Influence of selected reactive oxygen species on the long-term aging of bitumen

The aging of bitumen is a major contributor to the failure of asphalt pavements. Realistic and accurate laboratory aging methods can predict bitumen durability and guarantee the use of high-quality components in asphalt pavement. However, current standardized aging methods do not incorporate atmospheric parameters, besides elevated temperatures and molecular oxygen. Crucial chemical components like reactive oxygen species (ROS), e.g. nitrogen oxides (NOx) or ozone (O3), are completely neglected. This study focusses particularly on the reactivity of individual ROS, such as nitrogen monoxide (NO), nitrogen dioxide (NO2) and O3, in regards to the long-term aging (LTA) of three unmodified bitumen. For LTA an adapted version of the Viennese Binder Aging method was used and the aged bitumen samples were analyzed with the dynamic shear rheometer and Fourier-Transform-Infrared spectroscopy, respectively. The results show that NO as a single component does not induce significant aging, whereas NO2 leads to severe bitumen deterioration, which is even more accelerated when a second oxygen source is present. In comparison, the reactivity of O3 is rather mild and it did not cause additional aging for two of the investigated binders. This study provides evidence, that ROS play a crucial role in bitumen aging and should thus not be neglected when addressing realistic aging conditions in the laboratory.

sulfur, which makes it an organic material [6]. Hence, contact of bitumen with oxygen leads to aging, causing the material to become stiffer, more brittle and susceptible to cracking over time [7]. Bitumen aging is typically described by two processes and simulated separately: (1) Short-term aging (STA) of bitumen occurs during construction and production caused by high temperatures and exposure to air. This process is simulated in the laboratory with the Rolling Thin Film Oven Test (RTFOT) [8].
(2) Long-term ageing (LTA) occurs during the service-life of bitumen and is caused by a variety of environmental phenomena, such as elevated temperatures, light, moisture and O 2 and reactive oxygen species (ROS) in the atmosphere. LTA is simulated in the laboratory with the Pressure Aging Vessel (PAV) [9]. During PAV aging a high temperature and pressure is applied to accelerate the aging process and to induce strong oxidation over a short period of time.
However, important aging inducing factors of the environment (moisture, light and ROS) have not been taken into account and are therefore not incorporated in the standardized laboratory long-term aging methods. Several studies have already reported an effect of moisture and light on bitumen and asphalt aging. First of all, water in combination with mechanical stresses can cause the loss of strength and durability, leading to debonding of the binder and the aggregate and adhesive or cohesive failure [10][11][12][13]. Furthermore, the interaction of bitumen with aqueous solutions of NaCl and Na 2 SO 4 can change the surface roughness of asphalt mixtures and accelerates the erosion process according to Zou et al. [14]. Additionally, the effect of moisture damage coupled with oxidative aging has been addressed in literature using mostly a theoretical approach. Ma et al. state that the chemical structure and composition, viscosity and morphology are some of the important factors influencing the behavior towards aging induced by water and oxygen [15]. Das et al. show indications for the formation of water soluble thin films due to aging [16]. Albeit, there have also been studies indicating that water has very little influence on bitumen and asphalt aging [17] or that the influence of moisture varies among different bitumen samples [18], which underlines the need for further research concerning the impact of water.
A similar picture is shown regarding studies focusing on the effect of light on bitumen aging. There is a general agreement in literature that different wavelengths induce different degrees of aging [19,20], but the mechanisms behind this phenomenon have not yet been fully uncovered. Light in the energyrich UV range certainly has a severe impact on bitumen aging, but according to recent studies of Mirwald et al. also light at longer wavelengths, i.e. in the visible region, has a strong aging potential [21,22]. Feng et al. have separated bitumen into the SARA fractions and have shown that the fractions on their own are more prone to UV-aging than the unseparated, original bitumen. Especially the saturates display the most changes after exposure to UV light [23]. Furthermore, the effect of ultraviolet absorbers (UVA) has been studied to a certain degree in various experimental setups. It is reported that UVAs can improve the UV aging and also the thermal-oxidative aging resistance [24], but the effect is strongly influenced by additional modifiers (e.g. polymers) [25,26].
The third aforementioned parameter, that has mostly been disregarded when looking at bitumen aging in the past, is the influence of the atmosphere and therein occurring reactive oxygen species like ozone (O 3 ) and nitrogen oxides (NO x ). Ozone is formed in the atmosphere due to oxidation of hydrocarbons and carbon monoxide (CO) [27] and NO x are emitted from car engines [28] in close proximity to asphalt pavements in the traffic system. The number of literature references on the effect of realistic aging conditions (including ROS) on bitumen or on other related materials is very limited and scattered, however an attempt is being made to give a cohesive picture of the so far available data. The influence of SO 2 and NO x in combination with moisture on polymeric films was examined by Pagnin et al. [29]. The impact of the different gas species varied between the investigated films, whereby NO x caused more distinctive morphological changes, SO 2 interacted more with oily substances and styrene-acrylic films showed a higher sensitivity to degradation reactions in combination with humidity. Middleton et al. [30] studied how polymer degradation was influenced by ozone at elevated and at room temperatures and how it compared to aging in air at high temperatures. It was found that at a temperature of 140°C and for an aging duration of 90 days the effect of 1% ozone can be disregarded, since the high temperature is the driving force of the degradation process. However, it is believed that for shorter periods of time and lower temperatures closer to bitumen field conditions the impact of ozone and other ROS on the aging process is more severe. Focusing on the determination of the underlying mechanisms of oxidation Hoffmann et al. [31] investigated photo-oxidation of hydrocarbons, whereby laser pulses initiated the oxidation process and a time resolved detection of OH and NO 2 radicals was performed. With this technique the oxidation rates and products of hydrocarbons can be observed and it can be applied to the study of volatile organic compounds (VOC), which occur in bituminous materials. Furthermore, field studies with polymers and with bitumen have been performed, thereby naturally incorporating sunlight, moisture and atmospheric ROS. The aging of polyethylene and styrene-butadiene was investigated by Getlichermann et al., who exposed polymer films directly to solar radiation for 15 days to uncover specific degradation mechanisms in polymers [32]. Soenen et al. [33] examined the rheological and chemical properties of field aged binders, showing similar trends between field and laboratory aged samples when looking at the stiffness and the carbonyl index. However, the aging level of the field aged samples surpassed the laboratory aged samples, indicating that more advanced aging methods could lead to a better correlation. Masson et al. [34] performed a field study with styrene-butadiene modified bitumen used as sealants, whereby samples were collected after one, three, five and nine years of service and then subjected to physico-chemical analysis. Occurring aging processes were confirmed to result in the formation of sulphoxides, sulphones, ketones and carboxylic acids, in the loss of aromatics and in block copolymer degradation. It is reported that the performance of the investigated bituminous sealants did not correlate well with standard aging tests, since these could not predict the performance of the materials under natural weathering conditions. The authors therefore recommended developing a more realistic aging method for bituminous materials.
The process of developing an aging method for bitumen that is time-efficient and as close to reality as possible has been gaining more attention in the last years. Hofko et al. [35] investigated the impact of ROS on asphalt mixtures by exposing the asphalt mix specimen to air, O 2 , O 3 and NO x . The results show that O 3 alone does not have a high oxidation potential, but in combination with NO x heavy aging is being induced. Furthermore, the effect of ROS on bitumen was studied by Mirwald et al. [36,37], who developed the Viennese Binder Aging (VBA) method [38] by adapting the previously developed Viennese Aging Procedure (VAPro) for asphalt mixtures [39,40]. During this aging method the bitumen samples are kept at 80°C, which correlates with field conditions, and exposed to air enriched with O 3 and NO x . Although the main aging inducing factors during VBA are the high concentrations of O 3 and NO x at elevated temperatures, the goal is to further advance the aging setup and incorporate as many aging inducing factors as possible occurring in the environment. It has been shown that ROS play an important role in the aging process of bitumen, but the specific impact of each gas species (and possible cooperative effects) have not yet been identified.
The purpose of this study was the determination and comparison of the effect of single ROS on bitumen aging. Therefore, three unmodified short-term aged bitumen were long-term aged using an adapted version of the VBA method [38], which allows the separate exposure of the sample to O 3 , NO and NO 2 in different carrier gases. These experiments were then compared to standardized aging methods (RTFOT ? PAV). The properties of the binders after aging were characterized using Fourier-Transform-Infrared (FTIR) spectroscopy and the dynamic shear rheometer (DSR). FTIR is a very helpful tool for the assessment of the chemical composition of bituminous materials and has been used progressively in the last two decades. Together with the mechanical information obtained from the DSR measurements the chemo-mechanical correlation of the acquired data can be determined.

Materials
For this study, three non-modified binders with the penetration grade 70/100 were used, whereby each binder was provided by a different refinery. Their basic properties are summarized in Table 1.
Several aging methods were performed and the binders were then analysed in their respective aging states: a) unaged, b) short-term aged and c) long-term aged. The standardized methods used were the Rolling Thin Film Oven Test (RTFOT) according to EN 12607-1 [8] for STA and the Pressure Aging Vessel (PAV) at 100°C according to EN 14769 [9] for LTA. Furthermore, PreVBA was used for STA and an adapted version of the VBA method for LTA. A detailed explanation of these methods will follow below.

''Pre '' Viennese binder aging (PreVBA)
For the ''Pre'' Viennese binder aging (PreVBA) method the respective bitumen was heated up to * 110°C in a small metal can for approximately 5 min and stirred with a thermometer for homogenization and temperature detection. After reaching the desired temperature, 8 g of the material were poured into a PAV metal container as evenly as possible and then transferred into a ventilated oven, where the sample was kept at 163°C for 75 min (similar to the conditions during RTFOT). The purpose of PreVBA is not only to achieve a STA condition, but also to prepare the sample, as it results in a uniformly distributed bitumen film with a thickness of 0.5 mm. (Fig. 1). After PreVBA the bitumen sample is transferred into the VBA aging cell for LTA.

Viennese binder aging (VBA)
The VBA is an aging method introduced by Mirwald et al. [38] as a way to simulate LTA of bitumen in a more realistic way using ROS in combination with an elevated temperature as aging inducing factors. More detailed information on the reason for incorporating ROS, on the separate parts of the aging setup and the reactions taking place between bitumen and the different gas species can be obtained directly from Mirwald et al. [38]. In this method an ozone generator operated with air is used to produce O 3 and NO x , which are then led into an aging cell made out of stainless steel. Since the focus of the current study was to investigate the effect of single ROS on bitumen aging the VBA setup was adapted to allow separate exposure of the sample to O 3 , NO and NO 2 (see Fig. 2). Therefore, gas bottles with different concentrations of NO in N 2 and NO 2 in air and N 2 were connected to the aging cell. For the production of O 3 without the presence of NO x and to keep the overall oxygen content in the atmosphere approximately at the same level as in air, the ozone generator was operated with a mixture of 79% argon and 21% oxygen. The ozone generator was the type COM-AD-01 and the ozone analyzer, which was used for the determination of the ozone concentration in the gas flow, was the GM-OEM both from the company ANSEROS.
Each binder was exposed to 21 different atmospheres, whereby two bitumen films were aged simultaneously in order to increase efficiency. The composition of the respective atmospheres was as follows:   [41] and the average annual value for NO 2 is set at 30 lg/m 3 [42]. All experiments were conducted at 80°C, with an aging duration of 3 days and a gas flow of 50 ± 5 l/h, whereby the flow rate was modulated with floating ball flow meters.

Attenuated total reflection fouriertransformation infrared (ATR-FTIR) spectroscopy
For the detection of changes in the chemical composition FTIR spectroscopy was used. The respective bitumen samples were heated up to * 110°C in a spoon and stirred with a thermometer for homogenization and temperature detection. After reaching the desired temperature small droplets of sample were transferred on a silicone foil. The sample preparation is shown in Fig. 3. For the measurement a Bruker Alpha II with an DTGS detector and an attenuated total reflection (ATR) unit containing a diamond crystal was used. The ATR geometry allows the recording of FTIR spectra in adsorption equivalent mode and the bitumen samples can be applied directly onto the crystal in solid state. All spectra were measured from 4000-400 cm -1 at a resolution of 4 cm -1 (24 scans)  with four repetitions. For each aging state (which corresponds to a different atmosphere during VBA) four samples were prepared, resulting in 16 spectra per aging state. Each ATR-FTIR-measurement consisted out of the following steps: Firstly, the ATR crystal was cleaned with limonene and isopropanol and then a background spectrum was recorded. After applying the bitumen sample on the ATR crystal, the spectra were recorded. For spectra evaluation and processing the software OPUS was used. A normalization between 3200-2800 cm -1 was carried out and a full base line integration using the following frequency limits was performed: By calculating the Aging Index (AI) according to Eq. 1 the uptake in oxygen due to the aging process can be assessed.
The AI of each aging state shown in this paper is obtained by determining the mean value and standard deviation of the respective 16 spectra.

Dynamic shear rheometer (DSR)
For the detection of changes in the mechanical properties the DSR was used. The samples were heated up to * 110°C in a small metal can and stirred with a thermometer for homogenization and temperature detection. After reaching the desired temperature the bitumen sample was poured into a silicone mold and stored in the dark for 24

Chemo-mechanical correlation
The chemo-mechanical correlation illustrates the link between the mechanical properties and the chemical composition of the respective bitumen sample. The FTIR aging index, which represents the formation of carbonyls and sulfoxides, is plotted on the X-axis and the norm of the complex modulus |G*| or the phase angle d is plotted on the Y-axis. This gives visual and graphical information about how an increase in stiffness is correlated to an increase in incorporated oxygen and vice versa. For this study |G*| at 1.592 Hz and 46°C is used for the chemo-mechanical correlation.

NO in N 2
The impact of NO on bitumen aging was investigated by exposing the samples to an aging atmosphere consisting of N 2 with different concentrations of NO for 3 days at 80°C. In Fig. 4 the chemo-mechanical correlation of the three binders is displayed. The results show that the values for |G*| and AI FTIR are very close to the STA state, which indicates that no significant aging took place, regardless of the level of NO in the aging cell. For all three binders the DSR and FTIR measurements displayed no distinctive aging trend after aging with NO, indicating no dependency on the bitumen source or composition. The scattering between the data points for the different NO concentrations is in the expected range and can be explained by naturally occurring variations between experiments, e.g. inhomogeneity of the bitumen sample and fluctuations in gas flow or temperature. In principal, NO plays an important role in atmospheric photochemical processes including decomposition of ozone and the tropospheric cycle of OH radicals [44,45], however, NO is not an oxidizing agent and other reactants and also energy input through radiation are required for these reaction cycles. Therefore, it is assumed that higher nitrogen oxides or other strong oxidants have to be present for a reaction with bitumen to take place.

NO 2 in N 2
The chemo-mechanical correlation for the three binders after 3 days in N 2 ? NO 2 is shown in Fig. 5. By using NO 2 instead of NO oxidation occurs, causing the |G*| and AI FTIR to rise. When looking at the results, it can be seen that the complex modulus is increasing proportionally with the NO 2 concentration for all three binders. Binder A and B show very similar trends with an average gap of 7 kPa between the different concentrations and binder C shows a slightly stronger increase of |G*| with an average gap of 14 kPa. The highest value for the complex modulus is obtained after 3 days in N 2 ? 600 ppm NO 2 , surpassing the PAV aged sample by approximately 5 kPa for binder A, 7 kPa for binder B and 20 kPa for binder C. Based on the more distinct increase in |G*| and the considerably higher aging level after 600 ppm NO 2 compared to after PAV aging it can be assumed that binder C might be more susceptible to oxidation induced by NO 2 in N 2 .
However, when looking at the aging index the values partly deviate from the expected trend that a higher concentration of NO 2 leads to a higher aging level. In Fig. 6 it can be seen that for binder A the AI FTIR is increasing from 0.363 for 50 ppm , to 0.378 for 100 ppm and to 0.380 for 200 ppm NO 2 . However, for 400 and 600 ppm the AI FTIR is decreasing to 0.376 The aging index, which is used for the chemomechanical correlation, is calculated according to Eq. 1. When looking at the FTIR spectra of binder A (see Fig. 6), it can be seen that the sulfoxide band does not increase proportionally with the NO 2 concentration, but decreases, which results in lower aging indices. However, in the range of 1350-1100 cm -1 the absorption of the bitumen samples follows the expected trend by showing the highest absorption for 600 ppm NO 2 and the lowest for 50 ppm NO 2 . The decreasing sulfoxide band could be explained by subsequent reactions to sulfones or decomposition of the sulfoxides. Therefore, it is possible that the bitumen sample after 3 days in N 2 ? 600 ppm NO 2 did incorporate the highest amount of oxygen during the aging process, but this is not properly detected by the current definition of the FTIR aging index, as it merely includes carbonyls and sulfoxides. Unfortunately, this problem has not yet been solved, since it is not clear what bands result from the formation of specific chemical compounds and what bands increase simply because of a general rise in polarity in the material.
The FTIR spectra and aging index of binder B are depicted in Fig. 7. Contrary to binder A, the sulfoxide band of binder B between 1079 and 984 cm -1 increases with the NO 2 concentration, showing the lowest intensity for 50 ppm NO 2 (in blue) and the highest for 600 ppm NO 2 (in turquoise). Additionally, the area from 1350 to 1100 cm -1 follows the same trend, whereby the bands of the different concentrations can be seen even more clearly here. The spectrum after 50 ppm NO 2 at the bottom in blue, then 100 ppm NO 2 in pink, 200 ppm NO 2 in orange,  400 ppm NO 2 in brown and 600 ppm NO 2 in turquoise on the top. The increasing sulfoxide band in combination with the increasing carbonyl band results in a steadily rising aging index from 0.336 for 0 ppm to 0.371 for 600 ppm NO 2 , as opposed to the decreasing aging index of binder A. This could be explained by the fact, that the sulfoxides formed in binder B are not as prone to decomposition or additional reactions as the sulfoxides in binder A and can therefore be detected with ATR-FTIR at the end of the aging experiment. The reason for that could be the different crude oil source of binder A and B, which results in a different chemical composition and therefore in a chemical environment for the sulfoxides, where they might be more stable or the possibility for further reactions is not given. Another interesting observation in Fig. 7 is the very steep increase of AI FTIR from 0 to 100 ppm NO 2 , which then transitions into a less distinct incline for 100 to 600 ppm. Two hypotheses have been established for the explanation of this phenomenon: Firstly, it is possible that the aging process is accelerated immensely by the addition of NO 2 , though minor concentrations of NO 2 are sufficient and higher amounts only have a slight effect. Secondly, a similar problem as observed for binder A might occur, whereby higher concentrations of NO 2 could indeed cause more aging, however this is concealed by the current definition of AI FTIR.
In Fig. 8 the FTIR spectra and aging index of binder C can be seen. The bands follow the same trend that was observed for binder B (see Fig. 7). The sulfoxide band between 1079 and 984 cm -1 and the area between 1350 and 1100 cm -1 show the highest intensity for the highest NO 2 concentration, which leads to an increase in the aging index from 0.352 to 0.374 from 50 to 600 ppm of NO 2 . This indicates that binder B and C might have a more similar chemical composition compared to binder A, which causes the sulfoxides to be more stable and/or less reactive. On the right hand-side of Fig. 8 the AI FTIR is depicted as a function of the NO 2 concentration. The same steep rise of the aging index followed by a weaker increase, as was seen for binder B, can be observed here, although the flattening of the slope is slightly less distinctive for binder C. This could be explained by the fact, that binder C has to some extent a higher aging susceptibly under these aging conditions, resulting in a more pronounced effect at higher concentrations of NO 2 .

NO 2 in air
The chemo-mechanical correlation for the binders after 3 days in air ? NO 2 is shown in Fig. 9. The addition of NO 2 to air leads to heavy oxidation, causing a very high complex modulus and aging index in all the samples. When looking at the results for binder A on the left hand-side an almost linear increase in |G*| with AI FTIR can be observed. The addition of 50 ppm NO 2 to the aging atmosphere results in an aging index of 0.439 and a |G*| of 119 kPa, whereby the PAV aged level is surpassed by 0.035 and 51 kPa, respectively. From 50 to 100, from 100 to 200 and from 200 to 400 ppm of NO 2 the AI FTIR increases on average by 0.005 and the |G*| by 30 kPa. The most significant increase is observed for the step from 400 to 600 ppm of NO 2 , whereby the AI FTIR goes up by 0.018 and the |G*| by 124 kPa. From the STA state (after PreVBA) to the LTA state achieved after 3 days in air ? 600 ppm NO 2 the DAI FTIR is 0.131 and D|G*| is 306 kPa.
For binder B the chemo-mechanical correlation also shows a linear trend, with the largest gap between 400 and 600 ppm. However, the increase in aging index and complex modulus is with 0.008 and 42 kPa significantly lower than for binder A. Moreover, the data points for 50, 100 and 200 ppm of NO 2 are especially close together, with the results for 50 ppm even surpassing 100 and 200 ppm of NO 2 . This could be explained by binder B having a lower aging susceptibility for NO 2 concentrations up to 200 ppm, where in that range the exact concentration has no influence on the aging level. However, the more likely explanation is that the sample underwent additional aging during storage, processing or analysis in the course of the aging experiment with 50 ppm NO 2 and that this sample shows therefore an unexpectedly high AI FTIR and |G*|. From the STA state (after PreVBA) to the LTA state achieved after 3 days in air ? 600 ppm NO 2 the DAI FTIR is 0.125 and D|G*| is 195 kPa. When comparing these results to binder A the increase of the aging index is very similar, however for the |G*| binder A shows a significantly higher increase.
When looking at the chemo-mechanical correlation for binder C, it shows a linear correlation between the aging index and the complex modulus, however the largest gap with DAI FTIR = 0.015 and D|G*|= 58 kPa occurs from 50 to 100 ppm. The impact of the increase in NO 2 concentration is becoming less significant for the higher concentrations, especially for 200, 400 and 600 ppm with an average DAI FTIR of 0.005 and an average D|G*| of 35 kPa. Most notably is the very minor increase in the aging index, indicating a smaller amount of incorporated oxygen or that the incorporated oxygen could not be as well detected for binder C as for binder A and B. From the STA state (after PreVBA) to the LTA state achieved after 3 days in air ? 600 ppm NO 2 the DAI FTIR is 0.134 and D|G*| is 288 kPa, whereby the increase in the aging index is very similar to the other two binders and the increase in the complex modulus is very similar to the increase observed for binder A. Although the three investigated binders show some differences, it can be said that overall the influence of NO 2 in air on the aging process is the same. The added NO 2 causes a drastic increase in the occurring oxidation, therefore inducing the formation of ketones and sulfoxides. Furthermore, the complex modulus is increasing significantly making the sample stiffer and more brittle after the aging process. As of now, it is assumed that NO 2 acts as a ''door opener'' for the oxidation process in this setting, enabling increased oxidation reactions with the O 2 from the air [46]. It is possible that NO 2 decreases the activation energy for a reaction of O 2 with bitumen or that elevated oxygen states are formed through reactions of NO 2 with O 2 . These elevated oxygen states hold a higher energy and an oxidation reaction with bitumen is therefore facilitated.

O 3 in Ar ? O 2
The chemo-mechanical correlation for the binders after 3 days in 21% O 2 ? 79% Ar ? O 3 is displayed in Fig. 10. As already mentioned the reason for using a mixture of O 2 and Ar was firstly to allow the production of O 3 without the presence of NO x and secondly to keep the overall oxygen content in the atmosphere approximately at the same level as in air. When looking at the chemo-mechanical correlation of binder A on the left hand-side, it can be seen that with an average AI FTIR of 0.413 and an average |G*| of 69 kPa all experiments led to aging levels close to the PAV with AI FTIR = 0.404 and |G*|= 68 kPa. The data point for 2 g/m 3 of O 3 is shifted somewhat to a higher aging index and complex modulus, however minor deviations can be explained by additional aging that occurred during storage, processing or analysis. For binder A the exposure to higher O 3 concentrations did not result in a significant increase or decrease AI FTIR or |G*|, indicating that the O 3 concentration does not have a strong influence on the aging process.
Very similar results can be seen when looking at the chemo-mechanical correlation for binder B in the middle of Fig. 10. With an average AI FTIR of 0.409 and an average |G*| of 70 kPa the aging experiments with O 3 led approximately to the PAV aging level of AI FTIR = 0.405 and |G*|= 64 kPa. The scattering between the data points of different O 3 concentrations is also in the expected range, suggesting that the addition of more O 3 does not cause more aging to occur.
However, the chemo-mechanical correlation of binder C shows a different trend. The experiment with 2 g/m 3 of O 3 induced approximately the same amount of oxidation and stiffening as the PAV method, but contrary to binder A and B higher concentrations of O 3 did cause a rise in the aging index and complex modulus by an average of 0.012 and 15 kPa. This indicates that binder C is more susceptible for oxidation reactions with O 3 , which could be caused by differences in the chemical composition due to the specific crude oil source or production process.
A more detailed depiction of the influence of O 3 on the aging process can be obtained from Fig. 11, where the correlation between the AI FTIR and the O 3 concentration (left) and between |G*| and the O 3 concentration (right) is depicted.
For binder A and B, the addition of O 3 to the atmosphere in the aging cell does not cause a higher aging index or complex modulus. A possible explanation is that due to the high oxidation potential of O 3 the surface of the bitumen sample immediately oxidizes and thus, a passivation layer is formed. Subsequently, O 3 does not induce any oxidation of the bulk, leading to a constant AI FTIR and |G*|. A different picture is given when looking at the results for binder C, which show a consistent rise of AI FTIR and |G*|. A reason for this difference in the aging behaviour could be the previously mentioned variations in the chemical composition between the binders. An organic compound with a high susceptibility for an oxidation reaction with O 3 could be responsible for the formation of said passivation layer. A possibility for that would be chemical structures containing unsaturated hydrocarbons, whereby an ozonolysis of the double bond takes place. Thereby, a ozonide is formed as an intermediate, which then further reacts to an aldehyde, a ketone or also to a carboxylic acid (see Fig. 12) [47].
If binder A and B show a significant higher concentration of e.g. unsaturated hydrocarbons, it is possible that the formed passivation layer on the surface of binder A and B is significantly thicker than on the surface of binder C. That could subsequently explain why binder C shows an increase in AI FTIR and |G*| with the O 3 concentration while binder A and B do not show this dependency. However, to determine this kind of information the application of other more suitable chemical analysis methods, e.g. nuclear magnetic resonance spectroscopy (NMR) or also mass spectrometry, would be necessary.

Comparison of different atmospheres
In the previous chapters the impact of selected ROS on bitumen aging was discussed in detail. The comparison of these different ROS and the associated aging atmospheres is displayed in Fig. 13, where very similar trends for all three binders can be observed. The aging index and the complex modulus after 3 days in N 2 ? 600 ppm NO are in close proximity to the values after STA with PreVBA. This indicates that NO as the only oxygen source induces no aging, neither by a formation of carbonyls or sulfoxides, nor in the form of an increase in stiffness. When looking at the results for 3 days in N 2 ? 600 ppm NO 2 and 3 days in 21% O 2 ? 79% Ar ? 10 g/m 3 of O 3 it can be seen that although both experiments lead to approximately the same |G*| for all three binders, the AI FTIR shows different values. For binder A the DAI FTIR for the aging experiment with NO 2 and with O 3 is 0.034, for binder B it is 0.041 and for binder C it is 0.071. This means that in these cases an increase of the AI FTIR and  [47] therefore an increase of carbonyls and sulfoxides in the bitumen sample does not go hand in hand with an increase in stiffness. A possible explanation is that not only carbonyls and sulfoxides are formed but also other changes in the chemical composition occur, which compensate the increasing polarity gradient and therefore the material does not become stiffer. The increase of AI FTIR compared to the aging experiments with NO 2 in N 2 is believed to be caused by the additional O 2 in the aging cell. While all three binders showed significant aging in N 2 ? 600 ppm of NO 2 , the biggest impact was caused by the addition of 600 ppm of NO 2 to air, with an average DAI FTIR of 0.096 and an average D|G*| of 207 kPa in comparison to the same concentration of NO 2 in N 2 . The results show that NO 2 does induce a certain amount of aging when being the sole oxygen source, but its full oxidation potential is unfolded with an additional oxygen source present.

Conclusion
Understanding the aging processes in bitumen is a crucial aspect in making bituminous construction materials more durable, enabling successful recycling and prolong its overall service life time. Therefore, aging methods that are as close to reality as possible, need to be considered. Aging inducing factors that are not being taken into consideration in standard LTA methods, are humidity, light and ROS, whereby especially ROS have been greatly disregarded. Therefore, the effect of NO, NO 2 and O 3 on three unmodified bituminous binders has been investigated in this study. This was achieved by adapting the VBA method to allow separate exposure to NO, NO 2 and O 3 in different atmospheres and analysing the respective bitumen samples with ATR-FTIR spectroscopy and with the DSR.
The results show that while NO in N 2 does not induce any significant aging, NO 2 in N 2 causes oxidation reactions to occur, resulting in an increase in AI FTIR and |G*|. This means that the oxidation potential of NO 2 is sufficiently high to cause aging in bituminous binders, even with NO 2 as the sole oxygen source in the system. The addition of O 3 to the atmosphere did not have any effect on the aging process of binder A and B, but for binder C a dependency of AI FTIR and |G*| on the O 3 concentration was seen. A possible explanation for this observation could be differences in the chemical composition and therefore in the aging susceptibility regarding different ROS of the three binders. Furthermore, the comparison of the different aging atmospheres showed that NO 2 in N 2 and O 3 in Ar ? O 2 induced an approximately equal rise in the |G*| for all three binders. However, the experiments with O 3 resulted in a higher AI FTIR , which is assumed to be caused by the additional O 2 in the aging cell.
The most significant impact was observed for the addition of NO 2 to air, whereby the AI FTIR and |G*| of all other aging experiments were greatly surpassed.
The results indicate that in combination with an additional oxygen source NO 2 acts as an oxidation intensifier or accelerator. For a more detailed analysis of the underlying mechanisms and to find more comprehensive explanations for the specific aging behaviour of the binders, additional chemical analysis methods, e.g. NMR or fluorescence spectrometry, will be used in the future. Especially, the connection of a SARA separation of the aged bitumen samples and a subsequent examination of the different fractions with the previously stated analytical methods could provide a deeper insight.
Moreover, the combination of ROS in different ratios could lead to more severe aging and could offer more information about how bituminous products age during their service life. The VBA aging method could be further modified to include moisture and light, thereby providing a realistic aging setup for not only bituminous materials, but also for any other materials exposed to atmospheric aging and weathering. Certain questions about the aging of bitumen are still unanswered, but this study provides evidence that ROS play an important role and should not be disregarded when looking at the aging of bituminous materials.
Funding Open access funding provided by TU Wien (TUW). The financial support by the Austrian Federal Ministry for Digital and Economic Affairs, the National Foundation for Research, Technology and Development and the Christian Doppler Research Association is gratefully acknowledged. Furthermore, the authors would also like to express their gratitude to the CD laboratories company partners BMI Group, OMV Downstream and Pittel ? Brausewetter for their financial support.

Declarations
Conflict of interest No potential conflict of interest was reported by the authors.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.