Correlations among chemical and rheological aging indices/indicators of asphalt binder at high temperatures

Abstract

This study examines the relationships between chemical and rheological–viscoelastic aging indicators of asphalt binders to clarify how oxidation-induced changes are reflected in high-temperature performance. Three neat binders with penetration grades of 40–50, 60–70, and 85–100 were subjected to short-term and then progressive long-term aging using the Rolling Thin Film Oven and multiple Pressure Aging Vessel (PAV) cycles. Aging was characterized using Fourier Transform Infrared (FTIR) spectroscopy, dynamic shear rheometry, and rotational viscosity measurements. While most previous studies have either examined relationships between a limited number of chemical and rheological indicators or focused exclusively on rheological aging indices, this study evaluates multiple key indicators, including an FTIR-based combined carbonyl–sulfoxide aging index, the Superpave rutting parameter (G^*/sinδ), zero shear viscosity (ZSV), apparent viscosity, and selected parameters of the 2S2P1D viscoelastic model. The results show that all indicators increase with aging, with the most pronounced changes occurring during the first PAV cycle. The FTIR-based combined carbonyl–sulfoxide index effectively reflects aging progression, and strong correlations were observed between chemical and rheological indices, including linear relationships between the FTIR aging index and both Log(ZSV) and the 2S2P1D α parameter, as well as a strong relationship between apparent viscosity and G^*/sinδ. In addition, the Superpave parameter G^*/sinδ at 60 °C exhibited a strong power-law correlation with the FTIR aging index and a robust exponential relationship with Log(ZSV) at the same temperature. These findings demonstrate that chemical oxidation is consistently reflected in rheological–viscoelastic behavior and reveal robust, trend-based relationships that can support flexible interpretation of binder aging and provide a basis for simplified, data-driven aging assessment.

Introduction

Asphalt binders consist of a variety of organic compounds that are prone to oxidation when exposed to air. Over time, oxidative processes increase binder rigidity and reduce flexibility, altering mixture performance and affecting pavement durability¹. In addition to chemical aging, the physical response of asphalt binder is strongly influenced by temperature, time, and applied loading throughout its service life². The most significant factors in asphalt binder aging are oxidation due to air and volatilization due to heat³. The durability of asphalt pavement is strongly influenced by the extent of oxidative aging, which leads to both structural and chemical modifications in binder molecules. This process develops as oxygen gradually diffuses into the binder during exposure to the atmosphere⁴.

Aging of asphalt is generally categorized into short-term and long-term processes, each occurring under different conditions and influencing the binder in particular ways. Short-term aging takes place during production, mixing, and construction stages, where the binder is exposed to elevated temperatures, mechanical mixing, and air, resulting in rapid oxidation and volatilization of lighter components⁵. This process increases the asphalt binder’s stiffness and viscosity even before the pavement enters service. Therefore, hot mix asphalt (HMA) pavements begin their service life having already undergone short-term aging. The rolling thin-film oven test (RTFOT), developed as an enhancement of the thin-film oven test (TFOT), was incorporated into the Superpave protocol to simulate the short-term aging of asphalt binders⁶. In this procedure, binder samples are conditioned at 163 °C for 85 min with an air supply of 4000 mL/min. In contrast, long-term aging develops gradually during pavement service life as a result of prolonged exposure to oxygen, ultraviolet radiation, and moisture, combined with temperature fluctuations. This aging process further increases stiffness, reduces ductility, and can ultimately lead to fatigue and thermal cracking. The Pressure Aging Vessel (PAV) test is widely used to simulate long-term aging in the laboratory by subjecting RTFO-aged binders to elevated temperature and pressurized air, replicating years of in-service conditions.

The extent of binder aging varies with geographic and environmental conditions, and the standard 20-h PAV conditioning protocol may not adequately represent long-term aging effects across different climates and locations⁷. To address this limitation, several studies have investigated extended PAV-aging durations of 40, 60, and 80 h as potential alternatives to the 20-hour protocol specified in the Superpave system^8,48.

Due to the accelerating global warming, asphalt pavement has become more vulnerable to aging-related degradation; therefore, understanding asphalt binder aging has become significantly important⁴. Another major motivation for studying asphalt binder aging is the growing use of reclaimed asphalt pavement (RAP). Rising costs of pavement materials and growing environmental concerns have led to a significant increase in RAP utilization in HMA for both construction and maintenance applications. During asphalt pavement rehabilitation and maintenance, a considerable amount of RAP is inevitably produced. Utilizing RAP conserves natural aggregates, mitigates environmental pollution, and reduces maintenance expenditures, ultimately offering significant ecological and economic advantages⁹. However, progressive aging of asphalt binders during the pavement service life results in increased stiffness of RAP materials due to binder hardening. Consequently, it is well-established that incorporating RAP into HMA mixtures alters their characteristics. The extent of this alteration primarily depends on the RAP properties and the degree of blending between RAP and virgin binders within the HMA^10,11. Therefore, effective use of RAP materials in HMA mixtures requires a clear understanding of how the mechanical properties of the RAP binder have evolved due to aging.

Aging increases the stiffness of asphalt binders, which enhances the resistance of asphalt pavements to permanent deformation at high service temperatures. Consequently, the high-temperature performance of asphalt pavements tends to improve as aging progresses during service life¹². However, it is well established that such aging simultaneously adversely affects pavement performance at intermediate and low temperatures, reflected in increased susceptibility to fatigue and thermal cracking. Despite these deleterious effects, assessment of asphalt aging at high temperatures remains essential for material selection, mixture design, and performance prediction.

Aging indices are quantitative parameters used to evaluate the extent of physical or chemical changes in asphalt binders caused by short-term and long-term aging processes. These indices are typically derived by comparing specific binder properties, such as viscosity, complex modulus, or chemical functional group absorbance, before and after laboratory-simulated or field aging¹³. By normalizing the change in a given property relative to its original value, aging indices provide a standardized approach to assessing binder hardening, oxidation, and rheological shifts. Commonly used aging indices include viscosity-based aging indices, dynamic shear rheometer (DSR)-based aging indices, and Fourier-transform infrared (FTIR) spectroscopy indices, each capturing different aspects of binder aging behavior^14,15.

The selection of appropriate aging indices is critical for accurately characterizing changes in binder performance and predicting pavement durability. Rheological indices, such as the ratio of post-aging to pre-aging G^*/sin δ, are closely associated with high-temperature performance and rutting resistance, whereas chemical indices, such as carbonyl or sulfoxide indices obtained from FTIR, directly reflect oxidation levels^16,49. These indices not only facilitate comparisons among binder types and modification strategies but also provide valuable inputs for performance prediction models. Integrating multiple aging indices allows a more comprehensive evaluation of binder aging mechanisms, helping to better link laboratory findings with in-service pavement performance.

Investigating the correlations among different asphalt binder aging indices is essential for understanding the interdependence of rheological and chemical changes during aging. Establishing reliable correlations among these indices is particularly important for predictive modeling and performance-based specifications¹⁷. When strong and consistent relationships are observed, one index may serve as a surrogate for another, thereby simplifying testing protocols and reducing experimental effort. Moreover, understanding these correlations supports the development of mechanistic models that link chemical composition changes to rheological behavior and, ultimately, to pavement performance. This integrative approach enables more efficient evaluation of binder durability, facilitates validation of alternative aging simulation methods, and strengthens the connection between laboratory measurements and field performance.

Most previous investigations have either examined relationships between a limited number of chemical and rheological indicators or focused exclusively on rheological aging indices^18,19,20,21. Moreover, the literature reports inconsistencies regarding the reliability of specific FTIR-based chemical indices. The carbonyl index is generally regarded as a robust indicator of long-term oxidative aging, whereas the sulfoxide index often exhibits more variable and binder-dependent behavior, limiting its universal applicability as a predictive parameter^2,22,23,24. Meanwhile, zero shear viscosity (ZSV) has been widely recognized as a meaningful high-temperature rheological indicator associated with oxidative aging and rutting resistance^25,26. Although ZSV has been linked to performance-based parameters or conventional rheological indices (e.g., G^*/sinδ), its direct statistical relationship with FTIR-derived oxidation markers remains largely unexplored. Consequently, the coupled evolution of chemical oxidation indices and high-temperature rheological parameters across different aging levels and binder grades is not yet fully understood.

This study aims to investigate the correlations between chemical and rheological aging indicators to examine how aging-induced changes in binder chemistry are reflected in rheological properties. Particular attention is given to the combined carbonyl–sulfoxide aging index as a representative measure of aging severity. Furthermore, by identifying consistent relationship trends among multiple aging indicators, this study seeks to demonstrate the potential existence of meaningful correlations that may be useful for assessing and predicting binder aging behavior.

Materials and methods

Materials

This study examined three conventional asphalt binders with penetration grades of 40–50, 60–70, and 85–100, all obtained from Pasargad Oil Company. The fundamental properties of these binders are summarized in Table 1.

Table 1 Properties of the asphalt binders used in this study.

Sample preparation

The binders were subjected to four conditioning methods: short-term aging using the Rolling Thin Film Oven (RTFO) according to AASHTO T 240, and long-term aging through one, two, and three cycles of Pressure Aging Vessel (PAV) aging following AASHTO R 28. Following RTFO-aging, the samples were placed in a PAV chamber and aged at 100 °C under a pressure of 2070 kPa for 20 h. The 2PAV samples underwent an additional 20-h aging cycle, resulting in a total PAV-aging duration of 40 h, while the 3PAV samples were aged for two additional 20-h cycles, bringing their total PAV-aging duration to 60 h.

Test methods

Frequency sweep test using dynamic shear rheometer

The frequency sweep tests were conducted at temperatures ranging from 46 °C to 82 °C, with 6 °C intervals, to obtain the dynamic shear modulus (G^*), phase angle (δ), and dynamic viscosity (η^*). A 25 mm parallel plates geometry with a 1 mm gap was employed. A strain amplitude of 0.01% was applied to ensure that all measurements remained within the linear viscoelastic (LVE) region, over an angular frequency range of 0.628–628 rad/s (0.1–100 Hz). At the elevated temperatures investigated, asphalt binders exhibit relatively wide LVE limits, and very small strains are typically used to avoid non-linear effects, even for highly aged materials.

Fourier transform infrared (FTIR) spectroscopy

Asphalt binder is composed of millions of hydrocarbon molecules, each exhibiting distinct temperature-dependent viscoelastic behavior. This molecular complexity makes it impractical to characterize every individual component^23,27. Consequently, chemical analyses that provide an overview of the binder’s composition or its molecular groups are typically employed.

Over the past few decades, chemical analysis of asphalt materials has become increasingly common. Among the available techniques, FTIR spectroscopy has become one of the most widely used methods for monitoring chemical changes, particularly oxidation, in asphalt materials over their lifespan. Molecular bonds within the binder absorb infrared (IR) radiation, exciting them to higher energy states; this absorption behavior provides a fundamental mechanism for characterizing the binder’s chemical composition.

FTIR operates in two functional modes: projection and reflection. Researchers increasingly favor the reflection mode because it eliminates the need for the IR radiation to pass through the sample. Among reflection-based techniques, Attenuated Total Reflection (ATR) is the most commonly utilized method in Fourier spectroscopy. In this study, FTIR measurements were performed using a Bruker Tensor 27 spectrometer equipped with an ATR containing a diamond crystal and a DTGS detector. Spectra were recorded over a wavenumber range of 4000–400 cm⁻¹, with a spectral resolution of 4 cm⁻¹ and 24 scans. Prior to each measurement, a background spectrum of the clean ATR crystal was recorded and automatically subtracted from the sample spectrum. Asphalt binder samples were placed on the ATR crystal within approximately 1 min after background collection to minimize potential baseline drift.

Previous studies have shown that the sulfoxide (S═O) and carbonyl (C═O) functional groups serve as reliable indicators for evaluating asphalt binder aging. These groups form as a result of long-term aging, and their concentrations generally increase as the aging process progresses. Therefore, monitoring the evolution of these functional groups, enables effective tracking and quantification of asphalt binder aging^27,28. Accordingly, this study focused on variations in the carbonyl and sulfoxide regions of the FTIR spectra. The carbonyl region is typically associated with the absorption band near 1680 cm⁻¹, while the sulfoxide region corresponds to the band near 1030 cm^–1. The aliphatic group, represented by symmetric and asymmetric bending vibrations around 1460 and 1376 cm^–1, respectively, were used as a reference band because they are considered stable and largely unaffected by aging. The aging indicators for each functional group were calculated from the enclosed area of the spectroscopic chart within the corresponding wavelength region^27,29: 1650–1750 cm^–1 for carbonyls (AI_CO), 980–1030 cm^–1 for sulfoxides (AI_SO), and 1350–1525 cm^–1 for the reference aliphatic band (AI_CH3), as illustrated in Fig. 1.

Fig. 1

Schematic illustration of chemical indicators associated with the FTIR aging index.

Previous studies have shown that sulfoxide formation occurs rapidly during the initial stages of aging and stabilizes at later stages, whereas carbonyl formation continues progressively as aging advances^27,30. To capture these complementary contributions, the carbonyl and sulfoxide functional groups were combined into a single FTIR-based aging index, providing a comprehensive measure of oxidative aging. The FTIR aging index (AI_FTIR) used in this study is defined in Eq. (1)²⁷:

$$\:{\text{AI}}_{\text{FTIR}}\text{}\text{=}\text{}\frac{{\text{AI}}_{\text{CO}}\text{}\text{+}{\text{}\text{AI}}_{\text{SO}}}{{\text{AI}}_{{\text{CH}}_{\text{3}}}}$$

(1)

Rotational viscometer (RV)

RV tests were conducted to determine the apparent viscosity of the binders within the elevated temperatures relevant to asphalt production and construction. In this method, the torque required to maintain a constant rotational speed of a spindle immersed in the binder at a controlled temperature is measured. The measured torque is then converted into shear stress, while the rotational speed is converted into shear rate. The ratio of shear stress to shear rate yields the apparent viscosity, which is displayed by the instrument. In this study, RV measurements were performed at 135 °C using a spindle rotational speed of 20 rpm.

Theoretical basis

Master curve construction and viscoelastic modeling

The construction of a master curve involves conducting DSR tests at multiple temperatures and frequencies, followed by horizontal shifting of the data to a reference temperature (along the frequency axis) using an appropriate shift factor model. The shifted frequencies, referred to as “reduced frequencies”, enable the characterization of binder behavior over a frequency domain substantially broader than can be captured by direct testing alone. In this study, the William–Landel–Ferry (WLF) shift factor function was employed to calculate the reduced frequencies and align the data, as presented in Eqs. (2a) and (2b):

$$\:\text{Log(}{\text{a}}_{\text{T}}\text{)\:=\:}\frac{-{\text{C}}_{\text{1}}\text{\:(}\text{T}-{\text{T}}_{\text{ref.}}\text{)\:}}{{\text{C}}_{\text{2}}\text{\:+\:(}\text{T}-{\text{T}}_{\text{ref.}}\text{)}}\:\:\:\:\:\:\:$$

(2a)

$$\:\:{\text{f}}_{\text{r}}\text{}\text{=}{\text{}\text{a}}_{\text{T}}\text{} \times \text{}\text{f}$$

(2b)

$\:{\text{a}}_{\text{T}}$ is the shift factor; C₁, C₂ is the constant parameters; T is the temperature; T_ref. is the reference temperature; f_r is the reduced or shifted frequency; and f is the frequency.

The resulting master curve represents the complex modulus as a function of reduced frequency. When plotted on a log–log scale of dynamic modulus versus reduced frequency, the master curve typically exhibits a sigmoidal shape. Accordingly, several mathematical models have been proposed to describe this sigmoidal behavior in dynamic modulus master curves. Among the most commonly used are the Christensen–Anderson–Marasteanu (CAM) model and 2S2P1D models. In this study, both the CAM and 2S2P1D models were employed for analysis and application.

CAM model

The CAM model is a widely used mathematical representation for constructing master curves of the complex modulus (|G^*|) of asphalt binders and asphalt mixtures over a broad range of loading frequencies and temperatures³¹. This model captures the viscoelastic behavior and rheological properties of asphalt materials by fitting a sigmoidal function to frequency sweep data, which are typically time–temperature superposed to generate a continuous curve. The CAM model is expressed in Eq. (3):

$$\:\left|{\text{G}}^{\text{*}}\left({\omega}_{\text{r}}\right)\right|\text{=}\frac{{\text{G}}_{\text{g}}}{{\left(\text{1+}{\left(\frac{{\omega}_{\text{c}}}{{\omega}_{\text{r}}}\right)}^{\text{k}}\right)}^{\frac{\text{m}}{\text{k}}}}$$

(3)

$\:\left|{\text{G}}^{\text{*}}\left({\omega}_{\text{r}}\right)\right|$ is the the complex modulus at angular frequency of $\:{\omega}_{\text{r}}$; $\:{\text{G}}_{\text{g}}$ is the the glassy modulus (assumed ~ 1 GPa for neat asphalt binders); $\:{\omega}_{\text{r}}$ is the reduced frequency; $\:{\omega}_{\text{c}}$ is the crossover frequency; and k, m is the model’s fitting parameters (to control the shape and curvature of the master curve).

The CAM model offers improved fitting accuracy compared to simpler sigmoidal models by accommodating the non-linear transition between the glassy, viscoelastic, and flow regions of the material. It is particularly useful for evaluating the long-term performance characteristics of asphalt binders and for comparing the effects of aging, modification, or binder source on rheological behavior.

2S2P1D model

The 2S2P1D model, consisting of two springs, two parabolic elements, and one dashpot, is an advanced viscoelastic constitutive model used to characterize linear viscoelastic behavior of asphalt materials over a broad spectrum of frequencies and temperatures. Developed to better capture the complex time-dependent behavior of asphalt, the 2S2P1D model combines physical interpretability with high accuracy in fitting experimental data.

The model is composed of the following rheological elements: two linear elastic springs representing instantaneous and long-term elastic responses, two fractional derivative (parabolic or “Springpot”) elements to capture the power-law nature of the material’s relaxation behavior, and one dashpot to model viscous flow. The mathematical formulation for the complex modulus is presented in Eqs. (4a) and (4b)³²:

$${G^*}\left( \omega \right) = {G_0} + \frac{{{G_g} - {G_0}}}{{1 + \alpha {{\left( {i\omega \tau } \right)}^{ - k}} + {{\left( {i\omega \tau } \right)}^{ - h}} + {{\left( {i\omega \beta \tau } \right)}^{ - 1}}}}$$

(4a)

$$\:{\beta}\text{}\text{=}\text{}\frac{\eta}{\text{(}{\text{G}}_{\text{g}}\text{}\text{\--}\text{}{\text{G}}_{\text{0}}\text{)}{\tau}}$$

(4b)

$\:{\text{G}}^{\text{*}}\left(\omega\right)$ is the dynamic modulus at frequency $\:\omega$; $\:{\text{G}}_{\text{0}}$ is the static or equilibrium modulus (when ω → 0); $\:{\text{G}}_{\text{g}}$ is the glassy modulus (when ω → ∞); k, h is the exponents with 0 < k < h < 1; $\:{\tau}$ is the characteristic time; α, β is the constants; and $\:\eta$ is the Newtonian viscosity.

This model effectively describes the transition from the glassy to the rubbery and viscous domains of asphalt behavior, making it highly suitable for constructing master curves and analyzing the effects of aging, temperature, or additive modification. Compared to empirical models, the 2S2P1D model offers superior fitting quality while maintaining a physical basis, enabling its use not only for material characterization but also for predictive modeling in pavement design and simulation. Figure 2 illustrates the mechanical 2S2P1D model.

Fig. 2

The 2S2P1D mechanical model.

While both the 2S2P1D and CAM models are used to describe the rheological behavior of asphalt materials and construct master curves, they differ in their theoretical foundations, complexity, and application scope. The CAM model is an empirical sigmoidal equation designed primarily for fitting the complex modulus data across frequencies after time–temperature superposition. Its simplicity and ease of implementation make it suitable for routine binder characterization. However, it lacks a direct physical interpretation of its parameters, which may limit its predictive capability in mechanistic analyses.

In contrast, the 2S2P1D model is a semi-empirical viscoelastic model grounded in rheological theory, incorporating both mechanical analogs and fractional calculus to represent asphalt behavior over a broad frequency and temperature range. Its parameters, such as equilibrium modulus, characteristic time, and fractional orders, have physical significance, providing greater insight into material behavior, particularly for aged or modified asphalt materials. Moreover, the 2S2P1D model generally provides superior fitting to experimental data and more accurately captures transitions between elastic, viscoelastic, and viscous regimes. Consequently, it is particularly advantageous for advanced modeling tasks, including finite element simulations and constitutive modeling in pavement design.

In this study, the CAM model was selected over the 2S2P1D model for estimating the WLF shift factor parameters, C₁ and C₂, due to its lower number of fitting parameters. The fitted CAM model parameters for the frequency sweep data at the reference temperature of 60 °C are presented in Table A.1.

Complex viscosity modeling and zero shear viscosity (ZSV)

At elevated temperatures, unmodified asphalt binders exhibit Newtonian fluid behavior. Furthermore, at very low shear rates, the viscosity of asphalt binders becomes independent of shear rate or frequency. This limiting viscosity is referred to as the zero shear rate viscosity or Newtonian viscosity, denoted by η_o.

Several approaches have been proposed for determining the ZSV of asphalt binders. In this study, ZSV was estimated from frequency sweep measurements performed with a DSR over a loading frequency range of 0.628–628 rad/s (0.1–100 Hz). Direct measurement of viscosity at near-zero shear rates is generally impractical due to prolonged testing times and equipment limitations. Therefore, ZSV is typically inferred through rheological modeling, which involves extrapolating the binder’s viscosity behavior to very low shear rates where it exhibits Newtonian characteristics. Among the various models used to characterize this behavior, the Cross model is one of the most widely applied. It is a four-parameter empirical model that effectively captures the shear-thinning behavior of asphalt binders^33,34. The mathematical formulation of the Cross model is presented in Eq. (5):

$$\:\frac{{\eta}^{\text{*}}\text{\--}\text{}{\eta}_{\infty}^{\text{*}}}{{\eta}_{\text{o}}^{\text{*}}\text{\--}\text{}{\eta}_{\infty}^{\text{*}}}\text{=}\frac{\text{1}}{\text{1}\text{+}{\text{(}{k \omega}\text{)}}^{\text{m}}}$$

(5)

$\:{\eta}^{\text{*}}$ is the complex viscosity;$\:{\eta}_{\text{o}}^{\text{*}}$ is the ZSV; $\:{\eta}_{\infty}^{\text{*}}$ is the limiting viscosity in the second Newtonian region; $\:\omega$ is the angular frequency, rad/s; and k, m is the model’s constant parameters.

It should be noted that, since the behavior of neat asphalt binders at elevated temperatures (e.g., 60 °C) is generally considered Newtonian^35,36 (Rakha & Krishnan 2011), and given the small strain amplitude (0.01%) applied in the frequency sweep tests, the samples are expected to behave within the linear viscoelastic range. Accordingly, the Cox–Merz rule is assumed to be valid, and the complex viscosity is taken as equivalent to the steady-shear viscosity. The validity of the Cox–Merz rule for neat asphalt binders at high temperatures has been reported in previous studies³⁷. The Cox–Merz rule is an empirical principle stating that the magnitude of the complex viscosity, |η^*|, measured at an angular frequency ω, is equal to the steady-state shear viscosity, η, measured at a numerically equivalent shear rate, $\:\dot{\gamma}$. At very low frequencies and shear rates, this relationship can be expressed as $\:{|{\eta}^{\text{*}}\text{(}\omega\text{)}\text{|}}_{\omega \to \text{0}}\text{=}{|\eta\text{}\text{(}\dot{\gamma}\text{)}\text{|}}_{\dot{\gamma}\to\text{0}}$.

Several studies have demonstrated that the rutting resistance of asphalt pavements is strongly related to the ZSV of the binder. In particular, ZSV determined at 60 °C, representing the highest pavement temperature typically encountered, has been found to correlate closely with asphalt mixture wheel tracking tests results^38,39.

The logarithm of ZSV typically exhibits a linear relationship with the inverse of absolute temperature, 1/T, where T is expressed in Kelvin. This behavior is consistent with the Arrhenius equation⁴⁰, as represented in Eq. (6):

$$\:\text{ln}\text{(}{\eta}_{\text{o}}\text{)\:=\:}\text{ln}\text{(}\text{A}\text{)}\text{\:+\:}\frac{\text{1}}{\text{T}}\text{(}\frac{{\text{E}}_{\text{a}}}{\text{R}}\text{)}$$

(6)

$\:{\eta}_{\text{o}}$ is the ZSV; A is the pre-exponential factor; T is the absolute temperature, Kelvin; $\:{\text{E}}_{\text{a}}$ is the activation energy for viscous flow; and R is the universal gas constant.

In this study, complex viscosity data were available at 58 °C and 64 °C, the two temperatures closest to the target temperature of 60 °C. Arrhenius interpolation was applied over this narrow temperature interval to estimate ZSV at 60 °C. Using Eq. (6), ZSV for each sample was estimated by linear interpolation of the logarithmic ZSV values with respect to the inverse absolute temperature between these two reference points, as detailed in Eq. (7):

$$\:\text{log}\text{(}{\text{ZSV}}_{{\text{T}}_{\text{d}}}\text{)}\text{=}\text{}\frac{\frac{\text{1}}{{\text{T}}_{\text{d}}}\text{\--}\frac{\text{1}}{{\text{T}}_{\text{2}}}}{\frac{\text{1}}{{\text{T}}_{\text{1}}}\text{\--}\frac{\text{1}}{{\text{T}}_{\text{2}}}}\text{(}\text{log}\text{(}{\text{ZSV}}_{{\text{T}}_{\text{1}}}\text{)}\text{\--}\text{log}\text{(}{\text{ZSV}}_{{\text{T}}_{\text{2}}}\text{)}\text{)}\text{}\text{+}\text{log}\text{(}{\text{ZSV}}_{{\text{T}}_{\text{2}}}\text{)}$$

(7)

$\:{\text{ZSV}}_{{\text{T}}_{\text{d}}}$ is the ZSV at target temperature of $\:{\text{T}}_{\text{d}}$; $\:{\text{T}}_{\text{d}}$ is the target temperature (60 °C), Kelvin; $\:{\text{T}}_{\text{1}}$, $\:{\text{T}}_{\text{2}}$ is the two temperatures which their ZSV is available (58 °C and 64 °C), Kelvin; and $\:{\text{ZSV}}_{{\text{T}}_{\text{1}}}$, $\:{\text{ZSV}}_{{\text{T}}_{\text{2}}}$ is the ZSV at temperatures of T₁ and T₂.

Results and discussion

The superpave high-temperature performance parameter

After subjecting the asphalt binder samples to the aging processes, the high-temperature performance parameter, G^*/sinδ, was evaluated using the original Superpave test. The values of G^* and δ were determined by applying a sinusoidal load at a frequency of 10 rad/s. The G^*/sinδ values for all tested samples are summarized in Appendix Table A.2.

To facilitate comparison of the results, Fig. 3a,c present the G^*/sinδ values of the aged samples for the 40–50, 60–70, and 85–100 base binders, respectively.

Fig. 3

G^*/sinδ of RTFO and PAV aged samples for (a) 40–50; (b) 60–70; and (c) 85–100 base binders.

In the original Superpave performance grading system, the rheological parameter of G^*/sinδ serves as a quantitative indicator of an asphalt binder’s resistance to permanent deformation, specifically rutting, at service temperatures. The high-temperature performance criteria require G^*/sinδ to exceed 1.0 kPa for the unaged (original) binder and 2.2 kPa for the RTFO-aged residue. As expected, Fig. 3a,c show that aging increases the rutting parameter (G^*/sinδ), indicating enhanced resistance of the samples to permanent deformation.

Figure 4 illustrates the normalized G^*/sinδ of each PAV-aged sample relative to its corresponding RTFO-aged counterpart at 60 °C (the same temperature used for ZSV evaluation), thereby enabling comparison of aging progression. Figure 5 presents the ratio of G^*/sinδ for each PAV-aged sample to that of the immediately preceding aging stage at 60 °C. The G^*/sinδ values at 60 °C were interpolated from the measured values at 58 °C and 64 °C in accordance with ASTM D7643⁴¹.

Fig. 4

Normalized G^*/sinδ of PAV-aged samples relative to their corresponding RTFO-aged samples at 60 °C.

Fig. 5

Ratio of the G^*/sinδ of each PAV-aged sample to the G^*/sinδ of its preceding aging stage’s sample at 60 °C.

As shown in Fig. 4, G^*/sinδ increases substantially with successive PAV cycles, reflecting progressive stiffening and enhancement of the asphalt binder’s elastic response due to oxidative aging. Figure 5 further indicates that the largest relative increase occurs during the first PAV stage, after which the rate of increase diminishes sharply. These observations suggest that the binder’s resistance to permanent deformation is established primarily during the early aging phase, which also represents a critical period for the onset of potential age-related embrittlement, highlighting its key role in determining long-term high-temperature performance.

Zero shear viscosity (ZSV)

The Cross model (Eq. (5)) was fitted to the experimental data obtained at 58 °C and 64 °C using Excel’s Solver add-in based on a least-squares minimization approach. To reduce potential user bias, identical solving methods (GRG Nonlinear) and convergence settings (Convergence = 0.0001) were applied consistently across all samples. The fitted model parameters (k and m), the corresponding estimated ZSV values, and goodness-of-fit indicators, including R² and the standard error ratio (S_e/S_y), are presented in Table A.3. The Log(ZSV) at 60 °C was then determined by interpolation using Eq. (7), and the resulting ZSV values are illustrated in Fig. 6. As expected, the 40–50 base asphalt samples generally exhibit higher ZSV values, whereas the 85–100 base samples show lower values.

Fig. 6

ZSV of the samples at 60 °C.

Figure 7 shows the normalized ZSV of each PAV-aged sample relative to the corresponding RTFO-aged sample within each base asphalt group, while Fig. 8 presents the ratio of the ZSV of each PAV-aged sample to that of the preceding aging stage, providing insight into the progressive effects of aging.

Fig. 7

Normalized ZSV of the PAV-aged samples based on the ZSV of the Corresponding RTFO sample at 60 °C.

Fig. 8

Ratio of the ZSV of each PAV-aged sample to the ZSV of its preceding aging stage at 60 °C.

Consistent with the trend observed for G^*/sinδ, Fig. 7 demonstrates that after the production and laydown phase, represented by RTFO-aging, the asphalt binder undergoes a substantial increase in viscosity due to progressive long-term aging simulated by successive PAV cycles. Figure 8 further indicates that the largest relative change occurs during the first PAV cycle, highlighting its critical role in long-term pavement performance. In other words, the binder experiences the most significant structural and chemical changes during the initial oxidative aging phase, whereas subsequent stages have a less pronounced impact on ZSV.

Dynamic modulus modeling at high temperatures

The dynamic modulus master curve provides a fundamental representation of the viscoelastic behavior of asphalt binders over a broad range of temperatures and loading frequencies. It offers a comprehensive understanding of the material’s stiffness and time–temperature dependency, which is essential for predicting performance under various climatic and traffic conditions. Constructed using time–temperature superposition (TTS) principles, the master curve aligns modulus data obtained at different temperatures into a single smooth curve referenced to a selected reference temperature. In this study, 60 °C was chosen as the reference temperature for constructing the master curves, since the primary objective is to investigate and compare the high-temperature performance of the asphalt binder samples (Zhao et al., 2016; Zhang et al., 2017). Figures 9 and 10, and 11 present the dynamic modulus master curves at 60 °C for the 40–50, 60–70, and 85–100 base samples, respectively.

Fig. 9

Dynamic modulus master curves of the 40–50 samples at a reference temperature of 60 °C.

Fig. 10

Dynamic modulus master curves of the 60–70 samples at a reference temperature of 60 °C.

Fig. 11

Dynamic modulus master curves of the 85–100 samples at a reference temperature of 60 °C.

Table 2 presents the fitted parameters of the 2S2P1D model using Excel’s Solver add-in, indicating the model’s strong capability to capture the experimental behavior. The parameter β is associated with the Newtonian viscosity of the material (see Eq. (4b) and therefore strongly affects the material’s response at elevated temperatures and low frequencies. As aging leads to an increase in viscosity, the value of β is expected to increase correspondingly, as can be observed in Table 2. Similarly, the parameter α increases with progressive aging and hardening of the material. This trend is consistent with previous studies, which indicate that α may serve as an indicator of aging, increasing with the degree of oxidative aging for aged unmodified asphalt binders^32,42. Additionally, Table 2 shows that the τ values also increase with advancing aging levels across all three sample groups, suggesting that τ may likewise be considered a potential aging indicator of asphalt binder.

Table 2 2S2P1D model parameters fitted to the frequency sweep data at a reference temperature of 60 °C.

Within the framework of the 2S2P1D model, the parameter α controls the slope of the dynamic modulus master curve at high frequencies and low temperatures, as well as the height of the pinnacle point of the Cole–Cole diagram, which represents the relationship between G′ and G″. The parameter β governs the slope of the dynamic modulus master curve at low frequencies and high temperatures; higher values of β corresponds to higher values of viscosity (η) and G^*. A graphical illustration of the effects of β and α is provided in Fig. 12a,b.

Fig. 12

Graphical illustration of the effect of (a) β on dynamic modulus master curve, and (b) α on Cole–Cole diagram.

FTIR aging index

FTIR test results were obtained over a wavenumber range of 4000 to 400 cm⁻¹. To calculate the FTIR aging index, the functional group indicators corresponding to carbonyl, aliphatic, and sulfoxide groups were extracted from the FTIR spectra of each sample, as described in “Fourier transform infrared (FTIR) spectroscopy”. These extracted values were subsequently substituted into Eq. (1) to determine the FTIR aging index. The resulting indices for all samples are illustrated in Fig. 13.

Fig. 13

Aging indices using FTIR test results.

As shown in Fig. 13, the FTIR aging index exhibits its most significant increase following the first PAV step. With subsequent PAV-aging steps, the rate of increase in the FTIR aging index decreases. This trend reflects the behavior of the individual functional group indices: the sulfoxide indicator (AI_SO) increases markedly during the initial aging period and then stabilizes, whereas the carbonyl indicator (AI_CO) continues to increase with each additional aging step. Similar observations have been reported in previous studies³⁰, indicating that carbonyl formation continues during extended aging, while sulfoxide formation tends to plateau.

High temperature apparent viscosity

Apparent viscosity measured at high temperatures, such as 135 °C, is widely recognized as a practical and reliable indicator of asphalt binder aging^43,44,45. The temperature of 135 °C is commonly assumed to represent the mixing and construction temperature of HMA and is used in viscosity-based criteria for binder grading. This parameter reflects the binder’s resistance to flow under elevated temperature conditions during production, mixing, and compaction. As asphalt binders undergo oxidative aging, volatile loss, and polymerization, their viscosity increases due to changes in molecular structure and chemical composition. Measuring apparent viscosity at high temperatures therefore provides insight into the binder’s hardening rate, making it particularly relevant for evaluating short-term aging effects simulated by the RTFO testing and for monitoring long-term changes following PAV conditioning.

To evaluate the high-temperature viscosity of the samples, RV tests were conducted at 135 °C. Figure 14 presents the measured viscosity values for all samples.

Fig. 14

Measured apparent (rotational) viscosity of the samples at 135 °C.

Correlations between the aging indices at high temperatures

In high-temperature performance evaluation of asphalt binders, establishing robust relationships among chemical, rheological, and viscoelastic aging indices is essential for both mechanistic understanding and predictive modeling. This section focuses is on four key families of aging metrics measured at elevated service temperatures: the FTIR aging (carbonyl–sulfoxide) index, the Superpave rutting parameter G^*/sin δ, the high-temperature viscosity and ZSV, and the master curve parameters (τ, α, and β) obtained from the 2S2P1D viscoelastic model. Each of these indicators captures a distinct aspect of binder evolution under oxidative and thermal stress: FTIR quantifies molecular oxidation, G^*/sin δ reflects high-temperature stiffness and rutting resistance, ZSV and apparent viscosity characterize shear flow behavior, and the 2S2P1D parameters describe shifts in relaxation dynamics and modulus plateaus.

All indicators investigated in this study, including the FTIR aging index, ZSV, high-temperature apparent viscosity, G^*/sinδ, and the 2S2P1D model parameters (β, α, τ), exhibited increasing trends with progressive aging of the asphalt binder samples, consistent with expected behavior. By analyzing the correlations among these measures, this study aims to: (a) identify which aging indicators most closely track chemical oxidation, (b) evaluate the consistency between rheological and chemical metrics throughout the aging process, and (c) clarify the trend-based relationship between changes in the viscoelastic spectrum and traditional stiffness and viscosity indicators. Understanding these interdependencies not only supports the use of simpler or more accessible test methods as proxies for more complex spectral analyses but also provides valuable insight for developing improved accelerated aging protocols and predictive lifetime models for assessing asphalt binder performance under high-temperature conditions. Such predictive capabilities are essential for optimizing pavement design life and maintenance scheduling strategies.

As discussed previously in “Complex viscosity modeling and zero shear viscosity (ZSV)”, the ZSV of the samples was determined at 60 °C. To enable consistent comparison and correlation, G^*/sinδ and the 2S2P1D viscoelastic model parameters were also obtained at 60 °C. In contrast, the apparent viscosity was measured at 135 °C, corresponding to the conventional mixing and construction temperature of HMA and consistent with standard viscosity-based binder grading criteria. Although measured at a different temperature, the apparent viscosity is governed by the same aging-induced molecular changes that influence rheological and viscoelastic behavior over a wide temperature range. Accordingly, the trend-based correlations presented herein capture the overall effect of aging and binder stiffening rather than temperature-specific equivalence.

It should be noted that the correlations developed in this study are based on three specific neat binder grades (40–50, 60–70, and 85–100) and therefore should not be interpreted as universally applicable to all asphalt binders, particularly those derived from datasets combining multiple binder grades. The proposed relationships are intended primarily to demonstrate the feasibility of establishing simple correlations among high-temperature aging indicators and to highlight the general trends governing their interrelationships. The absolute form and coefficients of these correlations are expected to depend on binder source, composition, and chemical characteristics.

Linear correlations between the high-temperature aging indices/indicators

The FTIR aging index, typically focusing on the growth of carbonyl and sulfoxide groups, serves a chemical measure of oxidation. As an asphalt binder ages, its chemical composition changes, leading to corresponding alterations in its physical and rheological properties. Increased oxidative aging, reflected by a higher FTIR index, results in asphalt binder stiffening, which directly leads to higher viscosity and ZSV values, as well as an increase in the G^*/sinδ parameter. Accordingly, positive correlations are expected between the FTIR aging index and these rheological aging indicators.

The parameters derived from fitting the 2S2P1D model inherently reflect the effects of aging on the viscoelastic properties of asphalt binders. Specifically, the parameters β, α, and τ increase with advancing aging level. Therefore, positive correlations are anticipated between β, α, and τ and the FTIR aging index, viscosity, ZSV, and G^*/sinδ, reinforcing their potential as comprehensive aging indicators.

The Pearson correlation coefficient was used as an exploratory tool to provide an initial assessment of potential linear relationships among the investigated aging indicators, supported by visual inspection of the data. Table 3 presents the correlation matrix for the FTIR aging index, Log(ZSV), and G^*/sinδ, along with their correlations with other aging indicators, including the 2S2P1D model parameters and high-temperature viscosity.

Table 3 The pearson correlation coefficients between the investigated aging indices/indicators^*.

Evaluation of parameter pairs with Pearson correlation coefficients exceeding 0.8 indicates that the FTIR aging index exhibits linear correlations with both the 2S2P1D α parameter and Log(ZSV) when analyzed within individual binder grades. In contrast, the relationship between the Superpave rutting parameter (G^*/sinδ) and apparent viscosity shows a consistent linear trend not only within each binder group but also across the combined dataset. These observations suggest that certain correlations are grade-dependent, whereas others appear to be more robust across different binder grades. Accordingly, the identified correlations are interpreted as trend-based indicators of aging behavior rather than universal predictive relationships. Given the limited number of data points available for each binder, the analysis is exploratory in nature, and further studies involving a broader range of binder sources are required to evaluate the statistical robustness and generality of these relationships.

Figure 15a,c illustrate the linear correlation between the FTIR aging index and the Log(ZSV) for the 40–50, 60–70, and 85–100 base binder groups, respectively. The results reveal a robust linear relationship between these two aging indicators across all aging conditions within each base binder group. This strong correlation implies that, for a specific binder, one index can be reliably predicted from the other at any aging state (or equivalent service life) using a linear equation calibrated with as few as two data points (e.g., from RTFO and PAV-aged samples).

Fig. 15

Linear correlation between Log(ZSV) and FTIR aging index for the sample group of (a) 40–50; (b) 60–70; and (c) 85–100 base binder.

Figure 16a,c present the linear correlation between the FTIR aging index and the α parameter of the 2S2P1D model for the 40–50, 60–70, and 85–100 base binder groups, respectively. The results indicate a strong linear relationship between the α parameter and the FTIR aging index within each base binder group. This correlation enables accurate estimation of one indicator from the other using only two data points at any aging level or equivalent service life.

Fig. 16

Linear correlation between α parameter of 2S2P1D model and FTIR aging index for the sample group of (a) 40–50; (b) 60–70; and (c) 85–100 base binder.

Another pair of aging indicators exhibiting a strong linear correlation is apparent viscosity at 135 °C and G^*/sinδ, as shown in Fig. 17a,c.

Fig. 17

Linear correlation between apparent viscosity at 135 °C and G^*/sinδ at 60 °C for the sample group of (a) 40–50; (b) 60–70; and (c) 85–100 base binder.

It is noteworthy that, at relatively high temperatures (40 °C and above), a linear correlation is observed between the apparent viscosity at 135 °C and G^*/sinδ, irrespective of the temperature at which G^*/sinδ is measured. In other words, the apparent viscosity measured at 135 °C exhibits consistent linear relationships with G^*/sinδ, and these relationships remain valid regardless of the test temperature used to determine G^*/sinδ.

Interestingly, when the data from all three base binder groups are considered collectively, a strong linear correlation between the apparent viscosity at 135 °C and G^*/sinδ persists, as illustrated in Fig. 18.

Fig. 18

Linear correlation between apparent viscosity at 135 °C and G^*/sinδ for all samples.

Although the linear relationship presented in Fig. 18 may not be universally applicable to all neat asphalt binders, it highlights the potential to predict the properties of one binder sample with high accuracy using data from another one at any aging level or stage of service life, provided that the binders originate from the same crude oil source.

Non-linear relationships between the high temperature aging indices/indicators

Following the evaluation of the linear correlations between the investigated aging indices/indicators, simple logarithmic, exponential, or power-law relationships were also examined. From a mathematical perspective, it is almost always possible to establish a highly accurate relationship between two numerical datasets regardless of their complexity. However, such relationships, derived from a limited number of observations for specific samples, may lack fundamental significance when interpreting the underlying connections between material properties and governing parameters. In contrast, simple and interpretable relationships that reliably describe the variation trend of parameters are more desirable for developing robust predictive models and improving mechanistic understanding.

The non-linear correlations between the investigated indices/indicators were evaluated on a pairwise basis. The following presents the relationships between indices/indicators that exhibited robust correlations.

A strong non-linear correlation was observed between the FTIR aging index and the β parameter of the 2S2P1D model when the data from all three sample groups were analyzed collectively. Correlation analysis revealed that the FTIR aging index follows a semi-logarithmic relationship with β, as illustrated in Fig. 19, corresponding to a linear trend between the FTIR aging index and the logarithm of β on a semi-logarithmic scale.

Fig. 19

The logarithmic relationship between the FTIR aging index and the β parameter of the 2S2P1D model for all samples.

A similar semi-logarithmic relationship is observed between the τ parameter of the 2S2P1D model and the FTIR aging index within each base binder group, as shown in Fig. 20a,c. When the FTIR aging index and τ data from all samples are considered collectively, the overall correlation trend remains semi-logarithmic, although with a reduced coefficient of determination.

Fig. 20

The logarithmic relationship between the FTIR aging index and the τ parameter of the 2S2P1D model for (a) 40–50; (b) 60–70; and (c) 85–100 base binder samples.

It is worth noting that, although a direct mechanistic formulation linking FTIR aging indices to the 2S2P1D parameters (including α, β, and τ) is not currently feasible, the observed correlations reflect consistent structure–property relationships governed by oxidative aging mechanisms.

Fig. 21

The power-law relationship between the FTIR aging index and the rutting parameter G^*/sinδ for (a) 40–50; (b) 60–70; and (c) 85–100 base binder samples.

Another strong correlation identified is between the rutting parameter G^*/sinδ and the FTIR aging index, demonstrating that rheological stiffening due to aging is strongly linked to the chemical changes detected by FTIR. These two aging indicators exhibit a power-law relationship within each base binder group samples, as illustrated in Fig. 21a,c.

As expected, apparent viscosity at 135 °C and Log(ZSV) also exhibit a strong correlation, since both indicators reflect high-temperature viscosity characteristics. Considering the combined dataset from all three sample groups, a semi-logarithmic relationship was observed between the apparent viscosity at 135 °C and Log(ZSV) at 60 °C, as illustrated in Fig. 22.

Fig. 22

The logarithmic relationship between the apparent viscosity at 135 °C and Log(ZSV) for all samples.

Another strong correlation revealed from the analysis is between Log(ZSV) and the Superpave rutting parameter G^*/sinδ, both at 60 °C. Considering the combined data from all three sample groups, a strong exponential relationship exists between Log(ZSV) and G^*/sinδ, as shown in Fig. 23. This finding corroborates previous studies identifying ZSV as a reliable alternative to the Superpave G^*/sinδ parameter for the assessing asphalt binder rutting resistance^26,46.

Fig. 23

The exponential relationship between Log(ZSV) and G^*/sinδ observed in the combined data from all samples.

Fig. 24

The logarithmic relationship between the apparent viscosity at 135 °C and the FTIR aging index for (a) 40–50; (b) 60–70; and (c) 85–100 base binder samples.

The final strong correlation identified is between the FTIR aging index and apparent viscosity at 135 °C within each base binder group, exhibiting a semi-logarithmic relationship, as illustrated in Fig. 24a,c.

To clarify how the identified correlations may be applied in practice, a trend-based interpretation framework is proposed, as illustrated in Fig. 25. The applicability of surrogate aging indices depends on the purpose of analysis, the availability of experimental data, and the consistency of binder source and grade. For correlations that are valid only at the individual binder-grade level, estimation of one aging indicator from another requires calibration of the identified trend using a limited number of data points (at least two for linear trends and three for non-linear trends) at the relevant aging levels.

For correlations that may be valid across multiple binder grades derived from the same crude oil source, calibration using data from a single binder grade enables estimation of aging indicators for other grades from the same source. This framework does not propose universal replacement of complex tests; rather, it provides a practical pathway for using simplified surrogate indices when appropriate, while acknowledging the source-dependent and trend-based nature of the proposed correlations.

Fig. 25

Proposed workflow for the practical use of trend-based correlations among asphalt binder aging indicators.

Summary and conclusion

The results of this study demonstrated that oxidation-induced changes in the chemical structure of asphalt binders are consistently reflected in variations across all investigated aging indices/indicators. The correlation trends identified among these aging indices/indicators can be used not only as a basis for predicting their values at different aging levels (or stages of service life) but also for interconversion between them to estimate unavailable aging indices/indicators.

Among the rheological aging Indicators, apparent viscosity at 135 °C and α, β, Log(ZSV), and G^*/sinδ at 60 °C exhibited strong correlations with the FTIR aging index, which reflects chemical changes in asphalt binders resulting from oxidative aging. In addition, the rutting parameter G^*/sinδ at 60 °C showed strong correlations with both Log(ZSV) at 60 °C and apparent viscosity at 135 °C.

Table 4 summarizes the correlations identified among the investigated aging indices/indicators. These correlations were found to be valid for samples at different aging levels within a specific asphalt binder grade. However, several correlations may also hold when all samples are considered collectively, provided that the different neat binder grades originate from the same crude oil source, as demonstrated in Table 4.

Table 4 Summary of correlations found between the high temperature aging indices/indicators.

Future work

Suggested directions for future research include:

• Including asphalt binder samples derived from different crude oil sources and evaluating the correlations among these samples.

• Incorporating modified asphalt binders produced from the investigated neat asphalt binders and examining the correlations within these materials.

• Evaluating the correlation trends for laboratory-aged and field-aged samples.

• Determining J_nr​ parameters of the samples as aging indicators using the Multiple Stress Creep Recovery (MSCR) test and assessing their correlations with the aging indices/indicators investigated in this study.