Glass Transition and Phase Stability in Asphalt Binders

Glass Transition and Phase Stability in Asphalt Binders Pavel Kriz* — Jiri Stastna* — Ludo Zanzotto* *Bituminous Materials Chair Schulich School of Engineering University of Calgary 2500 University Dr. NW Calgary, AB, T2N 1N4 Canada [email protected] [email protected] [email protected] ABSTRACT.

Major aspects of the glass transition of asphalt binders are described and an extensive literature review of the phenomenon and its relation to chemical composition is presented. The glass transition of asphalt binders was studied by modulated differential scanning calorimetry and also via dynamic mechanical analysis. A certain analogy between the glass transition of amorphous polymers and asphalts is suggested. The overall transition was found to be very broad on the temperature scale. The effects of evaporation of light-end components and oxidation on asphalt phase stability and glass transition were studied. It was suggested that phase incompatibility may exist in asphalts; however, the phase separation is observable after long-term isothermal conditioning at a temperature within the glass transition range. Based on the presented results, it is suggested that phase incompatibility develops if there is a discontinuity in the molecular distribution. Such discontinuity may be present in some neat binders as well as in severely oxidized or aged asphalt binders. KEYWORDS:

glass transition, modulated differential scanning calorimetry, asphalt binder, molecular mobility, physical aging, oxidative aging.

Road Materials and Pavements Design. Volume X – No X/2007, pages 1 to n

2 Road Materials and Pavements Design. Volume X – No X/2007

1. Introduction The glass transition is, perhaps, the most important factor that determines the viscoelastic properties of amorphous material (Williams et al., 1955, Ngai, 2004). It is a reversible change in an amorphous domain from a viscous or rubbery state to a hard and relatively brittle glassy state, and vice versa. The glass transition occurs when the characteristic time of molecular motions responsible for structural rearrangements becomes longer than the timescale of the experiment. The timescale for structural relaxation increases rapidly with decreasing temperature (Moynihan, 1994). Therefore, the glass transition is usually considered as a predominantly rate and non-equilibrium phenomenon. The transition to a glassy state is accompanied by a sudden change in the mechanical, optical and thermodynamic properties of the material. The increase in viscosity is enormous (up to 1013 Pa.s). The material becomes glossy in appearance and extremely brittle and rigid. The thermodynamic change during the glass transition upon cooling is recognized as the stepwise drop in heat capacity and loss in free volume (Ngai, 2004). Note, that the free volume is not a real volume measurable in cm3/g, but an operational quantity dealing with the molecular mobility (Struik, 1991). The glass transition temperature (Tg) is the temperature arbitrarily chosen to represent the temperature range over which the glass transition takes place: it is usually assigned to the temperature where half of the sample is already vitrified or devitrified (Wunderlich, 1994). In the current study, the glass transition of asphalt is thoroughly described and investigated by two techniques – modulated differential scanning calorimetry (MDSC) and dynamic mechanical analysis (DMA). The similarities between the glass transition of amorphous polymer and asphalt are suggested.

2. Background The glass transition is a function of molecular mobility. Upon cooling, molecular mobility decreases. Molecules are considered “glassy”, if their molecular motion has appeared to cease with respect to observation time. Different molecules have different degrees of motion at particular temperatures. The molecular motion is a function of molecular weight and structure, as well as of inter- and intra-molecular interactions. Before the glass transition is studied, the asphalt composition and intermolecular relations should be understood.

Glass Transition in Asphalt Binders 3

2.1. Molecular Continuity Asphalt consists of tens of thousands different chemical species (Speight, 1999). The whole scale of molecules ranges from non-polar fully saturated linear alkanes to highly polar polycyclic porphyrins or hetero-hydrocarbons. Each molecular functional group is essential to the phase stability of the asphalt system. It is, therefore, necessary to view the material as a continuum of molecules with a gradual transition in polarity, molecular weight and functionality. It is a generally accepted fact that the molecular functional groups overlap and, thus, are impossible to separate (Andersen et al., 2001). It has been reported (Carlson et al., 1958) that there is no sharp transition between asphaltenes and aromates. It has also been shown that part of the asphaltene fraction (based on the solubility parameter) may cosolubilize the rest of the fraction and, thus, behave as a resin (Heithaus, 1962, Andersen, 1995). There is no sudden transition in the molecular structural properties between resins and asphaltenes (Groenzin et al., 1999). Due to the complexity of the material and the interactions between high molecular weight molecules, the commonly used methods for fraction separation of asphalt cannot provide well-defined chemical fractions (Selucky et al., 1981). If part of the material is removed from the system during the separation, the molecular distribution and interaction are affected. The solubility of a substance is not a function of its general hydrocarbon skeleton and its chemical functionality alone, but also depends on interactions with other substances that act as co-solvents. Whether a given compound appears, for instance, as a resin or an asphaltene depends upon the presence or absence of other substances (Farcasiu, 1977). Separation into fractions has not proved to be a reliable predictor of asphalt performance on test roads (Goodrich et al., 1986). Indeed, asphalt refining is based on another type of fraction separation – distillation (Corbett, 1984). For the sake of simplicity, we may still use the terms “asphaltenes”, “resins”, and so on; but instead of relating these terms to solubility, we relate them to their functionality in the system. The asphaltenes represent the polyaromatic fraction, which tends to be insoluble in the continuous phase at certain conditions and needs to be stabilized (cosolubilized). This fraction consists of different molecules at different conditions (temperature, pressure, change in the continuous phase composition). Resins are unique aromates that are not qualitatively different from asphaltenes, yet restrict the asphaltenes from further association and, due to complex interaction, keep them in a single phase providing the molecular transition between the polar and non-polar ends. And finally, the continuous (or oil) phase, with composition gradually ranging from semi-polar resins to non-polar paraffins, is present.


2.2. Asphalt Structure The concept of asphalt being a colloidal system was proposed by Nellensteyn (Nellensteyn, 1938). A colloidal system is an intermediate between a homogenous solution and a heterogenous system, with the particle size ranging from 1 nanometer to 1 micrometer (Hiemenz et al., 1997). There is strong experimental evidence that the biggest single molecules in asphalt do not exceed the molecular weight of 1000 g/mol (McKay et al., 1978, Groenzin et al., 2001), which corresponds to molecular diameters in the range of 1-2 nm (Katz et al., 1945, Groenzin et al., 1999). The most polar molecules usually associate into bigger particles. The size of these particles rarely exceeds 6 nm (van der Hart et al., 1990); however, if the system is destabilized, they may grow bigger and eventually separate from the system. The presence of molecules and aggregates bigger than 1 nm disqualify asphalt from being a homogenous solution. The upper limit of a colloidal system (1 μm) is generally met in asphalts, even though recent atomic microscopy experiments revealed different domains of a size exceeding 1μm (Masson et al., 2006a). The molecular interactions (especially between asphaltenes and resins) and the whole structure of the system are still topics for discussion. The lyophobic models, where the insoluble asphaltenes are peptized by the resins that provide a protective sheet, were actually motivated by early observations of the solubilizing power of resins (Mack, 1932, Nellensteyn, 1938, Eilers, 1949, Koots et al., 1975). This is inconsistent with the fact that the asphaltene precipitates can be re-dissolved in a good solvent, even in the absence of resins (Andersen et al., 1990, Andersen et al., 1991). Also, the asphaltene molecules are not many times larger than the size of the resin molecules, as they would logically have to be in order for many resin molecules to be able to fit around them and form the protective layer, similar to the water-surfactant-oil emulsion. The asphaltene and resin molecules are, in fact, not much different in size (Storm et al., 1995, Porte et al., 2003). Whether a particular molecule behaves as an asphaltene or resin always depends on actual conditions (temperature, pressure, surrounding molecules, etc.). 2.2.1. The Molecular Forces The interaction between the asphaltene molecules and/or resin molecules is crucial in order to keep the colloidal character of the material and prevent further aggregation towards bigger agglomerates, which would eventually lead to instability and potential phase separation. There is no reason to assume that the inter-molecular forces acting between asphaltenes, resins and the oil phase should be in any way different from those existing between other well-known organic molecules containing the same atomic groups. Inter- and intra-molecular forces affecting the phase stability and molecular structure are charge transfer forces, electrostatic interactions, van der Waals interaction, exchange-repulsion interaction, forces arising from induced dipole, and hydrogen bonding (Murgich, 2002).


2.2.2. Thermodynamics of the Micelle Formation and Further Association The formation and stability of molecular aggregates, such as micelles or disperse solids, present in asphalt are determined by the changes in the total free energy of the system. The binding of molecules is always an unfavorable entropic process (reduction in degrees of freedom) (Vinter, 1996). This implies that the change in enthalpy must be negative (Murgich, 2002), since only processes with negative change in free energy are spontaneous. This was proven experimentally, and the number of interaction sites on an asphaltene molecule was estimated to be between 1 and 2 and the heat of association in the range of 2-7 kJ/mol (Merino-Garcia et al., 2004a, Merino-Garcia et al., 2004b). The association of heavy aromatics into bigger entities, or even more pronounced structures, is an exothermic process. The association becomes more pronounced with decreasing temperature. At low temperatures, however, the increasing viscosity of the continuous phase impedes the molecular diffusion and, therefore, the association. The aggregates or micelles may further associate into more complex structures. In the case of a sufficiently high dissolving power of the medium, phase separation does not set in; and, with the accumulation of asphaltenes, internal structures (networks) may form (Rogacheva et al., 1980). The very complex distribution of the molecular aggregates (Sheu et al., 1992) suggests that asphalt cannot be described simply as a pure sol or gel system (Saal et al., 1940, Eilers, 1949) formed by solid asphaltene particles dispersed by resins, or as a simple and uniform micellar system similar to those found in aqueous solutions of pure surfactants. Diversity of asphalt molecules (in size, shape, polarity, etc.) disallows the organized structure (crystal) to be built, and most of the molecules conform in an amorphous organization (Murgich et al., 2001).

2.3. Parameters Affecting the Glass Transition 2.3.1. Molecular Weight The molecular weight of the glass forming molecules is an important factor in glass transition. The effect is not straightforward, and it is also dependent on the molecular structure. For linear molecules, the chain ends exhibit greater mobility over the repeat inner units, because they are bonded only on the one side. The shorter molecules have more mobile chain ends per volume unit than longer molecules and, thus, exhibit greater molecular mobility. Molecules with higher mobility (shorter chains) become motionless at a lower temperature than those with lower mobility (longer chains). This means that Tg increases with the increasing molecular weight (Fox et al., 1955). Contrarily, the limited data for small polymer rings surprisingly predicted the increase in Tg with decreasing molecular weight (Clarson et al., 1985). The free volume theory (Fox et al., 1955) is inconsistent with these results; however, one cannot omit the simple explanation that the free volume


decreases with decreasing ring size (molecular weight), as the rings become tighter and tighter. 2.3.2. Functional and Structural Groups The effect of various functional and structural groups on Tg may be very complex and difficult to estimate. In polymers, for instance, introduction of paraphenyl rings into the linear chain significantly increases Tg, due to fact that paraphenyl restricts the chain backbone rotation. On the other hand, introduction of a methylene group or oxygen (ether bond, C–O–C) into the backbone lowers Tg, due to an increase in flexibility of the chain. Introduction of side alkyl chains also significantly decreases Tg. It has been shown (Weyland et al., 1970, Krevelen et al., 1976) that Tg can be successfully predicted by summing the partial Tg (Tgi) for particular functional groups. Tgi is dependent on molecular weight and structure. Generally, saturated and linear molecules possess higher molecular mobility due to the rotational degree of freedom of the single bond. Therefore, saturated molecules vitrify at much lower temperatures than unsaturated molecules. Rings and multiple bonds, as well as heteroatoms (especially carbonyl groups), introduce a certain rigidity into the molecule and reduce the molecular mobility. For instance, Tgi of the carbonyl group is 964 K, while Tgi of the ether group is 232 K. The presence of sulfur in the chain (C–S–C, Tgi=234 K) may lower the overall Tg, while Tgi of the SO2 group (905 K) may do the opposite. Carbonyl and sulfonyl groups are often considered as products of oxidative aging of asphalts, and their contribution to Tg may be important in oxidized asphalts. The prediction of Tg using the method developed by Krevelen may be difficult for asphalt, due to the vast number of functional groups. It has been shown, however, that reasonable estimates of Tg can be obtained (Huynh et al., 1978). 2.3.3. Cross-linking and Molecular Interactions It is known in polymer science that introduction of cross-links into a polymer strongly affects the local segmental relaxation and, hence, the glass transition. The cross-links reduce the configurational degree of freedom, thereby increasing Tg. The increase in Tg can be understood by a possible decrease in the molecular motion, because the van der Waals interactions are replaced by shorter and firmer covalent bonds (Chang, 1992). In asphalt, no formation of new inter-molecular covalent bonds is expected at ambient conditions. Major interactions and associations between asphalt molecules are due to weak van der Waals forces, hydrogen bonds and other forces. The weak inter-molecular interactions in asphalt may reduce the configurational degree of freedom, at least to some extent, as the covalent crosslinks do, possibly contributing to an increase in Tg. One should not omit the fact that the association energies between the most polar molecules are in the order of 5 kJ/mol (Merino-Garcia et al., 2004a), while single covalent bonds (cross-links in polymers) are much stronger (approximately in the order of 180-450 kJ/mol (Israelachvili, 1985)).


2.3.4. Crystallinity Some asphalts may contain a certain amount of crystalline phase (Claudy et al., 1992); therefore, the effect of crystallinity on Tg should be also considered. Tg of amorphous phase may increase as well as decrease with the degree of crystallinity, depending on the relative density of the crystalline and amorphous phases. In most cases, the ordered crystalline phase possesses higher density; and, the molecular chains of the amorphous phase are entrapped in the crystalline lattice. Their mobility is reduced, and Tg is increased (Flocke, 1962, Bair, 1994). The effect of crystalline phase on the glass transition of asphalts was found to be significant (Kriz et al., 2007a). 2.3.5. Dilution The solvent power and Tg of the oily phase in asphalt are important factors in the overall Tg. In polymer science, the effect of diluents to lower or increase Tg is well known. Dilution with a solvent that has a lower Tg than the original material decreases Tg of the system. Essentially, solvent with a higher Tg than the original material increases the resulting Tg (Plazek et al., 1991). Furthermore, there are solutions in which the mobility of the solvent is increased by the presence of a polymer whose undiluted Tg is higher than that of the solvent. Free volume theory cannot explain these unpredictable effects, and the additional concept of intermolecular coupling was introduced (Santangelo et al., 1994). 2.3.6. Other Effects The effect of pressure on Tg is typically in the order of an increase by 20°C per 100 MPa of pressure (Ngai, 2004). Since the pressure applied on pavement by traffic is in the order of several hundreds of kPa or a few Mpa, the increase of Tg due to pressure would be in the order of several tenths °C. It has also been shown that, for thin films of polystyrene and other polymers on aggregate (silicone wafers), Tg of a particular polymer is strongly dependent on the film thickness (Fryer et al., 2000). However, this effect was observable for film thicknesses of 50 nm and thinner. Since the average binder film thickness in asphalt mix is in the order of μm (Frolov et al., 1983), the effect of the film thickness on Tg is probably negligible.

3. Experiment and Methods

3.1. Instruments A TA Instruments Q100 differential scanning calorimeter (DSC), equipped with the modulated temperature setup and liquid nitrogen cooling system, was used in


this study. Ultra pure helium (99.999 percent) was used to purge the experimental cell. The rate was 50 mL/min. Standard hermetic aluminum pans were used for all experiments. The sample mass was in the range of 7-10 mg. Sample pans were sealed under nitrogen. The thermal history of the sample, if not stated otherwise, was deleted prior the experiment. The procedure was as follows. The sample was heated up to 150°C and underwent isotherm for 15 minutes. The sample was then quenched to −100°C at a rate of 10°C/minute. The MDSC setup was developed and evaluated in our other papers (Kriz et al., 2007b, Kriz et al., 2007c). The modulation amplitude was 2°C; the modulation period, 60 seconds; and, the linear heating rate, 2°C/minute. The typical range was from −100 to 100 or 150°C. The glass transition temperature was assigned to a temperature where the reversible heat capacity reached half of the overall change. Table 1. Basic properties of asphalt binders used in this study Asphalt A

Asphalt B

AAC-11

AAV1

Crude oil source

Cold Lake

Ural Russia

Red-water

Alaska North Slope

Performance Grade (PG)

PG 58-282

PG 52-282

PG 58-16

PG 52-22

Penetration 25°C [dmm]

1623

1863

133

121

−25.82

−18.90

−23.59

−20.78

0.704

4.624

5.06

3.13

Tg (MDSC) [°C] Wax (% wt) 1

data obtained from the SHRP Materials Reference Library (Jones, 1993), except for the Tg determination. 2 AASHTO (American Association of State Highway and Transportation Officials) MP320 (Asphalt Institute, 1994). 2,3 data acquired from Bituminous Materials Chair at the University of Calgary. 4 UOP (Universal Oil Products) 46-85 method (UOP Inc., 1985). A Rheometric Scientific strain-controlled rheometer (ARES) was used for the dynamic mechanical analysis. The linear viscoelastic (LVE) region was determined as a linear region in the G ′ , G ′′ versus testing strain (Figure not presented). The dynamic frequency sweeps were run (0.1 to 100 rad/s) at a constant temperature and a constant strain within the LVE region. A TA Instruments Q500 thermogravimetric analyzer (TGA) was used. About 40 mg of binder was placed on standard platinum pan prior to the experiment. Each sample was first melted, and small droplets were prepared. These droplets were


stored in a freezer to minimize the evaporation and to ensure the same number of reheatings for each sample. The purging gas was either nitrogen (non-oxidative evaporation) or air (oxidation) at a flow rate of 60 mL/min. The heating rate was 20°C/min (from ambient to test temperature), and test temperatures varied between 60 and 200°C, depending on the experiment. The sample was held at the test temperature for 50-1440 minutes.

3.2. Asphalt Samples Four different asphalts were used in this study. Asphalt A was an asphalt from heavy Alberta crude oil (Cold Lake) of penetration grade 150/200, supplied by Husky Energy Inc. It represented a high-quality binder for low-temperature application. Asphalt B was an asphalt from Ural crude oil of penetration grade 150/200, supplied by SLOVNAFT Plc., Slovakia. It represented an asphalt binder produced from the crude oil typically processed in Central Europe. The other two binders were Asphalts AAC-1 (Redwater) and AAV (Alaska North Slope) from the original Strategic Highway Research Program (SHRP) Material Reference Library. These three binders had been stored in bulk (15kg) at room temperature for 15 years.

4.Results and Discussion

4.1. Dependence of the Glass Transition on Observation Time According to the definition of a glassy state, the domain is considered glassy if the molecular motion appears as nonexistent within the observation time. This means that the glass transition is a kinetic process, and it is dependent on the particular experiment’s observation time. Figure 1 presents a master curve constructed from several isothermal dynamic data measured for Asphalt A. The time-temperature superposition principle (tTs) was applied (Ferry, 1961), and horizontal shift factors ( aT ) were determined. Shift factors were fitted with the Williams-Landel-Ferry equation (WLF): log aT = − c1 (T − Tr ) / (c 2 + T − Tr )

[1]

where, c1 and c2 are the empirical fitting parameters, T is the temperature, and Tr is an arbitrarily selected reference temperature. The frequency where the loss modulus ( G ′′ ) attains its maximum value is usually assigned as the glass transition frequency (Wada et al., 1959). With the use of tTs, the frequency can be converted to temperature:


ω = aT ω

[2]

where, ω is the reduced frequency, and ω is the testing frequency. The combination of Equations 1 and 2 yields:

Tg = Tr +

(

c1 log ω g / ω

(

)

c 2 + log ω / ω g

)

[3]

where, ω g is the frequency at the maximum of G ′′ , and Tg is the glass transition temperature. Using Equation 3, the glass transition temperature of the material can be calculated as a function of testing frequency (Kriz et al., 2008). Results for Asphalt A are presented in Figure 2.

Figure 1. Asphalt A – master curve, time-temperature superposition, test temperatures from −30 to 80°C, G ′ (○), G ′′ (●), tan δ (−). Constant strains within the linear viscoelastic region (LVE) were used. Geometry varied according to the viscosity of the specimen: 50mm cone and plate (T>50°C), 25mm cone and plate (T~20-50°C), 10mm plate-plate (T~0-20°C) and rectangular torsion bar (T