1 Introduction

Asphalt mixtures are a widely used material in the road pavement industry. For example, Europe produced 280 million tons of hot and warm mix asphalt in 2019. This composite material uses bitumen obtained by crude oil refinement as a binder [1].

Due to the oxidation of bitumen as an organic material, the visco-elastic properties change over time. This leads to increased stiffness and brittleness, which results in the deterioration of the performance of the respective asphalt mixtures. The aging of asphalt binders and mixtures is divided into short-term aging (STA) and long-term aging (LTA). STA occurs during production and paving and is characterized by fast oxidation and vaporization of remaining volatile components due to high temperatures. LTA is a slower process occurring during the service life of asphalt pavements. Natural and anthropogenic impacts trigger slow oxidation, particularly in the upper centimeters of (dense) asphalt pavements. The main impact factors are reactive oxygen species (ROS) in the atmosphere, like ozone (O3) or nitrogen oxides (NOx), UV/solar radiation, and moisture [2,3,4,5,6,7,8,9].

Accelerated aging simulation of bituminous materials is a time-efficient tool to predict long-term changes in material behavior. This information can be used for mix design optimization and thus, help to extend the asphalt pavements’ lifetime and increase their recyclability by selecting the most suitable binder and ultimately developing more durable mixtures. Two methods are commonly used on the bitumen scale: For STA, the Rolling Thin Film Oven Test (RTFOT) is a reliable procedure to simulate production and mixing. For LTA, Pressure Aging Vessel (PAV) imitates a service life of 5 to 10 years [10, 11]. Various procedures have been developed on asphalt mix scale: They differ by either conditioning compacted specimens or loose mix and aging medium. Many methods use high temperatures (above + 100 °C) and/or high pressures that do not occur in the field [12, 13] to achieve long-term aging in a time-efficient manner. Due to these temperature and/or pressure differences, other reaction mechanics and kinetics can participate in the oxidation mechanism and thus differ from field aging.

A new procedure for aging compacted cylindrical asphalt mix specimens has been developed recently. The so-called “Viennese Aging Procedure” (VAPro) uses a highly reactive gas mixture to age the asphalt specimens in conditions similar to the field (+ 60 °C and pressure < 0.5 bar) [14, 15]. The gas mixture contains traces of reactive oxygen species (ROS), which are ozone (O3) and nitrogen oxides (NOx). These oxidants are also present in the atmosphere but in lower concentrations. Besides OH-radicals, O3 and NOx play a significant role in the tropospheric cycle. Thus, being one of the main drivers of tropospheric aging processes, they contribute to the idea of a realistic aging simulation [16,17,18,19]. Prior studies have shown the significant contribution of O3 and NOx on asphalt aging [20].

The aging of compacted specimens is reported to produce an aging gradient within the specimen when using oven aging or UV aging. This reflects the aging gradient occurring in the field [21, 22]. However, homogeneous oxidation is more suitable for material testing and assessment. Having this in mind, the idea of VAPro is as follows: By perfusing compacted specimens (1.0 l/min) with the above-mentioned reactive gas mixture, more homogeneous aging is targeted. Consequently, an air void content of at least 6.0% by volume is needed to maintain a sufficient gas flow through the specimens.

The main objective of this paper is to evaluate the applicability of VAPro. Therefore, asphalt mixes with the same grading curve but with bitumen of different origins are produced and aged with VAPro. At the same time, the binders are aged with RTFOT and PAV. For binder evaluation, the binders of the VAPro-aged specimens are extracted with tetrachloroethylene as a solvent and distilled according to EN 12697–3 [23]. The binders are analyzed by Dynamic Shear Rheometer (DSR) and Fourier-Transform Infrared (FTIR) Spectroscopy.

2 Materials and methods

2.1 Materials

This study uses asphalt concrete following EN 13108–1 [24] with a maximum nominal grain size of 11 mm (AC11) and a binder content of 5.2% by mass. Figure 1 shows the grading and limiting curves according to EN 13108–1. The fine and coarse aggregates of the mix are of porphyritic origin, and the filler material is powdered limestone.

Fig. 1
figure 1

Sieving curve of the used asphalt mixture

Full size image

Four asphalt binders of the same grade (70/100 pen according to EN 12591 [25]) were selected from different sources to produce four mixes differing only in the binder used. In Table 1, various binder specifications are given for comparison. Since they are all graded the same, all four binders have similar values within the allowed limits. According to EN 12697–35 [26], a laboratory mixer is used to prepare the mixture at a temperature of + 170 °C. HMA slabs (50 × 26 × 4 cm) are compacted using a steel roller segment compactor according to EN 12697–33 [27].

Table 1 Binder characteristics
Full size table

From each slab, eight specimens with a diameter of 100 mm are cored. According to EN 12697–8 [28], the air void content of the specimens ranges from 4.8 to 9.2% by volume. The bulk density required for the void content calculation is determined according to EN 12697–6, procedure B [29]. For each mix, three specimens with an air void content of 7.0% ± 1.0% by volume are selected for aging. This relatively high air void content is necessary for VAPro-aging. The reasons for this are described in detail below.

2.2 Methods

2.2.1 Aging procedures

Rolling thin film oven test (RTFOT) and pressure aging vessel (PAV). All four virgin binders are aged in RTFOT according to EN 12607–1 [30] at + 163 °C. Subsequently, PAV is carried out according to EN 14769 [31] at + 100 °C and 2.1 MPa.

The Viennese aging procedure (VAPro). In the meantime, two independent aging cells have been developed: VAProcyl for aging up to three cylindrical specimens with a diameter of 100 mm and VApropri to obtain prismatic specimens [14, 32]. The asphalt mix specimens in this study are aged with VAProcyl. A single specimen is conditioned during each aging cycle. The centerpiece of this method is an ozone generator that produces a gas mixture containing traces of ROS. The gas mixture is produced by directing air from the central laboratory supply through the ozone generator. Using a dielectric discharge tube, the generator enriches the air with ozone and nitrogen oxides. The aging cell of VAProcyl is a triaxial cell in which the specimen is located and covered with a silicone membrane. Figure 2a shows the aging cell. An overpressure of ~ 1.0 bar in the triaxial cell ensures that the elastic membrane is pressed against the specimen to maintain the gas flow through the specimen. A flow meter and a pressure regulator upstream of the generator ensure a constant flow (~ 1.0 l/min) and pressure (p ≤ 1 bar) in the system. A heating device is installed between the generator and the triaxial cell to heat the gas. The triaxial cell and the heating device are placed in a heating cabinet at a temperature of + 60 °C (= Toven). In Fig. 2b, schematics of the VAPro principle are shown.

Fig. 2
figure 2

a photo of the aging cell (with cut-out to see the specimen), b VAPro schematics

Full size image

The aging duration was reduced compared to earlier studies since VAPro was continuously improved [15, 33]. With the latest setup, the aging duration was set to three days since the extracted binders achieved at least a level of RTFOT + PAV-aging after this duration [20]. After passing the triaxial cell, the gas mixture is led through a washing flask filled to degrade the ROS before the gas is sucked away by a fume hood.

2.2.2 Mechanical analysis

Binder extraction and recovery. To analyze the binders from the asphalt mixes, they are extracted with tetrachloroethylene as a solvent and distilled according to EN 12697–3 [23].

Binder testing. A DSR, according to EN 14770 [34], is used to investigate the rheology of the binders. A 25 mm plate with a gap of 1 mm is used in the upper-temperature range from + 40 °C to + 82 °C; from −4 °C to + 40 °C, an 8 mm plate with a gap of 2 mm is used. A strain-controlled frequency sweep is carried out from 0.1 Hz to 40 Hz. Dynamic Shear Modulus |G*| and phase angle δ are obtained from these measurements.

Mastercurve construction. Based on the time–temperature superposition principle (TTSP), the relationship between frequencies and temperatures allows us to gain information on material behavior on a broader scale than what is measurable. The modified Williams–Landel–Ferry (WLF) function proposed by Kaelble [35] is used to determine the time–temperature shift factors. Rowe and Sharrock [36] rearranged the original function for easier application, as shown in Eq. (1).

$$ \log a_{T} = - C_{1} \left( {\frac{{T - T_{k} }}{{C_{2} + \left| {T - T_{k} } \right|}} - \frac{{T_{ref} - T_{k} }}{{C_{2} + \left| {T_{ref} - T_{k} } \right|}}} \right) $$
(1)

The mastercurves of |G*| and δ are determined using the Generalized Logistic Sigmoidal Model [37, 38]. Originally proposed for asphalt mixtures, Yusoff, et al. [39] demonstrated that the Generalized Logistic Sigmoidal Model shows a better correlation between measured and fitted data than CA or CAM models. The function used to calculate |G*| and δ mastercurves are shown in Eqs. (2) and (3).

$$ |G^{*} | = \partial + \frac{\alpha }{{\left[ {1 + \lambda e^{{\beta + \gamma \left( {\log \left( {fr} \right)} \right)}} } \right]^{{{\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 \lambda }}\right.\kern-0pt} \!\lower0.7ex\hbox{$\lambda $}}}} }} $$
(2)
$$ \delta = \frac{{90\alpha \gamma e^{{\beta + \gamma \left( {\log \left( {fr} \right)} \right)}} }}{{\left[ {1 + \lambda e^{{\beta + \gamma \left( {\log \left( {fr} \right)} \right)}} } \right]^{{\left( {1 + {\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 \lambda }}\right.\kern-0pt} \!\lower0.7ex\hbox{$\lambda $}}} \right)}} }} $$
(3)

The mastercurves are constructed with a widely known calculation software using non-linear least squares regression techniques. The best fit is found by minimizing the sum of squared errors (SSE) between measured and fitted data. From the mastercurve calculations, a variety of parameters can be obtained:

  • Crossover Temperature Tc

  • Crossover frequency fc

  • Crossover Modulus Gc

  • The rheological parameter R

  • G-R parameter

The crossover temperature Tc (where the phase angle is 45°) increases with aging. As aging progresses, the elastic behavior becomes more dominant at the given frequency. Thus, a higher reference temperature must be applied to reach the desired transition point. This parameter is successfully used in literature to describe the aging of asphalt binders [40,41,42,43].

At the same time, the crossover frequency fc becomes lower with aging. The crossover modulus Gc (at fc where G’ = G’’) undergoes the same evolution. The rheological parameter R (R-value) is an interesting parameter for mastercurve assessment. It describes the relation between Gc and the glassy modulus Gg (Gg = 109 MPa) and is expressed by the log of the ratio of Gg and Gc. The R-value increases with aging. Thus, lower values mean more favorable mechanical performance. It was identified in the SHRP program as sensitive to aging and is widely used to evaluate bituminous binders [44–46].

For evaluation of susceptibility to embrittlement and cracking, the G-R parameter introduced by Glover, et al. [47] and rearranged by Rowe, et al. [48] (see Eq. 4) is used. The G-R parameter is calculated at 15 °C and 0.005 rad/s. For this study, the input for the G-R parameter is obtained from constructed mastercurves at 15 °C.

$$ G - R = \frac{{G^{*} *\left( {\cos \delta } \right)^{2} }}{\sin \delta } $$
(4)

Rowe, et al. [48] suggests a warning limit for damage onset at 180 kPa and a critical limit for significant cracking at 600 kPa. Thus, a higher G-R parameter indicates increased embrittlement.

Four mastercurves for each sample are constructed at Tc, 15, 25 and 35 °C. This allows for a small statistical analysis of the master curve design: The temperature-independent parameters R and crossover modulus show a very small standard deviation of a maximum of 0.1%.

The Crossover temperature is obtained by running through the calculation procedure with the reference temperature as a variable. The frequency turns into the reference value and is set to 1.59 Hz. In addition, the requirement is set that at the given frequency phase angle has to be at 45°. Thus, the reference frequency becomes the crossover frequency at the same time here. The calculated crossover temperature (and the mastercurve construction) is validated by comparing it with the crossover temperature retrieved from the 8 mm plate DSR measurements by linear interpolation. Figure 3 shows an excellent correlation between the measured and calculated crossover temperature. The model slightly underestimates the crossover temperature.

Fig. 3
figure 3

Correlation between measured and calculated crossover temperature

Full size image

2.2.3 Chemical analysis

Fourier-transform infrared (FTIR) spectroscopy. A Bruker Alpha II with an ATR diamond crystal module is used for the FTIR measurements. For each binder sample, four individual samples are measured. A background scan of the empty crystal is recorded before every measurement. For each sample, the device performs four consecutive measurements considering 24 scans. This results in 16 single measurements per binder sample. The spectra are recorded within the wavelength range of 600—4000 cm−1 at a resolution of 4 cm−1.

Sample preparation is as follows: A small amount of binder is heated to 130–160 °C on a spoon and homogenized. Heating and homogenizing are done within 60–120 s. The temperature is measured with a thermometer which is used for stirring at the same time. Next, four droplets per sample are applied on small pieces of silicone foil, which are then covered with a tin lid. The droplets cool down for 5 to 10 min before being applied to the ATR crystal for measurement.

The measured spectra are normalized in the fingerprint area from 600–1900 cm−1 using min/max normalization. The minimum can be found in the wavelength range of 1900–1800 cm−1, where no functional group appears. The maximum is located around 1460 cm−1, where the aliphatic band for reference appears. This approach results in less scattering due to the noise/artifacts generated by the ATR crystal in the wavelength range of 1950–2250 cm−1. Figure 4 shows two example spectra of a virgin and an LTA-aged binder: The increase in carbonyl and sulfoxide bands with aging is easily identified. The box illustrates the random noise found in the lowest part of the spectra that can bias the analysis.

Fig. 4
figure 4

Normalized FTIR spectra examples of an unaged and LTA-aged binder, including the illustration of the signal noise part around 1950–2250 cm−1

Full size image

For binder assessment, the following integrations are made by using the full baseline integration method:

  • Carbonyls (AICO): wavelength band: 1665–1725 cm−1

  • Sulfoxides (AISO): wavelength band: 984–1070 cm−1

  • Reference aliphatic band (AICH3): wavelength band: 1355–1490 cm−1

The Carbonyl Index IC=O and Sulfoxide Index IS=O are calculated by formulas (5) and (6) from these integrals. The Sulfoxide Index IS=O is not adopted for further analysis: Inconsistencies can be found in IS=O data when using recovered binders from asphalt mixtures. Some interactions with the aggregates during mixing or minimal amounts of left filler in the bitumen samples might bias the measurements, leading to dubious results [46, 49]. Similar effects are found within the analyzed data for this paper. Therefore, this work uses only the carbonyl index for chemical analysis. More research on this specific issue is needed.

$$ I_{CO} = \frac{{AI_{CO} }}{{AI_{{CH_{3} }} }} $$
(5)
$$ I_{SO} = \frac{{AI_{SO} }}{{AI_{{CH_{3} }} }} $$
(6)

3 Results and discussion

As described in part 2.2, all four binders in their five aging states (Virgin, RTFOT, RTFOT + PAV, LAB MIX (= after slab production), VAPro) are analyzed by DSR and FTIR.

Figure 5 shows the master curves of all samples at + 25 °C. The aging-induced stiffening and embrittlement can be observed on all four binders: The complex modulus |G*| is increasing, especially in the lower time–temperature domain, and the phase angle δ is decreasing. Binder 4 seems to be the most aging susceptible one with the highest stiffness increase and phase angle loss, followed by binder 3. Binders 1 and 2 show quite similar evolution with aging at first glance. The following parameters are used for a thorough assessment of the aging behavior. The results for each sample are listed in Table 2.

  • The rheological parameter R (R-Value)

  • Crossover temperature Tc

  • Carbonyl index IC=O

  • G-R parameter

Fig. 5
figure 5

Mastercurves for shear modulus G* and phase angle δ of all samples at 25 °C; a Binder 1, b Binder 2, c Binder 3, d Binder 4

Full size image
Table 2 Results of the binder analysis
Full size table

Comparing the crossover modulus Gc given in Table 2, one can observe different development of the respective binders: Binders 1, 2 and 3 have the highest values in the virgin state, ranging from 16 to 27 GPa. Binder 4 is far below with a Gc of 8 GPa, which is roughly in the range of the LTA-aged samples of Binders 1, 2 and 3. This is quite an interesting observation since all four binders have the same grade.

For further mechanical analysis of the investigated binders, attention is put on the R-value and the crossover temperature Tc. Figure 6 shows the link between these two parameters: Considering all samples, a linear correlation with a coefficient of determination of 0.84 can be identified. For each binder, a good linear relationship with aging is observed (R2 ≥ 0.95). Every single data point consists of three individual measurements. Standard deviation is between 4 and 5%.

Fig. 6
figure 6

Link between R-value and crossover temperature Tc

Full size image

In the virgin state (circles), binder 1 shows the lowest values with an R-value of ~ 1.57 and Tc = 5.6 °C. Binder 2 and 3 behave pretty similar with an R-value of ~ 1.75 and a Tc of ~ 10.0 °C. Binder 4 has the highest values in virgin state with an R-value of 2.1 and a Tc of 11.6.

The STA samples (rhombuses) of binders 1, 2, and 3 range from 1.7 to 1.9 for R-value and from 9 to 14 °C for Tc. This shows that binder 4, in its virgin state, is already at the same level as the STA samples of the other three binders. Although graded the same, binder 4 performs worse than binders 1, 2, and 3 in virgin state and appears already aged.

A similar trend can be observed when looking at the LTA samples (triangles): The STA-sample of binder 4 is already in the range of the LTA samples of binders 1, 2 and 3. The sample of Binder 4 aged with VAPro has the highest values, indicating that it is the most aged or most susceptible to aging. VAPro-aging is also more severe than PAV-aging for all the binders examined.

Similar trends can be seen when looking at the G-R-Parameter in Table 2: VAPro-aging always produces higher values than PAV. The ranking observed in the R-Tc-Diagram is also visible within the G-R-Parameters, identifying binder 4 as most prone to cracking. This shows that VAPro helps in identifying aging susceptible binders. It can point out the differences more pronounced than PAV does. R-Value, crossover temperature Tc and G-R-Parameter are helpful parameters for evaluating the mechanical data. The results show that bitumen of the same grade can have different aging behavior.

To investigate the oxidation state of the bitumen samples, an ATR-FTIR is used. This method is fast and easy to apply. As described in part 2.2.3, only the Carbonyl Index is used for evaluation. Figure 7 shows the correlation between R-Value and the Carbonyl Index Ico.

Fig. 7
figure 7

Correlation of R-value and carbonyl index Ico

Full size image

For each bitumen, a good linear correlation is observed. The carbonyl index supports the assumption that bitumen 4 is already aged in its virgin state: It is 0.05 for binder 4, which is on par with the STA samples of binders 1, 2, and 3, with an average of 0.03. You also see the STA of binder 4 is already at the LTA level of binders 1, 2 and 3 and binder 4 is the most oxidized in this diagram. That shows similar behavior as described in Fig. 6. This confirms that binder 4 is more susceptible to aging than the others and is already oxidized in virgin state.

This data set shows that the toolkit of mechanical analysis and chemical analysis, the chemical–mechanical analysis so to speak, is excellent for identifying suitable asphalt binders and sorting out those with poor aging behavior. VAPro-aging points out these differences more clearly than PAV does. Its close-to-field conditions may help evaluate new asphalt mix designs and new additives such as recycling agents or antioxidants more reliably.

4 Summary

This paper shows the application of a recently developed asphalt mix aging method–The Viennese Aging Procedure. VAPro simulates aging processes by perfusing cylindrical specimens with ozone- and nitrogen oxide-enriched air at realistic ambient conditions in terms of temperature and pressure (T = 60 °C and p ≤ 1 bar). Besides the challenge of achieving a homogeneous air void content required for sufficient perfusion, the application of the new aging method was efficiently executed.

The aging behavior of four bitumen of the same grade but different origins is investigated. For reference, the binders are aged in RTFOT and PAV as well. A DSR and an FTIR are used for chemo-mechanical analysis. Three days of VAPro-aging is more severe than RTFOT + PAV–the stiffness increase of the extracted bitumen after VAPro is 1.2 to 2.6 times the RTFOT + PAV level. The dynamic shear modulus |G*| or phase angle δ from DSR and the carbonyl index derived from FTIR show a good correlation and further help evaluating the binders. The sulfoxide index shows differences between VAPro and RTFOT + PAV, possibly due to interactions with the aggregates not present in the PAV. Thus, it is not used for analysis. More research on that phenomenon is needed.

Although being the same penetration grade, the binder source significantly impacts aging susceptibility. This is observable from PAV, but VAPro can distinguish these differences more clearly. Differences seen on the mechanical side can also be overserved by the oxidation state of the binders. FTIR is an excellent tool for evaluating the oxidation of the material [50–53]. This method reveals that one of the virgin binders is already oxidized. Subsequently, this binder shows a higher aging susceptibility. The shown trends in aging susceptibility also correspond to the results of the G-R parameter. In summary, chemo-mechanical analysis offers a valuable tool-set for sorting out poor performing binders. Further research is in progress, discussing various adaptions of the aging setup, applying the method on binder level for a more fundamental approach, analyzing field aging and VAPro mechanisms and investigating additive behavior with the new approach [54, 55].