimetric measurements offer higher precision in the determination of the T g by isolating the scans from other thermal-driven events. Nevertheless, the exact evolution of the T g of crosslinking materials by performing those measurements was hard to obtained. A broad glass transition region was provided with T g values closer to those of fully cured binders. During the modulated calorimetric measurements, the multi-phase binders developed in the research were crosslinking and hence the glass transition regions were shifting over scanning. Overall, the determination of optimum curing conditions based on calorimetric measurements as performed in this study can assist the future material formulations and asphalt processing technologies to avoid the uncontrolled over-cure of epoxyasphalt binders and thus the negative effect of reduced mix workability during construction. In combination with the in-depth understanding of the chemical thermodynamics of binders with thermosetting nature, the material designers would be able to control various reaction-induced phenomena, such as the phase separation, after obtaining appropriate calorimetric information. Thus, the microstructure of cured epoxy-asphalt binders is possible to be predicted and desired miscibility of various incorporating phases to be achieved.