Understanding Bitumen Aging through Interrelationships and Aging Ratios

Abstract

Aging of bitumen can lead to fatigue and non-load-associated cracking of asphalt pavements. Performance grade specifications include threshold limits for rheological properties from specific ranges of binder behavior after laboratory aging. Some specifications include aging ratios (ARs) to limit the rate at which properties deteriorate. This study compares the evolution of interrelationships and ARs of performance parameters in specific areas in Black Space to understand the relative susceptibility of the binder behavioral spectra to aging. Unmodified and polymer modified binders are aged with the rolling thin film oven and pressure aging vessel, and tested with the dynamic shear and bending beam rheometers. It is seen in the study that there are both related and unrelated aging trends in the distinct stiffness-related ranges. Correlations in parameter development significantly improve when binders are grouped by similar features such as modification type or origin. Binder source and modification type and degree greatly dictate aging trends in the high-stiffness range. In the low-stiffness range, the initial hardness of the binder is indicative of aging potential. The evolution of properties may be influenced by either modification or binder source in the intermediate domain. When considering ARs for specifications, combining complementary ARs may provide reciprocal insights into binder aging.

Aging of bitumen can lead to fatigue and load-associated cracking of asphalt pavements ( 1 ). To ensure sufficient in-field behavior and durability, performance grade (PG) specifications include threshold limits for rheological properties from specific ranges of binder behavior after laboratory aging. Some specifications include aging ratios (ARs) to limit the rate at which properties will deteriorate ( 2 ).

Black Space has been used to monitor bitumen aging and performance ( 1 , 3 , 4 ). The complex shear modulus, G*, is plotted versus the complex phase angle, $\delta$ . Specific areas in Black Space are associated with degrees of viscoelastic behavior. This study compares the development of performance parameters in each spectrum and their ARs to understand relative susceptibility of the binder behavioral spectra to aging and the influencing factors.

Molecular Structure and Aging of Bitumen

Bitumen is divided into two groups of primarily hydrocarbon molecules, that is, asphaltenes and maltenes, without a finite boundary between them. Asphaltenes are insoluble in n-heptane, highly polar solids of relatively high molecular weight. Maltenes have lower molecular weights and are divided into polar resins, and non-polar saturates and aromatics ( 5 ). Redelius and Soenen note the presence of trace amounts of metals such as vanadium, nickel, and iron ( 6 ).

Models exist that attempt to mimic the structure of bitumen molecules. These vary in complexity and reasoning. Some accept the generalization that bitumen is a colloidal suspension with higher molecular weight asphaltenes dispersed in lower molecular weight maltenes. The maltenes form the continuous phase. The polar resins will act as the dispersing agents, while the solvents will inhibit the solvating strength of the maltenes ( 5 ).

Two boundary types of bitumen composition, the sol-type and gel-type, are commonly accepted. In the sol model, asphaltenes are completely dispersed and mobile within the maltenes. In the gel model, the maltene fraction is not sufficient to accomplish asphaltene distribution. The asphaltenes are allowed to associate and form irregular, disconnected structures. The gel behavior of bitumen decreases as it is heated. Most bitumens behave somewhere between the two ( 5 ).

Redelius and Soenen stress that, although bitumen is often analyzed in its fractional components (saturate, aromatic, resin, and ashpaltene [SARA]), the behavior of bitumen is not the sum of the parts ( 6 ). They characterize bitumen as a continuum of unique and intricate molecules with complex interactions rather than a colloidal suspension. The size of the largest molecules is dependent on the crude source, while the size of the smallest molecules is dependent on the cut point of the specific refinery. The larger the molecules, the higher the molecule interactions, and the higher the viscosity of the bitumen.

Physical properties and performance of bitumen are influenced by three main molecular interactions ( 6 ):

• London dispersion forces: This interaction is most important for larger, non-polar asphaltene molecules. Temporary dipoles are induced that can dominate bitumen behavior.

• Polar and hydrogen bonds: Although these interactions are important for smaller molecules, they play a lesser role in the interactions of larger bitumen molecules. Polar interactions correlate with the elastic stiffness component of bitumen.

• Pi-pi forces: These interactions are caused by the different shapes of the molecules, which are unlimited.

Redelius and Soenen describe the behavior of asphaltenes during aging and cooling as precipitation (versus association) ( 6 ). The degree of precipitation depends on the insolubility of the asphaltenes and the solving power of the maltenes.

This theory is supported by the findings of Bukka et al.: fourier transform infrared (FTIR) analysis of two binders with an inverse relationship of asphaltene content and viscosity revealed that, although a similar amount of oxygen was present, the distribution of the oxygen content in aromatic portions was a significant indicator of viscosity ( 7 ).

In some ways, there are similarities in the opposing views. Hunter et al. state that the degree of dispersion of asphaltenes, along with the viscosity of maltenes, will ultimately determine the viscosity of the bitumen ( 5 ). They agree that research (such as Christensen et al.) has shown that increased asphaltene content leads to increased viscosity at a constant temperature ( 8 ). Bukka et al., and Redelius and Soenen state that larger proportions of asphaltenes alone do not produce larger viscosities ( 6 , 7 ). Bukka et al. also apply the sol-gel concept; Redelius and Soenen do not ( 6 , 7 ). Redelius and Soenen do concede that it is possible that some molecules in the asphaltenes, such as the metal groups, may be completely insoluble and form associations ( 6 ).

Bitumen hardens with age because of an interaction with the oxygen in the air. Three hypotheses are proposed where the reaction with oxygen causes increased:

• Molecular size,

• Polarity, or

• Aromaticity (which leads to additional pi-pi and dispersive interactions) ( 6 ).

Redelius and Soenen conclude that physical properties, such as viscosity and stiffness, are highly dependent on molecular interactions ( 6 ). These interactions change over time because of chemical reactions with oxygen. Properties such as viscosity relate well to molecular weight and aromaticity.

Performance Measurement of Asphalt Binders

Because of the chemical complexity of bitumen, engineers often use mechanical test properties to measure performance and durability for road applications. Performance specifications measure specific rheology indicators to determine the suitability of a binder for environmental and traffic conditions ( 2 , 9 ). These specifications are based on the principal that controlling fundamental properties can prevent or limit failure. King et al. note that, although aging can lead to increased risk of fatigue cracking, damage caused by thermal, non-load-associated cracking is also induced ( 1 ). Table 1 indicates common failure mechanisms of bituminous surfacing applications and the parameters that may be used to measure binder resistance or susceptibility.

Table 1.

Stiffness Ranges, Failure, and Performance Parameters

Stiffness range	Associated failure	Specification and durability parameters*
Low	Rutting	High-temperature grade (Tmax), non-linear creep compliance ( $J_{\mathrm{nr}}$ )
Intermediate	Load-associated fatigue cracking	Intermediate stiffness ( $G_{\mathrm{int}}$ ), Glover–Rowe parameter (G-R), viscoelastic transition temperature (Tvet), viscoelastic transition stiffness (Gvet)
High	Non-load-associated, thermal cracking	Low-temperature grade (Tmin), critical low-temperature difference ( $\Delta T_c$ )
All	Premature failure	Aging ratios (ARs)

*

Please refer to Table 1 for definitions of the abbreviations as used throughout the rest of the paper.

Binder Performance Grade (PG)

PGs aim to ensure sufficient high-, intermediate-, and low-temperature performance during service life. Table 2 gives an example of the PG specifications currently used in South Africa. PG definitions are based on the minimum and maximum expected road temperatures, $T_{\max }$ and $T_{\min }$ . Where the maximum temperature is based on the highest 7 day average, the minimum is determined from the single lowest temperature ( 2 ).

Table 2.

Performance Grade (PG) Bitumen Specification Example ( 2 )

a Test property	Requirement
Max design temperature (°C)	b $T_{\max }$
Min design temperature (°C)	b $T_{\min }$
G* and $\delta$ @ $T_{\mathrm{int}}$ (°C)	Report only
G*/sin $\delta$ @10 rad/s @ $T=T_{\max }$	Report only
Viscosity ( $\ge 30\mathrm{se}{c}^{-1}$ ) @ 165°C	$\le 0.9$
Storage stability (G* ${\%}_{diff}$ at $T_{\max }$ )	$\le 15$
Flash point (°C)	$\ge 230$
After e rolling thin-film oven (RTFO) aging
G* and $\delta$ @ $T_{\mathrm{int}}$ (°C)	Report only
Mass change (% m/m) ≤	S: 0.3, H to E: 1.0
c $J_{\mathrm{nr}}$ at $T_{\max }$ ( ${\mathrm{kPa}}^{-1}$ ) ≤	S: 4.5, H: 2.0, V: 1.0, E: 0.5
$G_{\mathrm{RTFO}}^{*}/G_{\mathrm{Original}}^{*}$	$\le 3.0$
After RTFO plus f pressure aging vessel (PAV) aging
G* and $\delta$ @ $T_{\mathrm{int}}$ (°C)	Report only
d S < 300 Mpa, m > 0.3	$T_{\min }$
Test temperature @ 60 s (°C)
b $\Delta T_c{(}^{°}C)$	$\ge -5$
$G_{\mathrm{PAV}}^{*}/G_{\mathrm{Original}}^{*}$	$\le 6.0$

a

Test properties related to G* and $\delta$ are determined according to ASTM D 7175 with a dynamic shear rheometer (DSR) ( 10 ).

b

The continuous grade of a binder may be determined according to ASTM D 7643 ( 11 ). In South Africa, standard PG grades are 58-22, 64-16, and 70-10. T _min is determined with ASTM D 6648 and $\Delta T_c$ with ASTM D 7643 ( 11 , 12 ).

c

$J_{\mathrm{nr}}$ at $T_{\max }$ (ASTM D 7405) determines the traffic grade as defined by Bredenhann et al.: S = standard; H = heavy; V = very heavy; E = extreme ( 2 , 13 ).

d

Flexural creep stiffness, S, and the log slope, m, are determined using the bending beam rheometer (BBR) and ASTM D 6648 ( 12 ).

e

Rolling thin-film oven (RTFO) test represents in-plant aging of asphalt and is done in accordance with ASTM D 2872 ( 14 ).

f

Pressure aging vessel (PAV) test represents in-field aging is done in accordance with ASTM D 6521 ( 15 ).

Rheological properties are tested to determine binder characteristics that are expected to control in-field performance. Dynamic shear rheometer (DSR) testing is used to determine the complex shear modulus, $G^{*}$ , phase angle, $\delta$ , and non-linear creep compliance, $J_{\mathrm{nr}}$ , under oscillatory loading at intermediate and high temperatures, while the bending beam rheometer (BBR) determines the normal stiffness and slope of the stiffness graph under monotonic loading at low-temperature conditions. The intermediate temperature [T_int = (T_max+T_min) / 2 + 4] and stiffness ( $G_{\mathrm{int}}$ ) represent the general operating conditions in which repeated loading would lead to fatigue failure.

It is important to distinguish between the specification grade, that is, environmental requirements, and the continuous grade of a binder ( 11 ). In this paper, T_max,cont, T_min,cont, and T_int,cont [(T_max,cont+T_min,cont) / 2 + 4] denote the continuous equivalent of T_max, T_min, and T_int. For example, a binder with T_max,cont = 67.65°C and T_min,cont = -25.17°C has T_int,cont = (67.65 - 25.17) / 2 + 4 = 25.24°C.

Critical Low-Temperature Difference

$\Delta T_c$ quantifies the low-temperature cracking resistance of a binder ( 11 ):

$$\Delta T_c=T_{\mathrm{c},\mathrm{S}(60)}-T_{\mathrm{c},\mathrm{m}(60)}$$

(1)

with S and m as defined in Table 2, and the critical temperatures where specification limits are reached, interpolated as ( 16 ):

$$T_{\mathrm{c},\mathrm{S}(60)}=T_1+\frac{\log 300-\log S{(60)}_1}{\log S{(60)}_1-\mathrm{logS}{(60)}_2}·(T_1-T_2)-10$$

(2)

$$T_{\mathrm{c},\mathrm{m}(60)}=T_1+\frac{0,3-m{(60)}_1}{m{(60)}_1-m{(60)}_2}·(T_1-T_2)-10$$

(3)

where the current specification limits are defined in Table 2, and $T_{\min }$ is the highest of $T_{\mathrm{c},\mathrm{S}}$ or $T_{\mathrm{c},\mathrm{m}}$ . $T_1$ and $T_2$ are the two isotherms to each side of the compliance limit. $T_1$ and $T_2$ are not necessarily the same for S and m.

$\Delta T_c$ indicates the difference in the low-temperature grade between S and m. When $\Delta T_c>0$ , the binder is S-controlled, and when $\Delta T_c<0$ , the binder is m-controlled, that is, $T_{\mathrm{c},\mathrm{m}}$ is reached before $T_{\mathrm{c},\mathrm{S}}$ . As m indicates the ability of the binder to relax stress, it is undesired that this is the limiting factor to the performance of the binder, that is, it is preferred that the binder is S-controlled. Interestingly, $\Delta T_c>0$ has been correlated to the rheological index, R < 1.82 ( 17 ).

Aging Ratios (ARs)

ARs indicate the rate at which binder properties will deteriorate from their original state. They are often expressed in relation to the stiffness measured with increasing binder ages (Figure 1):

$$A{R}_{\mathrm{RTFO}}=\frac{G_{\mathrm{RTFO}}^{*}}{G_{\mathrm{Original}}^{*}}$$

(4)

$$A{R}_{\mathrm{PAV}}=\frac{G_{\mathrm{PAV}}^{*}}{G_{\mathrm{Original}}^{*}}$$

(5)

The South African PG specification shown in Table 2 requires ARs as determined at the intermediate temperature ( $T_{\mathrm{int}}$ ) and 10 rad/s.

Figure 1.

Age-hardening of binders.

Other Durability Parameters for Binders

Durability relates to the ability of a binder to maintain its performance and, therefore, its ability to resist the effect aging has on beneficial properties ( 18 ). Figure 1 shows the expected change in properties over time. Aging indicators include an increase in G* and viscosity or a decrease in ductility. Durability parameters thus aim to evaluate the evolution of rheological properties over time in in-service road conditions. Table 1 also include parameters not commonly included in PG specifications, but that have been proven by research to measure the performance of bitumen.

Ductility and the Glover–Rowe (G-R) Parameter

Ductility is measured according to ASTM D 113 and defined as the distance to which a binder sample elongates before it fractures when pulled apart at a specific temperature and speed (cm/min) combination ( 19 ). Although empirical, research has suggested that ductility represents a measure of the state of the colloidal system, or molecular interactions, in the binder and the shear susceptibility at a specific temperature ( 20 ).

Kandhal recovered binders from a range of asphalt surfacings and found a correlation between the ductility of the binder and the visual condition of the pavements ( 20 ). For binders with similar penetrations, those with lower ductility are more susceptible to cracking than those with higher ductility.

Generally, ductility is measured at low temperatures, not only to establish the resistance to thermal cracking but because of the consistency of binders at those temperatures. Because of the nature of binder deformation, ductility measurements at temperatures above 25°C have low reproducibility and little meaning ( 20 ).

Ductility is an efficient means to rank binder condition and performance. Because of its empirical nature and binder requirements (75 g), surrogates for this test method have been investigated and developed. Glover et al. investigated the use of a Maxwell mechanical model to describe the extensional flow a binder undergoes during a force-ductility test ( 21 ).

The model combines a spring and a dashpot element in series to represent elastic (G) and viscous ( $\eta$ ) responses, respectively. The model replicates binder behavior when subjected to a force: internal stress builds as a result of the immediate elastic response, and, as the internal stresses build, viscous flow commences, lowering these stresses. Although the model can describe the curvature of stress versus deformation, it lacks quantitative accuracy ( 21 ).

Displacement described by the Maxwell model is dependent on stiffness, G, and the ratio between G and viscosity, $\eta$ . Since the in-phase components are more sensitive to changes as the binder ages, Glover et al. examined the correlation between ductility, $G\prime$ and $\eta \prime /G\prime$ ( 21 ). They found that ductility is dependent on the combination of the two parameters. From this, the Glover parameter is defined as:

$$\frac{G\prime }{\eta \prime /G\prime }$$

(6)

where $G\prime$ and $\eta \prime$ are measured with a DSR at 0.005 rad/s (as this corresponds to an elongation rate of 1 cm/min) and the ductility cracking limits of 5 and 3 cm are equivalent to $9·{10}^{-4}$ and $3·{10}^{-3}$ MPa/s ( 21 ).

For modified binders, the relationship between ductility and the Glover parameter is less defined. The measure to which ductility improves is dependent on the type of modifier and the interaction with the binder. However, polymers degrade with age, decreasing the benefit to ductility. Glover et al. note that polymer degradation could even contribute to a loss of ductility of the binder, as unaged stress elongation, influenced by both the binder and polymers, changes when the polymers are no longer active ( 21 ). When comparing the Glover parameter with ductility, the general correlation improved when grouped according to a similar type. The differences between the unmodified binders, however, converge with age.

Rowe et al. noted that the Glover parameter could be reduced as follows ( 22 ):

$$\begin{array}{c}\eta \prime =\frac{G″}{\omega }\hfill \\ G\prime =\frac{G″}{\tan \delta }\hfill \end{array}$$

thus

$$\begin{array}{c}\multicolumn{1}{c}{\frac{\eta \prime }{G\prime }=\frac{G″}{\omega }·\frac{\tan \delta }{G″}}\\ \multicolumn{1}{c}{=\frac{\tan \delta }{\omega }}\end{array}$$

and

$$\begin{array}{c}\multicolumn{1}{c}{\frac{G\prime }{\eta \prime /G\prime }=\frac{G\prime \omega }{\tan \delta }}\\ \multicolumn{1}{c}{=\frac{G^{*}\cos \delta }{\tan \delta }·\omega }\\ \multicolumn{1}{c}{=\frac{G^{*}·\stackrel{2}{\cos }\delta }{\sin \delta }·\omega }\end{array}$$

As the test is performed at a constant frequency, the G-R parameter is then defined as ( 22 ):

$$\mathrm{Glover}-\mathrm{Rowe}=\frac{G^{*}·\stackrel{2}{\cos }\delta }{\sin \delta }$$

(7)

with limiting values equivalent to 5 and 3 cm ductility, of 180 and 600 kPa, respectively ( 20 ).

Similarities in G-R and $\Delta T_c$ are noted. Both are a measure of stiffness (i.e., G and S) and relaxation (i.e., $\delta$ and m) but at different operational temperatures (i.e., intermediate and low). For this reason, Mensching et al. explored the possibility of a low-temperature G-R equivalent for $\Delta T_c$ using a study of Rowe indicating when S ≤ 300 MPa, G ≤ 111 MPa, and when m ≥ 0.3, $\delta \ge$ 26.2° ( 4 , 23 ). They found some variance in the low-temperature grade, though both correlated well with field performance. The authors note that the selection of a temperature-frequency combination relevant to in-service performance should be considered carefully.

Viscoelastic Transition

The viscoelastic transition (VET) is associated with $\delta$  = 45°. It is characterized with the crossover modulus (G_vet), temperature (T_vet), and frequency ( ${\omega }_c$ ).

G_vet is independent of temperature and frequency. It is simply the stiffness of the binder when G' = G". Widyatmoko et al. determined T_vet at a loading rate of 0.4 Hz ( 24 ). More recently, Porot and Eduard, and Garcia Cucalon et al. used 10 rad/s as this loading rate is commonly applied in specifications ( 25 , 26 ). The ${\omega }_c$ is temperature dependent and is usually reported at the master curve reference temperature ( $T_{\mathrm{ref}}$ ) ( 27 ).

As a binder ages, G_vet and ${\omega }_c$ decrease, while T_vet increases ( 24 – 28 ). Christensen and Anderson, Widyatmoko et al., and Porot and Eduard relate this to hardening and reduced temperature susceptibility ( 24 , 25 , 29 ). Farrar et al. noted a correlation between G_vet and oxygen content ( 30 ). Widyatmoko et al. found a lower T_vet was associated with greater cracking resistance ( 24 ). Garcia Cucalon et al. further noted interrelationships between T_vet and G-R ( 26 ).

Black Space Diagram

A Black-Space diagram plots G* versus $\delta$ , eliminating the need for the time-temperature superposition principle ( 27 , 31–33). Figure 2 indicates the expected changes in binder behavior with increased aging. Figure 2a specifically highlights various temperature-associated regions. Figure 2b shows the expected progression of the G-R parameter of an aging binder in Black Space from a softer binder with a higher $\delta$ (viscoelastic behavior) to a stiffer binder with a lower $\delta$ (elastic behavior).

Figure 2.

Age hardening of binders in Black-Space: (a) mechanical test data in Black Space and (b) evolution of Glover–Rowe parameter (G-R).

The G-R parameter is helpful to monitor the combined effect aging has on both the stiffness and relaxation properties of binders at in-service temperatures. King et al. associate the 180 and 600 kPa boundaries for G-R with areas (i.e., combinations of G* and $\delta$ ) where cracking could be present, and Mensching et al. suggest that different areas in Black Space could apply to different failure types ( 1 , 4 ).

Research Approach

When considering (i) observations of ductility and the influences of modification and its deterioration by Glover et al., (ii) the distinct evolution of binder properties in each temperature range presented in Figure 2a, and (iii) the established development of G-R and associated performance boundaries shown in Figure 2b, the authors ask (a) how the changes in each region relate, (b) whether the development of parameters in each region follow similar trends observed for G-R, and (c) whether the observations of these parameters improve when specific groupings (such as modification type and source) are considered ( 21 ).

To investigate these questions, neat binders are artificially aged and tested with the DSR and BBR at various loading rate and frequency combinations ( 10 , 12 , 14 , 15 ).

Parameters from each temperature region are evaluated to determine interrelationships and the change in bitumen stress-strain response with increased aging:

• High-temperature: $T_{\max ,\mathrm{cont}}$

• Low-temperature: $T_{\min ,\mathrm{cont}}$ and $\Delta$ $T_c$

• Intermediate-temperature: VET, G-R, and $G_{\mathrm{int}}$

• ARs for all parameters

Material Selection and Procurement

Unaged, original binders originate from refineries and suppliers across South Africa and are subjected to artificial aging processes. Various ages are included to establish trends in parameter development.

Binders are considered based on:

• Binder type: Binders are chosen to represent those commonly applied in industry. The study includes 70/100, S-E1, SC-E1, S-E2, and SC-E2. Three asphalt binders are available as a part of the larger research initiative. These are included in the analysis to indicate possible differences in aging of modification types: 50/70, A-E2, and A-P1. (S, A, C, E, and P represent spray grade, asphalt, emulsion, elastomer-modified, and plastomer-modified, respectively. 1 and 2 indicate the degree of modification) ( 34 ).

• Origin: Binders are sourced either directly from suppliers or road-building projects. Those obtained from suppliers are selected to ensure diversity in base binders and modification. These binders originate primarily from the Western Cape (WC), Gauteng (GP), and Kwa-Zulu Natal (KZN) provinces. Where original binders are sourced from specific projects in specific provinces, the original binder is labeled from the province the road project is located in, that is, the Eastern Cape (EC), Limpopo (L), or Free-State (FS) provinces.

• Ages: Binder ages are chosen with the intent to represent ages along the full spectrum of surfacing life. RTFO-aging is specified in the South African PG specification for all binders, that is, both asphalt and chip seal binders. For this reason, the majority of binders are subjected to RTFO testing before PAV aging. However, since chip seal binders do not undergo in-plant aging represented by RTFO aging, some specimens are only aged with the PAV to consider possible disparities in aging. Artificial ages comprise: Original (unaged), RTFO, RTFO & PAV1, PAV2, RTFO & PAV2, PAV4, and RTFO & PAV4. (PAV1 indicates 20 h aging, and PAV4 indicates 80 h aging.)

In total, 92 binder samples are considered: 18 penetration grade, 49 modified spray-grade, 15 emulsion, and 10 asphalt-modified binders; 19 unaged and 73 binders between RTFO to RTFO & PAV4 aged were tested. For anonymous analysis, binders are labeled by their type and alphabetic markers, that is, S-E1-D, when the binder specific evolution of parameters are analyzed. Other projects that reference these binders are Engelbrecht, Van Zyl et al., and Tredoux et al. ( 35 – 37 ).

Testing

Aging

The RTFO test simulates plant aging, and the PAV simulates short-term in-service aging. For RTFO, ASTM D 2872 (unmodified) and TG 1 (modified) are followed, and for PAV, ASTM D 6521 is followed ( 14 , 15 , 18 ). (PAV aging is applied in multiples of 20 h, that is, PAV4 is equivalent to 4 times 20 standard PAV hours.)

DSR

The DSR is used to conduct strain and frequency sweeps. ASTM D 7175 is followed with equipment and test set-up as follows ( 10 ):

• Parallel 8 mm with 2 mm gap, and 25 mm with 1 mm gap

• Sample preparation: Silicone moulds

• The corresponding test variables, as well as variations of the specification, are as follows:

• Applied strain: For unaged binders, strain percentages as per Bredenhann et al. are used ( 2 ). For aged binders, strain sweeps are conducted between 0.01 and 10% strain. When the stiffness reaches 95 % of its initial value, the linear viscoelastic (LVE) limit is reached. If the LVE limit is less than 1%, that percentage strain is applied during frequency sweeps. Otherwise, 1% is used.

• Test temperatures: For most samples, the 8 mm geometry is used to test the specimen at 35°C, 25°C, 15°C, and 5°C, and the 25 mm geometry at 35°C, 45°C, 60°C, and 70°C. Strain sweeps are conducted at the lowest temperature for each geometry.

• Test frequencies: Strain sweeps are conducted at 10 radians/s, and frequency sweeps between 0.251 and 25.1 radians/s at each temperature.

• Sample preparation: Before testing, binders are heated at 165°C and poured into DSR moulds.

Bending Beam Rheometer (BBR)

The BBR is used to determine cold temperature stiffness and relaxation. ASTM D 6648 is followed with equipment and test set-up as follows ( 12 ):

• Sample preparation with steel moulds.

• Binders are tested at two to three temperatures, which are determined by the binder type.

The corresponding sample preparation and conditioning procedure is as follows:

• Binders are heated at 165°C for 25 to 30 min while steel moulds are prepared.

• After pouring, the binder is left to cool at room temperature for 60 min, and trimmed after 20 to 30 min. For harder binders buttering (refer to ASTM D 6648, note 10) is required ( 12 ).

• Before demoulding, the binder is placed in a fridge for 5 min.

• The demoulded beams are then left in the BBR bath for 60 min at the test temperature (± 0.1°C).

• Following testing, the specimen width and height is measured with an electronic calliper to the nearest 0.01 mm.

For specific samples, BBR tests could not be completed. For these binders, only T_max,cont is reported; other continuous and low-temperature parameters are not included in the data analysis.

Performance Parameters

For this study, DSR and BBR data are shifted to T_ref = 15°C. The modified Kaelble shift and generalized logistic master curve functions are used to calculate specific parameters where needed ( 28 ). The evolution of binder properties with aging is subsequently evaluated by examining the progression of performance and durability parameters:

• $T_{\max ,\mathrm{cont}}$ : ASTM D 7643 defines $T_{\max ,\mathrm{cont}}$ as the temperature at which G* / sin $\delta$ = 1.00 kPa ( 11 ).

• $T_{\min ,\mathrm{cont}}$ : ASTM D 6648 defines the low-temperature grade as the maximum between Equations 2 and 3 ( 12 ).

• $\Delta T_c$ parameter: This is investigated as a measure of low-temperature behavioral changes, and determined according to Equation 1 ( 11 ).

• VET parameters: T_vet is the temperature at which ${\omega }_c$ is equal to 10 rad/s. The shift from T_ref to T_vet is firstly determined by calculating the shift from ${\omega }_c$ to 10 rad/s. This shift is then substituted into the modified Kaelble equation and solved for the corresponding temperature (i.e., T_vet). G_vet is calculated from the master curve model at ${\omega }_c$ .

• ${\omega }_c$ : The crossover frequency is calculated from $\delta$ -isotherm data. An isotherm that crosses $\delta$ = 45° is used to interpolate to ${\omega }_c$ . Subsequently, the pairwise shift factors are used to calculate ${\omega }_c$ at T_ref.

• G-R parameter: G* and $\delta$ are calculated from master curve models at 15°C and 0.005 rad/s. G-R is subsequently calculated according to Equation 7.

• $G_{\mathrm{int}}$ : $G_{\mathrm{int}}$ is considered the binder stiffness at the intermediate temperature of its intended climatic (PG) grade. Three intermediate temperatures are considered, i.e., 22°C, 28°C, and 34°C. The corresponding stiffness, $G_{\mathrm{int},22}$ , $G_{\mathrm{int},28}$ , and $G_{\mathrm{int},34}$ , for each binder is determined from master curve models at a loading rate of 10 rad/s.

• ARs: ARs are calculated for each of the above performance parameters to evaluate relative aging for the load-response region of binder behavior it represents. Predominantly, ARs are calculated as:

$$A{R}_X=\frac{X_{\mathrm{aged}}}{X_{\mathrm{unaged}}}$$

(8)

where X represents the specific performance parameter. The exception is $\Delta T_c$ and $T_{\mathrm{vet}}$ :

$$A{R}_Y=\frac{Y_{\mathrm{aged}}-Y_{\mathrm{unaged}}}{|Y_{\mathrm{unaged}}|}$$

(9)

The variation is required because of the nature of $\Delta T_c$ and $T_{\mathrm{vet}}$ , represented above by Y. Some initial values are positive, while others are negative. If Equation 8 is applied, the resulting ARs will contain both positive and negative values for a decreasing value of $\Delta T_c$ , or an increasing value of $T_{\mathrm{vet}}$ . Instead, the change is normalized by the absolute value of the initial value to represent a relative change.

The relative susceptibility of binders to age in various regions of behavior is evaluated by considering interrelationships. The correlations between performance parameters are indicated by the Pearson correlation coefficient, $\rho$ . This calculation is done in Matlab with corrcoef, defined as:

$$\rho (X,Y)=\frac{1}{n-1}\sum _{i=1}^n(X_i-{\mu }_X)(Y_i-{\mu }_Y)$$

(10)

where

X and Y = correlated variables,

${\mu }_X$ and ${\mu }_Y$ = the means of each variable, and

n = the number of data points.

The summary of the experimental flow, including material selection, testing, and data analysis, is shown in Figure 3. It is noted that shift and master curve function optimization is not in the scope of this paper, but is reported by Goosen ( 28 ).

Figure 3.

Experimental and data flow design.

Evolution of Durability Parameters

Black Space Diagram

Black Space represents the entire range and domain of possible mechanical responses, at the particular binder age, regardless of loading rate and temperature. Parameters were chosen to represent various stiffness and relaxation spectra indicated in Figure 2a. The typical progression of G* versus $\delta$ in Black Space for unmodified and modified binders are shown in Figures 4 and 5. The figures indicate G*, $\delta$ , and temperature values associated with $T_{\max }$ , G-R, VET, and $\Delta T_c$ . (Corresponding predictive values for G* and $\delta$ models for optimized the GL models are also indicated for reference. For the most part, GL models can still accurately describe binder behavior to G* ≈  ${10}^4$ Pa.)

Figure 4.

Typical progression of Black Space diagrams for unmodified binders: (a) unaged, (b) rolling thin-film oven (RTFO)-aged, and (c) pressure aging vessel (PAV)-aged.

Figure 5.

Typical progression of Black Space diagrams for modified binders: (a) unaged, (b) rolling thin-film oven (RTFO)-aged, and (c) pressure aging vessel (PAV)-aged.

As binders age in Black Space, the high-temperature tail of modified binders straightens out to the right—an indication of decaying modifiers. For similar test temperatures, the curve as a whole flattens and moves to the upper left. As a result, VET and G-R move closer together and G*/sin $\delta$ = ${10}^3$ moves down the tail. In general, this confirms that elastic dominant behavior and the temperatures associated with each parameter increases. The differences between modified and unmodified binders decrease with increased aging.

Figure 6 shows the progression of G-R in Black Space. With an increase in age, G-R is expected to move from the lower right corner toward the upper left of Black Space. This movement is also known as the march of death as it is indicative of increased age and oxidation, and ultimately signals distress and possible failure. It is observed that the data converges significantly with age. Scattering for artificial ages in Figure 6a suggests that location after artificial aging depends on the initial position. Thus, when grouped by similar grade, that is, $T_{\max ,\mathrm{cont}}$ , in Figure 6b, cluster paths are more defined.

Figure 6.

Glover–Rowe (G-R) in Black Space: (a) artificial ages, (b) grouped by $T_{\max ,\mathrm{cont}}$ , and (c) grouped by @ $T_{\mathrm{int},\mathrm{cont}}$ per modification.

To investigate the scattering of data point with G* <  ${10}^5$ Pa, the influence of modification on the march of death is considered in Figure 6c. The data is grouped according the $T_{\mathrm{int},\mathrm{cont}}$ . Scattering for lesser-aged binders is significantly reduced. It is observed that all binder types approach the onset of cracking line as $T_{\mathrm{int},\mathrm{cont}}$ approaches 40°C. Remembering that the binders represent aged conditions, fixed G-R boundaries indicating the start of distress may not be accurate for all climates. A binder with a $T_{\mathrm{int},\mathrm{cont}}$ of 40°C is expected to also be less ductile than one with a $T_{\mathrm{int},\mathrm{cont}}$ of 34°C. These binders are thus expected to perform differently in 55-22 and 70-10 environments. Even the binder with a $T_{\mathrm{int},\mathrm{cont}}$ of 34°C will behave more brittle in the first mentioned, cooler climate. It may be prudent to investigate failure onset boundaries on specific climatic conditions.

Continuous Temperature Grades

Figure 7 shows the relationship between T_max,cont, the useful temperature interval, that is, T_max,cont - T_min,cont, and T_int,cont. In Figure 7a, the useful temperature interval is observed to be proportional to T_max,cont. The correlation implies that (i) the 80°C difference enforced by specifications for $T_{\max }$ ≥ 58°C is achievable for the binder sources considered, and (ii) T_min,cont increases at a lower rate than T_max,cont. It is reasonable to question whether molecules and interactions dominating the high-temperature behavior of bitumen are more sensitive to aging than those dominating low-temperature behavior.

Figure 7.

Development of continuous grades: (a) useful temperature interval and (b) $T_{\max ,\mathrm{cont}}$ versus $T_{\mathrm{int},\mathrm{cont}}$ .

The relationship between $T_{\max ,\mathrm{cont}}$ and $T_{\mathrm{int},\mathrm{cont}}$ shown in Figure 7b may appear counter-intuitive because of the observed differences in low and intermediate G* behavior, and specifically the associated $\delta$ values, in Figures 4 and 5. However, it should be considered that G* and $\delta$ are not directly evaluated, rather a ratio between the two, for example, G* = 1.00 · sin $\delta$ kPa, and the temperatures at which specific values are reached. Therefore, the temperatures associated with behavioral ranges in Black Space may indeed evolve proportionately, more so for specific binder types.

Pertinent interrelationships for continuous grade parameters are indicated in Figure 8. Figure 8a shows the temperature grades versus G-R. Performance boundaries associated with G-R are included. When considering the graph, it is useful to remember the movement of G-R away from high- and toward low-temperature behaviour in Black Space as aging increases. Subsequently, it is seen that the correlation of G-R and $T_{\max ,\mathrm{cont}}$ diverges as binders increase in age. For non-aged unmodified binders, or binders grouped by similar modifications, G-R could be an indication of binder temperature grade.

Figure 8.

Pertinent interrelationships for continuous grades: (a) continuous temperature grades versus Glover–Rowe (G-R) and (b) $T_{\mathrm{int},\mathrm{cont}}$ versus $T_{\mathrm{vet}}$ grouped by binder type.

The interrelationship with $T_{\mathrm{int},\mathrm{cont}}$ is most significant. For all binders and $\rho$ = 0.943, the values corresponding with G-R < 180 kPa and 600 kPa are $T_{\mathrm{int},\mathrm{cont}}$ < 37.84 and 42.09°C, respectively. Additionally, the correlation between $T_{\mathrm{int},\mathrm{cont}}$ and $T_{\mathrm{vet}}$ is highlighted in Figure 8b. Given recent research interest in G-R and VET, it is noted that the continuous grade, although more empirical, is similar to certain fundamental properties. It also implies that some information about the continuous grade and useful temperature interval may be gauged from G-R and $T_{\mathrm{vet}}$ .

Further to the above, the study distinguishes between the continuous intermediate temperature of the binder and the intermediate temperature of the climate, or PG grade. Temperatures associated with intermediate stiffness behavior of the binder will increase with time. As defined by the PG specification, the intermediate temperature of the climate will remain constant at either 22°C, 28°C, or 34°C. It may be interesting to investigate whether there is a relationship between the difference in $T_{\mathrm{int},\mathrm{cont}}$ and $T_{\mathrm{int}}$ , and fatigue failure.

Intermediate-Temperature Stiffness

Intermediate temperatures are defined per specification binder grade, specifically for 55-22, 64-16 and 70-10. Binder stiffness at intermediate temperatures of 22°C, 28°C, and 34°C, and a loading rate of 1 rad/s, is investigated as a measure of the expected in-service stiffness response. The specific intermediate temperature is dependent on the environment for which the binder has been chosen. In principle, it remains constant unless temperatures associated with specific regions are adjusted, for example, because of climate change. As binders age, it is expected that stiffness at intermediate temperature will increase, as will the associated risk of fatigue damage.

As expected, decreasing temperature causes increasing stiffness. For the binders considered, $G_{\mathrm{int}}$ correlated most significantly with G-R, $T_{\mathrm{vet}}$ , and ${\omega }_c$ (0.70 $<\rho <$ 0.91). Interestingly, correlations with $G_{\mathrm{vet}}$ were comparatively low (0.44 $<\rho <$ 0.65).

General correlations between the intermediate temperature stiffness responses and G-R in Figure 9 are similar to those of $T_{\mathrm{vet}}$ and ${\omega }_c$ . The interrelationships improved with increasing intermediate temperature, and were influenced by binder type and source. It implies that stiffness at 34°C and 10 rad/s is more aligned with 15°C and 0.005 rad/s than at lower intermediate temperatures. This may be expected, as the 35°C isotherms, for which test frequencies include 10 rad/s shown in Figures 4 and 5, are located closer the G-R parameters than the 25°C isotherms in Black Space.

Figure 9.

Intermediate stiffness versus Glover–Rowe (G-R): (a) G_int,22, (b) G_int,28, and (c) G_int,34.

Scattering observed for G-R < ${10}^4$ Pa becomes more pronounced with an increase in temperature. This may be expected, as modifiers become increasingly active at higher intermediate temperatures. In Figure 9b, the influence of binder type is considered. Here, it can be observed that degree of modification defines the interrelationship for less-aged binders. Unmodified binders indicate lower G-R values for similar $G_{\mathrm{int}}$ responses than highly modified S-E2 binders, or vice-versa. This implies that, for similar intermediate stiffness behavior, modified binders may indeed be further developed in relation to aging measured by G-R. Alternatively, for similar ductility, modified binders are more stiff than unmodified binders—assuming the relationship with ductility is valid at such low values for G-R. When approaching the G-R limits, the influence of binder type and modification appears increasingly insignificant.

When considering $G_{\mathrm{int}}$ at each temperature, it may be observed that stiffness associated with G-R limits increases as the temperature decreases. Although the phase angle would certainly also influence the observation, it is reasonable to expect that a binder with higher stiffness would withstand fewer load repetitions at T = 22°C than one with a lower stiffness response. Binder fatigue tests done at the different intermediate temperatures for various binder ages may give additional insight into whether G-R limits may have to be adapted for specific climates, or PG grades.

Critical Low-Temperature Difference

$\Delta$ $T_c$ is calculated from the differences in critical low temperatures of bending stiffness, S, and the associated log-slope, m. The parameter is an indicator of susceptibility to cracking at low temperatures. As $\Delta$ $T_c$ increases negatively, the binder increases in elastic brittleness.

General trends in development are considered in Figure 10. $\Delta$ $T_c$ correlates least with other parameters. When considering the evolution of $\Delta$ $T_c$ versus $T_{\min ,\mathrm{cont}}$ per neat binder, there is generally a consistent decrease in $\Delta$ $T_c$ as $T_{\min ,\mathrm{cont}}$ increases. For most binders, the degree of hardening appears to be dependent on the initial parameter value. Unmodified and asphalt binders indicate $\Delta$ $T_c$ initial values above 0°C and S-E2 binders below 0°C, while those of S-E1 binders vary. $\Delta$ $T_c$ versus $T_{\min ,\mathrm{cont}}$ data is superimposed in Figure 10a. The development in $\Delta$ $T_c$ versus $T_{\min ,\mathrm{cont}}$ is not significant when general trends are considered.

Figure 10.

$\Delta$ $T_c$ interrelationships: (a) $T_{\min ,\mathrm{cont}}$ by artificial age, (b) Glover–Rowe (G-R) by source, and (c) $G_{\mathrm{vet}}$ by binder type.

Figure 10b shows $\Delta$ $T_c$ versus G-R. When taking the source into account, the correlation notably improves for binders from WC and GP. This phenomenon certainly implies that the base binder, and perhaps the cut-off points of the refineries, may play a role in the observed distribution. The majority of binders from KZN are S-E1 binders. Considering the distribution of initial unaged $\Delta$ $T_c$ for S-E1 binders, this may be a contributing factor to the general lack of correlation.

Figure 10c indicates the correlation of $\Delta$ $T_c$ with $G_{\mathrm{vet}}$ . Both are shape parameters located to the upper left in Black Space, and this is reflected in their interrelationship—correlations with VET are significantly higher than with other parameters. The consideration of binder type improves the correlation, although it does not appear to influence the rate of change in $\Delta$ $T_c$ . Lesueur et al. and Goosen further discuss correlations between $\Delta$ $T_c$ , $G_{\mathrm{vet}}$ , and other shape parameters such as R ( 17 , 28 ).

From the above analysis, it appears that $\Delta$ $T_c$ may correlate with other parameters, but that this correlation is significantly dependent on both the source of the binder and the modification. The discussion of Redelius and Soenen is recalled ( 6 ). The precipitation during cooling is influenced by both larger asphaltene molecules and the solving power of the maltenes. These are influenced by the crude source and the cut-off point and refineries. It stands to reason, then, that cold temperature behavior and aging are dependent on the source of the binder.

Viscoelastic Crossover

The development of VET is shown in Figure 11. VET has been linked to oxidative-age hardening, reduced temperature susceptibility and binder distress ( 24 , 25 , 29 ). $G_{\mathrm{vet}}$ and $T_{\mathrm{vet}}$ are the stiffness and temperature associated with ${\omega }_c$ (at $T_{\mathrm{ref}}$ and 10 rad/s). It is expected that $G_{\mathrm{vet}}$ and ${\omega }_c$ decrease, while $T_{\mathrm{vet}}$ increases, with age. Essentially, this implies that, as a binder ages, the range of stiffness for which more brittle, elastic behavior is dominant will increase. In addition, the temperature at which relaxing, viscous behavior can counteract pure elasticity will increase as well. This increase in $T_{\mathrm{vet}}$ may be detrimental when it exceeds operational temperatures, as repeated loading may lead to cracking. $T_{\mathrm{vet}}$ above operational temperature implies the binder will not be able to operate outside of elastic dominant behavior. This occurrence may have a higher degree of damage implication compared with a $T_{\mathrm{vet}}$ below operational temperatures.

Figure 11.

Influences on the development of viscoelastic transition (VET): (a) artificial age, (b) source, and (c) binder type.

Viscoelastic Transition (VET)

The VETs in relation to $G_{\mathrm{vet}}$ and $T_{\mathrm{vet}}$ for all binders are shown in Figure 11. These observations correspond to literature where an inverse proportionality between stiffness and temperature is observed. The influence of source and binder type on the correlation between $G_{\mathrm{vet}}$ and $G_{\mathrm{vet}}$ is also shown. Mostly, the correlations are at least equal to the general, although most indicate a marked improvement. Only S-E2 binders show a reduced correlation, of which differences in the binder origin may be the cause.

As observed in Figure 11c, unmodified binders appear to tend toward the upper part of the data band. This occurrence suggests modified binders are slightly less stiff whilst transitioning from viscous to elastic-dominant behavior. The impact of this observation on construction should be considered and investigated further. The rate of change for all sources and binder types appears similar.

Crossover Frequency

The ${\omega }_c$ may be measured at any temperature when $\delta$ = 45°. It represents the theoretical crossover from viscous to elastic-dominant behavior. For this analysis, ${\omega }_c$ is represented at $T_{\mathrm{ref}}$ , that is, 15°C.

The VET is compared with ${\omega }_c$ in Figure 12. Generally, there is a strong correlation between $G_{\mathrm{vet}}$ and ${\omega }_c$ . The near-perfect correlation between $T_{\mathrm{vet}}$ and ${\omega }_c$ may be explained mathematically. $T_{\mathrm{vet}}$ is calculated from the shift function when ${\omega }_c$ = 10 rad/s. The shift function is derived from the time-temperature superposition principle in the LVE range of binders. Fundamentally, it follows that either a constant T =  $T_{\mathrm{vet}}$ or constant $\omega$ = ${\omega }_c$ will result in identical G* and $\delta$ master curves, and the two parameters are thus interchangeable.

Figure 12.

Viscoelastic transition (VET) versus ${\omega }_c$ : (a) G_vet and (b) T_vet.

Figure 13 exhibits significant interrelationships for ${\omega }_c$ . Figure 13a shows a consistent decrease in ${\omega }_c$ for an increase in $T_{\max ,\mathrm{cont}}$ . The progression of a binder depends on the initial parameter value and the binder type. Unmodified binders are mostly located to the lower side of the spectrum and asphalt modified binders to the upper side, whereas seal-modified binders are spread throughout. The initial observation suggests that aging rates for various binder types may be similar. This may only be the trend for the selected binders and should be investigated further in future studies.

Figure 13.

Pertinent interrelationships for ${\omega }_c$ : (a) $T_{\max ,\mathrm{cont}}$ and (b) Glover–Rowe (G-R).

G-R and ${\omega }_c$ also indicate a significant interrelationship in Figure 13b. It is observed that lesser-aged, modified binders follow a different trend to unmodified binders. This may be understood when considering Figures 4 and 5. For non-aged, modified binders, G-R is observed on the tail that curves inward, and ${\omega }_c$ on the straight upper portion of the curve. The two parameters are thus located on separate curvatures of the behavioral line. As these binders age, the behavioral line straightens out, and both ${\omega }_c$ and G-R are located on a single line of curvature. It may also be observed that G-R moved up the behavioral line as a binder ages. It could be considered that aging could be described by the distance between ${\omega }_c$ and G-R.

Glover–Rowe Parameter (G-R)

The general development of G-R in Black Space is indicated in Figure 6. G-R has been shown throughout the paper to correlate well with a significant number of rheological indicators. G-R also indicates notable interrelationships with VET as indicated in Figure 14. The specific evolution of G-R for individual binders is shown in Figure 15.

Figure 14.

Glover–Rowe (G-R) versus viscoelastic transition (VET): (a) $G_{\mathrm{vet}}$ by source and (b) $T_{\mathrm{vet}}$ by binder type.

Figure 15.

Development of Glover–Rowe (G-R) versus $T_{\mathrm{int},\mathrm{cont}}$ .

The interrelationship with ${\omega }_c$ , previously shown, was most significant. Correlations with $G_{\mathrm{vet}}$ in Figure 14 are not as defined, but improve when binder sources are considered. Correlations with $G_{\mathrm{vet}}$ are expected because of the significant interrelationships between VET parameters presented in Figures 12 and 13. The limits-related crack initiation and propagation is also indicated in interrelational plots.

In Figure 6c, grouping by $T_{\mathrm{int},\mathrm{cont}}$ represents G-R data points in distinct clusters as they progress in the march of death. It appears that the 180 kPa G-R boundary coincides with $T_{\mathrm{int},\mathrm{cont}}$ approaching 40°C. When considering the specific evolution of G-R for individual binders in Figure 15, this can be further refined to between 35°C and 40°C. None of the binder reached the 600 kPa boundary.

It is a further point of clarity that a single binder follows a particular trend. This distinction may be assumed to be caused by various aspects identified, including the base binder and not excluding modification products. Some distinction between binder types can be observed since their initial values tend to be in similar ranges. The slope also appears to converge with increased artificial age, in most cases possibly approaching an asymptote.

Aging Ratios (ARs)

Based on observations of general evolution and interrelationships, ARs are considered for parameters that indicate sensitivity to triggers of binder aging. Parameters are chosen to represent high-, intermediate-, and low-temperature or stiffness responses. Interrelationships are considered to assess relative susceptibility to aging at different regions of binder behavior.

It should be stressed that ARs normalize the parameters, and disregard the initial conditions. Thus, one binder may yield a greater AR than another, and still behave relatively less aged in absolute terms. Interrelationships between ARs point out where behavior is most susceptible to aging. The onus is then on the analyzer to determine whether the aging mechanisms will lead to premature failure.

Continuous Temperature Grade

The ratio of aged to initial $T_{\max ,\mathrm{cont}}$ is shown in Figure 16. This representation is indicative of the aging potential of binders in relation to hardness at high temperatures. For the laboratory ages considered, $T_{\max ,\mathrm{cont}}$ increases with a factor between 1 and 1.5. Trends for binders vary significantly. For binders with numerous samples, 70/100 binders indicated the least variable aging. This occurrence is consistent with observations of the evolution of the performance parameters. SC-E2 and S-E1 binders are the most variable in response. S-E1 binders exhibited the widest band of initial high temperatures, that is, between 58°C and 70°C. This scope suggests that the aging potential may also vary to a similar extent.

Figure 16.

$T_{\max ,\mathrm{cont}}$ AR versus artificial age per binder type.

Interestingly, S-E2 binders yields a considerably lower aging rate than other modified binders—this is also observed for the 50/70 and A-E2 binders, respectively. From literature, it is understood that bitumen approached a stiffness asymptote with age. This trend may translate to the equivalent high-temperature grade. Subsequently, a binder that is initially stiffer or aged may have lower aging potential.

Further, it may also be mentioned that, in South Africa, it would be assumed that SC-E2 emulsion binder is produced from a S-E2 base binder. Given the parallel between aging of the S-E1 and SC-E2 binders, this may be called into question.

When interrelationships in Figure 17 are considered, significant insights are gained. The AR of $\Delta$ $T_c$ appears to have an exceptionally poor correlation with that of $T_{\max }$ until further investigations are done into origins. Generally, binders from WC and KZN are significantly more susceptible to high- than low-temperature aging. Binders from GP correlate less, although most indicate greater susceptibility to low-temperature aging.

Figure 17.

High-temperature AR interrelationships: (a) $\Delta T_c$ , (b) $T_{\mathrm{int},\mathrm{cont}}$ , (c) useful temperature interval, and (d) Glover–Rowe (G-R).

The correlation in development between $T_{\max ,\mathrm{cont}}$ and $T_{\mathrm{int},\mathrm{cont}}$ observed in Figure 7b, is further confirmed in Figure 17b. Although $T_{\mathrm{int},\mathrm{cont}}$ increases at a lower rate, it is proportional to the increase in $T_{\max ,\mathrm{cont}}$ for both modified and unmodified binders. Similarly, the rate of increase in the useful temperature interval is proportional to that of the high-temperature parameter.

Binder type is expected to influence the aging and aging rate in Black Space between $T_{\max }$ and G-R. This is the region of binder behavior most influenced by modification. For lesser-aged binders, these two parameters are measured close to one another on the behavioral line. As a binder ages, G-R moves further away in the direction of the viscoelastic transition. In Figure 17d, a divergent correlation between the first-mentioned parameters with increased age is coherent with this trend. As modification increases, for example, from 50/70 to A-P1, the binders yield greater changes in $T_{\max }$ than G-R. This implies that, as modified binders age and modifiers deteriorate, they may indicate greater changes in high-temperature susceptibility than unmodified binders.

Intermediate-Temperature Stiffness

The interrelationship between G-R and $G_{\mathrm{int},34}$ is given in Figure 18. Generally, G-R grouped by binder type indicates the highest degree of correlation.

Figure 18.

Intermediate stiffness versus Glover–Rowe (G-R).

Trends in ARs for in $G_{\mathrm{int}}$ at the individual intermediate temperatures are similar. Correlations generally improve with an increase in temperature for $T_{\max }$ , $G_{\mathrm{vet}}$ , and G-R, and decrease for ${\omega }_c$ and $\Delta$ $T_c$ . A divergence in correlation is observed for an increased binder age at all temperatures. The relationships are significantly influenced by binder type, less so by origin. It is also observed that the differences between these are amplified as the intermediate temperature increases.

In Figure 18, the separation of modified and unmodified binders observed in Figure 17d is further noticeable. It could be argued that this distinction is because of the difference in behavior observed in Figures 4 and 5. The 35°C isotherm and G-R parameter in Black Space coincide with the deviation in the modified binder data between ${10}^4$  < G* <  ${10}^4$ Pa. As the modified binder ages, the deviation reduces and the trend straightens. This trend is not observed for unmodified binders, and could contribute to the observation in Figure 18.

Critical Low-Temperature Difference

The aging rate of $\Delta T_c$ versus artificial age is shown in Figure 19. Opposite to $T_{\max }$ , S-E2 binders indicate significant variability. $\Delta T_c$ previously indicated sensitivity to the origin, which may influence the observed inconsistency. The majority of binders indicate a low rate of increase in $\Delta T_c$ .

Figure 19.

$\Delta T_c$ aging ratio (AR) versus artificial age per binder type.

Interrelationships are considered in Figure 20. In general, $\Delta T_c$ indicates the least correlation with other ARs when the entire dataset is considered. However, when groupings in relation to binder type and origin are considered, the correlation often improves significantly. The comparison with $G_{\mathrm{vet}}$ is done for origin. Binders from WC and KZN are relatively less susceptible to aging in relation to $\Delta T_c$ . This difference in susceptibility suggests that origin, as well as the cut-off point and modification strategies in certain regions, will determine whether a binder is more susceptible to loss of intermediate- or low-temperature ductility.

Figure 20.

$\Delta T_c$ aging ratio (AR) interrelationships: (a) $G_{\mathrm{vet}}$ and (b) Glover–Rowe (G-R).

The influence of binder type on the correlation between aging rates of $\Delta T_c$ and G-R is significant and possibly the clearest of those considered thus far. There is an almost radial movement in trends. A change in G-R almost entirely dominates the aging of unmodified binders. Opposite, the aging of those which are expected to have the highest degree of modification, that is, S-E2 binders, is controlled by $\Delta T_c$ .

If details about modifications were available for S-E1 binders, it would be interesting to assess whether considering the degree of modification, or possibly type of modification, would further improve correlations. From a chemical perspective, it may also be valuable to investigate the molecular composition and its influence on low- and intermediate-temperature-aging susceptibility.

Viscoelastic Crossover

Viscoelastic Transition

ARs for $G_{\mathrm{vet}}$ versus artificial age are given in Figure 21. $G_{\mathrm{vet}}$ decreases with increased age. Subsequently, the ARs are below 1. Most seal binder types appear to approach a limiting decrease in the order of 0.9. Aging rates for $G_{\mathrm{vet}}$ are comparatively less variable than other parameters. This observation is consistent with previous findings.

Figure 21.

$G_{\mathrm{vet}}$ aging ratio (AR) versus artificial age per binder type.

Interrelationships are considered in Figure 22. Correlations with $T_{\mathrm{vet}}$ are marginally influenced by binder origin. Binders from KZN indicate a slightly higher susceptibility to aging in relation to $T_{\mathrm{vet}}$ than $G_{\mathrm{vet}}$ , compared with binders from WC. The high proportionality between ${\omega }_c$ and $T_{\mathrm{vet}}$ implies a similar correlation would be observed.

Figure 22.

G _vet aging ratio (AR) interrelationships: (a) T_vet and (b) Glover–Rowe (G-R).

The influence of binder type on the correlation between $G_{\mathrm{vet}}$ and G-R is similar to $\Delta T_c$ and G-R, although less exaggerated. Apart from the S-E2 and A-P1 binders, neither $G_{\mathrm{vet}}$ nor G-R dominates aging. One simply outweighs the other depending on what appears to be the degree of modification. When comparing this aging of the A-P1 binder with $\Delta T_c$ and G-R in Figure 20b, there is a significant difference. The A-P1 data point previously fell in the middle of the unmodified and highly modified binders. Here, it sides with the highly modified binders, as in Figure 17d. This difference could indicate that the type of modifier in the case of A-P1 leads to relatively more aging at the behavioral range that includes $T_{\max }$ and $G_{\mathrm{vet}}$ than the ranges including $\Delta T_c$ and G-R. This is an odd anomaly; it could be investigated whether the binder origin also plays a role.

Crossover Frequency

ARs for artificial binders in relation to log ${\omega }_c$ at $T_{\mathrm{ref}}$ = 15°C are shown in Figure 23. Ratios mostly range between 1 and −1. Some binder types appear to have a decreasing change in the aging rate. Those that do not may still have aging potential. A decreasing gradient of aging rate for ${\omega }_c$ appears rational, depending on the shape of the G* master curve, and the location of VET on it. When VET is located on the flatter slope of G* versus ${\omega }_r$ , a change in stiffness will translate to a relatively large change in frequency. If this point moves down the curve to the extent that it is located on the steeper gradient, a similar change in stiffness will relate to a relatively smaller change in frequency. As the point at which G' = G" moves down the G* master curve, the associated rate of change in ${\omega }_c$ should decrease up to the point of the inflection of the master curve. It should also be considered that artificial ages are not at linear intervals, and thus the apparent change in aging rate is simply qualitative.

Figure 23.

Crossover frequency aging ratio (AR) versus artificial age per binder type.

Notable interrelationships for ${\omega }_c$ are shown in Figure 24. Given the observed correlation between ${\omega }_c$ and $T_{\mathrm{vet}}$ in Figure 12b, it would be expected that the ARs compared with $G_{\mathrm{vet}}$ would evolve similarly. From Figure 22a, this is not the case. The difference in observation may indicate the problematic nature of calculating ARs for T, and perhaps $T_{\mathrm{vet}}$ as well, as shown in Equation 9. ARs for parameters that may have both negative and positive initial values should be carefully considered.

Figure 24.

Crossover frequency aging ratio (AR) interrelationships: (a) $G_{\mathrm{vet}}$ and (b) Glover–Rowe (G-R).

Similar to the observations with $T_{\mathrm{vet}}$ and $G_{\mathrm{vet}}$ , relative susceptibility to either G-R or ${\omega }_c$ aging, in Figure 24b, appears to depend highly on the degree of modification. Binders with a higher degree of modification appear more susceptible to aging at the crossover. Unmodified binders are more receptive to aging in relation to G-R. This difference implies that modified binders age more on the high intermediate portion of the behavioral curve, while unmodified binders age more in the lower intermediate domain.

G-R Parameter

Interrelationships that involve G-R are referred to throughout the discussion. ARs in relation to artificial ages are shown in Figure 25. Unlike previous parameters, which indicated a greater variability for modified binders, G-R is higher and more scattered for unmodified binders. Values range between 1 and 2 for modified and between 1 and 4 for unmodified binders. This observation supports previous findings that unmodified binders are more susceptible to age hardening on the lower intermediate portions of the behavioral line.

Figure 25.

Glover–Rowe (G-R) aging ratio (AR) versus artificial age per binder type.

Summary of Pertinent Observations and Discussions

The literature presented indicated molecular size and interactions highly affect mechanical binder properties such as stiffness and viscosity ( 6 ). Molecular size is influenced by the crude source and processing by refineries, while oxidative aging will both increase molecule sizes and affect their interactions.

Further literature has shown the relationship between ductility and G-R is distinct for modified and unmodified binders, and more distinguishable when similar binders are grouped ( 21 , 22 ). Polymer degradation is thought to contribute to loss of ductility. Differences in behavoir converged with age.

The authors set out to understand binder aging by asking how changes in distinct behavioral ranges in Black Space relate to each other; how these changes relate to known aging trends for G-R; and to what extent the observations are influenced by binder origin and modification.

The investigation was done by considering interrelationships in performance-related parameters development and ARs for various laboratory-aged modified and unmodified binders.

The aging of binder in Black Space is notably different for modified and unmodified binders. The decay of modifiers is observed in high-temperature behavior, as the tail straightens, and becomes similar to that of unmodified binders. The movement of isotherms indicates increasing elastic dominant behavior for operational temperatures. In general, parameters adjacent in Black Space indicate notable correlations.

Low-Stiffness Behavior

Aging in the low-stiffness or high-temperature behavior range is considered by analysing the development of $T_{\max ,\mathrm{cont}}$ . $T_{\max ,\mathrm{cont}}$ , and $T_{\mathrm{int},\mathrm{cont}}$ and the useful temperature intervals increase proportionately with age. Strong correlations between temperature parameters and G-R are observed. ARs indicate that $T_{\max ,\mathrm{cont}}$ increased between a factor of 1 and 1.5 for the laboratory ages evaluated. S-E1 and SC-E2 binders have the widest range of initial $T_{\max ,\mathrm{cont}}$ and age most variably in this behavior domain. Binders from specific sources indicated a higher degree of high-temperature than low-temperature aging. Harder, unaged binders, such as S-E2, indicate lower ARs, and thus less aging potential, even when modified.

Consequently, it is seen that softer binders, whether modified or unmodified, are most susceptible to a higher degree of aging in the high-temperature behavior spectrum, but that any aging in this region will coincide proportionally with aging throughout the useful temperature interval.

Intermediate-Stiffness Behavior

Intermediate-stiffness or temperature aging is analyzed through the progression of G-R at 15°C and 0.005 rad/s, $G_{\mathrm{int}}$ at the PG intermediate temperatures and 10 rad/s, and the viscoelastic crossover represented by $G_{\mathrm{vet}}$ , $T_{\mathrm{vet}}$ at 10 rad/s, ${\omega }_c$ at 15°C. Trends in evolution of behavior differed for parameters in the lower-intermediate and higher-intermediate stiffness ranges.

The evolution of G-R is sensitive to modification of non-aged binders because of location on the mechanical test data in Black Space. G-R grouped by $T_{\mathrm{int},\mathrm{cont}}$ improves clustering for the parameters in Black Space. Further investigation indicates that G-R = 180 kPa coincides with 35 < $T_{\mathrm{int},\mathrm{cont}}$ < 40°C. Fixed G-R boundaries may not be optimal for all environmental conditions—climate-specific thresholds should be investigated. It is suggested that environmental conditions may determine allowable aging. Fatigue tests of aged binders at intermediate temperatures may indicate the thresholds for aging to allow for suitable binder performance.

G-R indicates compelling interrelationships between a significant number of performance parameters. Converging interrelationships with G-R, VET, and ${\omega }_c$ are observed for increased binder age, while the correlation with $T_{\max ,\mathrm{cont}}$ diverges. This may be because G-R moves toward the upper left in Black Space—away from $T_{\max ,\mathrm{cont}}$ and toward VET. Interrelationships are sensitive to binder type and origin, except for ${\omega }_c$ that converges for all binders when G-R > ${10}^4$ Pa.

G-R ARs are more scattered for unmodified binders. Values range between 1 and 4 for these, and only between 1 and 2 for modified binders. Interrelationships indicate that unmodified binders are relatively more susceptible to aging in the lower-intermediate-stiffness domain measured by G-R than in low- and high-stiffness ranges.

A distinction is made between intermediate specification, or environmental, temperatures, that is, $T_{\mathrm{int}}$ and intermediate binder properties, measured at $T_{\mathrm{int},\mathrm{cont}}$ . Stiffness responses, $G_{\mathrm{int}}$ , at intermediate temperatures associated with PGs for the South African climate are considered, that is, 22°C, 28°C, and 34°C.

Intermediate stiffness correlates most with parameters located in proximity on the behavioral curve, that is, G-R, $T_{\mathrm{vet}}$ , and ${\omega }_c$ . The correlation with $G_{\mathrm{vet}}$ is less meaningful. Interrelationships and in-service evolution of $G_{\mathrm{int}}$ are primarily influenced by binder type and, to a lesser extent, by origin. Correlations are generally higher for an increased $T_{\mathrm{int}}$ . Binder fatigue tests at various artificial ages and intermediate temperatures are recommended to establish aging limits for suitable performance. It may also be prudent to investigate the relationship between $T_{\mathrm{int},\mathrm{cont}}$ - $T_{\mathrm{int}}$ and fatigue failure.

The ARs of $G_{\mathrm{int}}$ versus G-R notably separated aging tendencies in modified and unmodified binders. This distinction may be because $G_{\mathrm{int}}$ and G-R measure exactly the area in Black Space where modified binder behavior starts to deviate from unmodified binder behavior.

A decrease in $G_{\mathrm{vet}}$ and ${\omega }_c$ , and an increase in $T_{\mathrm{vet}}$ , is observed with increased artificial aging. The correlation between $G_{\mathrm{vet}}$ and $T_{\mathrm{vet}}$ improves for binder types and origins. Unmodified binders are located toward the upper portion of the data band, implying modified binders are less stiff when transitioning from viscous to elastic dominant behavior. In VET, S-E2 binders age most variably. The correlation between $T_{\mathrm{vet}}$ and ${\omega }_c$ is near perfect, since they are fundamentally linked by the time-temperature superposition principle and subsequently interchangeable.

Interrelationships at the viscoelastic crossover begin to indicate transition in dominating influencers of binder behavior and aging. Correlations with $G_{\mathrm{vet}}$ are predominantly influenced by binder source, as with $\Delta$ $T_c$ , and those of $T_{\mathrm{vet}}$ and ${\omega }_c$ by modification type.

ARs for $G_{\mathrm{vet}}$ are comparatively less variable than other parameters, and mostly between 0.9 and 1; that of ${\omega }_c$ is between −1 and 1. A comparison of ARs between $G_{\mathrm{vet}}$ , ${\omega }_c$ , and G-R reveals that unmodified binders are more susceptible to aging in the domain of VET than G-R.

Subsequently, it is seen that modified binders are more susceptible to aging in the higher stiffness domains, and unmodified binders relatively more in the lower-intermediate stiffness domains.

High-Stiffness Behavior

High-stiffness of low-temperature aging is analyzed through the evolution of $\Delta$ $T_c$ . Initial values of $\Delta$ $T_c$ appear dependent on binder type. Interrelationships with other parameters are comparatively poor if the source of the binder is not considered. $G_{\mathrm{vet}}$ correlates most significantly with $\Delta$ $T_c$ , regardless of the influence of modification or binder origin. This is reasonable, as both parameters are shape parameters in a similar range in Black Space. Should $G_{\mathrm{vet}}$ be considered as an alternative for $\Delta$ $T_c$ in specifications, it could reduce testing requirements significantly, as BBR data would not be necessary.

The comparison of ARs with G-R is substantial. G-R dominates aging for unmodified binders, while the $\Delta$ $T_c$ aging overshadows the progression of modified binder behavior.

In the high-stiffness domain, aging is highly influenced by the modification and crude source of the binder. It may be beneficial to investigate the molecular compositions of binders and the influence on low-temperature aging susceptibility.

Influence of Modification and Source on Interrelationships

For the binders and aging considered in the study, it is deduced that specific binder types are more susceptible to aging in particular distinct stiffness domains in Black Space:

• In the low-stiffness domain measured by $T_{\max ,\mathrm{cont}}$ , softer binders are more susceptible to aging than harder binders.

• In the lower-intermediate stiffness domain measured by G-R, unmodified binders are more susceptible to aging than modified binders.

• In the higher-intermediate stiffness domain measured by VET and ${\omega }_c$ , dominating factors vary depending on specific parameters and binder types.

• In the high-stiffness domain, susceptibility to aging is higher for modified binders and those from specific crude sources and refineries.

The development of $\Delta$ $T_c$ is more correlated to source than binder type. Aging in term of $T_{\mathrm{vet}}$ , ${\omega }_c$ , $G_{\mathrm{int}}$ , and G-R is correlated to binder type and source in varying degrees. The progression of $G_{\mathrm{vet}}$ is least influenced by binder type. The evolution of $T_{\max ,\mathrm{cont}}$ is dependent on modification type and initial hardness of the specific binder. As an example, S-E2 binders indicate relatively little aging in relation to $T_{\max ,\mathrm{cont}}$ , but significant variability in $\Delta$ $T_c$ . Aging potential is lower at high temperatures because of the initial hardness, and greater at low temperatures because of modification. The variability at lower temperatures may be attributed to differences in the origin of the various S-E2 binders.

Implications of Aging Ratios (ARs)

ARs normalize parameters and disregard initial conditions. One binder may indicate a greater AR than another, and still be relatively less aged in absolute terms. Interrelationships between ARs point out where behavior is most susceptible to aging. The onus is then on the analyzer to determine whether the aging mechanisms will lead to premature failure.

The concept of an AR appears contradictory to uniform parameter limits per specification grade. Uniform limits imply greater absolute allowances for changes in properties. An initially softer bitumen, with a lower G-R, may age significantly more than a harder bitumen before reaching the performance limits, although they were chosen for significantly different environments. Given the correlation between G-R and $T_{\mathrm{int},\mathrm{cont}}$ , fatigue tests at these temperatures could be correlated to G-R limits to investigate differences in thresholds for specification grades.

When considering ARs for specifications, primary failure mechanisms should be considered. Additionally, a combination of poorly correlating ARs may be considered. Complementary parameters will monitor distinct aging aspects and can provide reciprocal insights into binder aging. Thus, including ARs for G-R, ${\omega }_c$ , and $T_{\mathrm{vet}}$ may be redundant if the interrelationships are as strong as the binders considered for the study. Combining ARs, for example, for $G_{\mathrm{vet}}$ and G-R will provide information about aging of the high-to-intermediate and intermediate-to-low stiffness spectra, respectively.

Conclusion and Recommendations

It is seen in this study that there are both related and unrelated aging trends in the distinct temperature- or stiffness-related behavioral ranges in Black Space. Correlations in parameter development significantly improve when binders are grouped by similar features such as modification type or origin. In the high-stiffness range, binder source and modification type and degree greatly dictate aging trends; in the low-stiffness range, the initial hardness of the binder is indicative of aging potential; and in the intermediate domain, the evolution of properties may be influenced either modification or binder source.

Based on observations in the analysis, further investigations into specific areas are recommended:

• G-R and other parameter boundaries based on PG grade

• Whether binder fatigue tests at intermediate temperatures may assist in distinguishing parameter performance thresholds for the various PG grades

• The possible correlation of $T_{\mathrm{int},\mathrm{cont}}$ - $T_{\mathrm{int}}$ and fatigue failure

• The parallel between molecular composition and low-temperature aging

• Describing aging as a combination of G-R and ${\omega }_c$

• $G_{\mathrm{vet}}$ as a possible replacement parameter for $\Delta$ $T_c$ in specifications

• The calculation of ARs for parameters where both positive and negative values are present, such as $\Delta$ $T_c$ and $T_{\mathrm{vet}}$

Finally, it would be constructive to validate observations of interrelationships and ARs for binders that are from other crude and refinery sources, specifically in global regions where modification strategies differ from South Africa.