Multiscale mechanisms of asphalt performance enhancement by crumbed waste tire rubber: insight from molecular dynamics simulation

The recycling of waste tires is a major environmental problem facing mankind, and the addition of crumbed waste tire rubber (CWTB) to the base asphalt is an extremely promising recycling method. However, the modification mechanism of CWTB to asphalt is not well understood, which restricts the development of CWTB modified asphalt. In this study, the mechanism of CWTB modification of asphalt was explored by dynamic mechanical analysis (DMA), fluorescence microscopy, scanning electron microscopy (SEM), Fourier transform infrared spectroscopy (FTIR), and molecular dynamic (MD) simulations. After verifying the asphalt reasonableness using glass transition temperature, CWTB modified asphalt was simulated and experimented. The results showed that CWTB enhanced the high temperature performance of the base asphalt. The microscopic mechanism by which this phenomenon occurs is that CWTB has the largest binding energy with the aromatics (1150–1350 kcal/mol), followed by the saturates (740–830 kcal/mol), followed by the resins (90–330 kcal/mol), and the smallest binding energy with the asphaltenes (100–140 kcal/mol), which causes CWTB to absorb the light components of the asphalt (aromatics and saturates). In addition, the introduction of CWTB reduces the diffusion coefficient of asphalt. In this process, CWTB will gradually swell, which causes CWTB to bind more and more tightly with the base asphalt, and eventually the good high temperature performance of CWTB is transferred to the base asphalt. The macroscopic manifestation of this process is that the rutting factor of CWTB-modified asphalt is significantly higher than that of virgin asphalt. This study can provide basic theoretical support for the application of CWTB-modified asphalt.


Introduction
The number of cars worldwide has exceeded 1 billion, and these vehicles generate more than 2.5 billion scrap tires per year [1,2]. Worryingly, the number of waste tires is still growing at a rate of 2-3% per year [3,4]. Therefore, the recycling of waste tires has attracted worldwide attention. At present, the common methods of reuse of waste tires are mainly burning power generation and underground landfill. However, both of these methods have the disadvantages of serious secondary pollution and long metabolic cycles. If rubber is used as a modifier for asphalt, it can significantly improve the low temperature crack resistance of asphalt [5,6]. For asphalt mixtures, the addition of rubber powder can improve the compaction and improve the damage resistance and mechanical properties of the mixtures. In addition, rubbermodified asphalt mixtures can also significantly reduce noise [7,8]. Therefore, using crumbed waste tire rubber (CWTB) to modify asphalt cannot only improve pavement performance but also realize the great recycling of waste tires.
Currently, researchers generally agree that the change in the properties of asphalt originates from the interaction between CWTB and the base asphalt [9] and CWTB was found to have two phases of swelling and dissolution in asphalt [10]. Swelling is the phenomenon of volume increase of CWTB, while dissolution is a phenomenon in which the CWTB is desulfurized and depolymerized at high temperatures [11,12]. Both the dissolution and swelling stages have a great influence on the road performance of CWTB-modified asphalt. Therefore, the literature [3,13] qualitatively investigated the interaction between various rubbers and asphalt. However, the mechanism of rubber modification of asphalt is still not obtained although researchers have conducted many studies on CWTB-modified asphalt, the mechanism of CWTB modified asphalt has not been well explained. The main reason is that most previous studies have used macroscopic performance experiments to evaluate the effect of modifiers on asphalt properties, or quantitative characterization of the microscopic state of modified asphalt using techniques such as scanning electron microscopy (SEM) or fluorescence microscopy. However, these methods can only lead to macroscopic, empirical conclusions and cannot fundamentally explain the generation mechanisms of macroscopic properties and phenomena at the atomic scale. Therefore, there is a need to use new methods that can quantitatively characterize the microscopic scale.
Molecular dynamic (MD) simulation is an excellent tool for interpreting material properties from the molecular scale and has attracted strong interest from asphalt researchers [14]. In recent years, the use of MD simulation has spread in all areas of asphalt. The application of MD simulations in interface problems originated from Guo Meng, who pioneered the development and validation of an interface model between asphalt and aggregate, which has profoundly influenced subsequent research [15][16][17]. The self-healing model usually consists of two asphalt layers with a vacuum zone in between to allow molecular movement [18] and to study the changes in the movement of asphalt molecules under a particular condition [19]. Although MD simulations cannot reproduce the dynamic processes of chemical reactions, the construction of molecular models of asphalt before and after aging allows the evaluation of the effects of aging on the microstate of asphalt [20][21][22]. The optimization of asphalt molecular models is also a very important topic, which often needs to be explored by professional asphalt chemists. A twelve-component molecular model has been proposed in the literature [23], and this model has been widely used. The literature [24] proposed an alternative fourcomponent model based on real asphalt molecules, making an exploration for finding a suitable model from real molecules. In order to achieve the connection between the microstructure and macroscopic properties of asphalt, a larger scale model of asphalt is required; therefore, a mesoscale model of asphalt was proposed in the literature [25], and the rationality of the model was verified using experiments. In addition, the time duration of MD simulations is always short and far from the actual asphalt service state. The literature [26] significantly reduced the time difference between molecular simulations and experiments using time-temperature superposition, which improved the credibility of molecular simulations. MD simulations have many other applications, such as physicochemical properties [27][28][29], modifier-asphalt matrix interaction [13,30], and glass transition temperature [19,31]. In the near future, MD simulations will be applied to other problems.
The objective of this study was to explore the multi-scale mechanism of CWTB modification of asphalt using MD simulations. CWTB-modified asphalt samples were prepared and subjected to dynamic mechanical analysis (DMA). The twophase structure of CWTB-modified asphalt at different developmental stages was observed by fluorescence microscopy, the microscopic morphology of CWTB modified asphalt at different developmental stages was observed by scanning electron microscopy, and the chemical mechanism of CWTB modification of asphalt was analyzed by Fourier transform infrared spectroscopy (FTIR). Finally, natural asphalt was chosen as a representative of CWTB, and the nature of the interaction between CWTB and various molecules of asphalt was explored quantitatively using MD simulations.

Construction of CWTB-modified asphalt model
In this study, natural rubber was selected as a representative of CWTB. However, natural rubber is a polymer made of cis-1,4-isoprene as monomer polymerization, and its molecular weight varies with the degree of polymerization, and the computational cost of MD simulation is proportional to the molecular weight of the system [13,32]. Therefore, a natural rubber model with suitable polymerization degree needs to be selected for molecular dynamic simulation. The method used in this study is to calculate the solubility parameter of natural rubber with different polymerization degrees and take the sudden change in the slope of the solubility parameter curve as the model needed for the simulation [33]. The solubility parameter can be calculated by the following equation.
where δ stands for the solubility parameter, CED stands for the cohesion energy density, E coh stands for the cohesion energy, and V stands for the system volume. Amorphous cells of natural rubber with different degrees of polymerization were constructed in Materials Studio (MS), they were all geometrically optimized in 500,000 steps, and finally the solubility parameters were calculated for different systems. The above processes were performed in the Forcite package of the MS software. The solubility parameters of different natural rubbers are shown in Fig.1, it can be found that the solubility parameters first increase and then basically equilibrate as the degree of polymerization increases, and the critical point polymerization degree is 12. Therefore, the natural rubber used in this study was with a polymerization degree of 12.
The asphalt molecular model used in this study is proposed in the literature [24], where 12 molecules are used to represent the four components of the asphalt, as shown in Fig.2. Before constructing the molecular model of asphalt, the content of the four components in the model must be determined, which requires testing asphalt samples. The virgin asphalt sample used in this study is GS-90, whose four-component content is shown in Fig.3b, and the measurement method is ASTM D4124-09. Based on this, the number of various asphalt molecules (Fig. 3a) is calculated so that the model four-component content is consistent with the experimental value (Fig. 3b), and finally the asphalt model is constructed in the amorphous cell (AC) module of MS software. The number of molecules of the asphalt model in this study is shown in Fig.3a. It can be found that the content of each component of the asphalt sample is highly consistent with that in the molecular model, which proves the correctness of the asphalt model. The molecular model of CWTB-modified asphalt was further constructed, and the construction method was as described previously. The virgin asphalt model and CWTBmodified asphalt models in this study are shown in Fig.4. It must be mentioned that the content of natural rubber in the CWTB modified asphalt model is 5%, which is to make the model consistent with the experiment.
A geometry optimization operation with a step number of 500,000 was performed for both the virgin asphalt model and the CWTB-modified asphalt model, with the aim of achieving an energy-minimized system. Next, a dynamic simulation at NVE of 50 ps was performed for each model to allow the model to reach a steady state. Finally, dynamic simulations were performed for each asphalt model at 100 ps, under NPT, and the last frame and trajectory file were taken to calculate the micro parameters.
The construction of the model in this study is implemented in the amorphous cell module of MS software, while the energy minimization and molecular dynamic simulations are proposed to be performed in the Forcite module of MS software, which is the most commonly used module for molecular dynamics.
In this study, the MD simulation temperature was set to 298.13 K and the pressure was set to 1 atm, so as to simulate the system variation of modified asphalt at room temperature and pressure. Quality was adopted as ultra-fine, which is to ensure the accuracy of MD simulation. The energy deviation is set to 40,000 kcal/mol to control the energy change of the simulation. Velocities and forces are included for the kinetic simulation trajectory, which is to facilitate our analysis of the various properties of the system.
During the simulation, nose was used for Thermostat, berendsen was used for Barostat, decay constant was set to 0.1 ps, time step was set to 1 fs, number of steps was set to 100,000, frame output every was set to 5000 steps, electrostatic was calculated by Ewald method, and van der Waals was calculated by Atom based method. In addition, the force field used in this study is the COMPASS force field, which is the preferred force field in the field of materials and has been adopted by many asphalt research institutes with good results [16,[34][35][36][37]. The functional expression of the COMPASS force field is shown in Eq (3).  where b stands for the bond, θ stands for the angle, ф stands for the torsion, K is the parameter, and OOPA stands for the out of plane angle. In addition, i, j stands for different atoms, ε stands for the potential well depth, r stands for the action distance of different atoms, and q stands for the atomic charge. Detailed information on the function can be obtained from the literature [38,39]. The trends of energy, amorphous cell edge length, density, and temperature of the CWTB-modified asphalt system during the MD simulation are shown in Fig.5. It can be clearly seen that the energy and temperature of the CWTB-modified asphalt system have reached the equilibrium after 15 ps. And after 40 ps, the system edge length and density also tend to equilibrium. That is, after 40 ps, the modified bitumen system has reached the kinetic equilibrium state, where the model and trajectory files can be used to calculate the microscopic characterization parameters.
In order to verify the soundness of the molecular model, the glass transition temperature of the native asphalt molecules was calculated. The glass transition temperature is an important property of polymeric materials and has a decisive influence on the properties of the material [40]. The method used was to calculate the density variation of the model at different temperatures (150, 200, 225, 250, 275, 298, 323, and 348 K), to obtain the variation of the model specific volume versus temperature, and finally to find the glass transition temperature of the model [19,40].
The density variation curves of the virgin asphalt model at various temperatures are shown in Fig. 6a. It can be found that regardless of the temperature, the system tends to equilibrate when the simulation proceeds to 40 ps and the density of the asphalt system gradually decreases as the temperature increases, a phenomenon consistent with previous studies [19]. The average value of the density from 70 to 100 ps was taken to calculate the specific volume, and the glass transition temperature was obtained using a linear regression method [41], and the results are shown in Fig. 6b. It can be found that the glass transition temperature of asphalt is lower than 273.15 K (0°C) and is 242.74 K, which is consistent with the results of previous studies [41,42]. Therefore, the asphalt model in this study is reasonable, and the simulation results obtained are highly credible.
The indicator used in this study to characterize the stability of the interaction is the binding energy, which is currently one of the most commonly used and most effective microscopic characterization quantities. There are many definitions of binding energy, but the most accepted one is the amount of energy or work required to separate two systems. The binding energy can be used to quantify the strength of the bond between two substances. For modified asphalt systems, the state of the bond between the modifier and the asphalt has a direct effect on the performance. The binding energy can be calculated by Eqs. (3) and (4).
where E bond stands for the bonding energy, E lrc stands for the long range correction, E coulomb stands for the electrostatic energy, E edW stands for the vdW energy, and E non-bond stands for the non-bonding energy. Characterization of the effect of natural rubber on the molecular movement of asphalt using diffusion coefficients. According to Einstein's diffusion law, the diffusion coefficient of molecules can be calculated using Eq. (6). It has to be mentioned that for a system like asphalt, it cannot be calculated using the mean square displacement (MSD) data of the simulated full process but only the first small part can be taken and found in the log function. In this study, the MSD-time data of the first 30 ps were taken to calculate the molecular diffusion coefficient.  where r(t) stands for the position vector of the molecule, t stands for the time, and MSD stands for the mean square displacement.

Materials and experiment
The asphalt used in this study was GS-90 provided by Sinopec, whose technical parameters are shown in Table 1. CWTB of 30 mesh was prepared from waste tires of Henan Huihua Waste Tire Recycling and Treatment Co. Since the CWTB are very hard and have a high melting point, the CWTB cannot be modified to be directly ground into powder and added into the base asphalt. In this study, the CWTB were cracked first, and the amine cracking agent was used, and the rubber powder was made at the end. The technical parameters of the prepared CWTB are shown in Table 2, and its particle size distribution is shown in Fig.7. The rubber powder obtained from the cracking was mixed with the base asphalt in a high-speed shear in a certain ratio and sheared at 185°C for 3 h at a speed of 5000 rpm, with the aim of allowing the rubber powder to fully interact with the base asphalt to obtain a stable CWTB-modified asphalt. In this study, FTIR tests, SEM tests, fluorescence microscopy tests, and DMA were performed successively on virgin and CWTB-modified asphalts with the aim of exploring the multiscale modification mechanism of CWTB-modified asphalt in combination with MD simulations. The SEM uses the TM3030PLUS™. FTIR by Bruker Spectroscopy, Germany, model VERTEX 70™. The model number of fluorescence microscope is BSM-600E™, the optical magnification is 40X-2000X, provided by Shanghai BIMU Instruments Co., and the DSR model is Bohlin GEMINI™ by Xi'an Minx Testing Equipment Co., Ltd., and the test standard is AASHTO T 315. The technical route of this study is shown in Fig.8.

Results and discussion
Rutting resistance of CWTB-modified asphalt and virgin asphalt The rutting factors of CWTB-modified asphalt and base asphalt are shown in Fig. 9. It can be found that the rutting factor of both base asphalt and CWTB-modified asphalt decreases significantly with the increase of temperature. This phenomenon is a fundamental property of asphalt materials, also known as temperature sensitivity, and is responsible for the occurrence of high temperature rutting distress.
The rutting factor of the CWTB modified asphalt is greater than that of the virgin asphalt regardless of the temperature, which indicates that CWTB have an enhanced effect on the high temperature performance of the asphalt. This is because CWTB absorbs some of the aromatics and saturates in the base asphalt, causing the relative asphaltene and resin content of the base asphalt to rise. At the same time, CWTB itself has good high temperature performance, so the rutting factor of CWTB-modified asphalt is higher than that of the base asphalt.

Two-phase structural state of CWTB-modified asphalt
The microscopic images of CWTB-modified asphalt at different developmental stages are shown in Fig. 10. It can be found  Fig. 7 Particle size distribution of CWTB that the CWTB-modified asphalt is divided into obvious black and red two-phase structures, where black is the CWTB phase and red is the asphalt phase. This phenomenon implies that CWTB is poorly compatible with the matrix asphalt, which means that CWTB is easily separated from the asphalt. In addition, the area of CWTB phase in the modified asphalt increases significantly and the volume of waste tire floc becomes larger as the development time increases. This supports the conclusion that CWTB absorbs light components, and the reason for the larger CWTB phase with longer development time is that more light components are adsorbed by CWTB as the development time grows, which often predicts better road performance.

Micro-interface morphology of CWTB-modified asphalt
The SEM morphology of CWTB-modified asphalt in different development stages is shown in Fig. 11. It can be found that CWTB is not dissolved in the base asphalt, which indicates that CWTB is not completely compatible with the base asphalt, and this result is consistent with the fluorescence microscopy observation. However, the boundary between CWTB and asphalt is not clear, which indicates that CWTB will interact with the base asphalt.
As the development time increased, the volume of CWTB increased significantly, which is the result of the association of CWTB with the light components of the asphalt. In addition, it has to be mentioned that the surface of CWTB has become smoother when the development time is 4 h, which will be the compatibility of CWTB with the base asphalt. This phenomenon predicts that the development time of CWTB should not be too long, because it will increase the tendency of phase separation of CWTB modified asphalt.

Effect of waste tires on functional groups of base asphalt
The FTIR results of the base asphalt and CWTB-modified asphalt are shown in Fig. 12. It can be clearly found that the fluctuations of FTIR curves of base asphalt and CWTBmodified asphalt are basically the same, which indicates that the functional groups of the two asphalt systems are basically the same. In other words, there is no chemical reaction between the waste tire and the base asphalt, and CWTB is mainly a physical modification of the base asphalt.
The FTIR curves of both the base asphalt and the CWTB modified asphalt have distinct peaks at wavenumbers 1300-1500 cm −1 and 2800-3000 cm −1 . The functional groups corresponding to this wave number are saturated hydrocarbon groups (CH) and methylene (CH 2 ), respectively. Observation of the wave intensity reveals that the peak of  the base asphalt is between 0.08-0.14%, while the CWTBmodified asphalt is less than 0.06%. The reason why the peak of the CWTB-modified asphalt is lower than that of the base asphalt is that the CWTB-modified asphalt will join and cross link with the light components, which causes a significant reduction in the CH and CH 2 functional groups. It is worth mentioning that the FTIR result is also consistent with the MD simulation results.

Stability of CWTB interaction with asphalt
The binding energy of CWTB with various asphalt molecules and its components (vdW) long range connection and are shown in Fig. 13. Since the absolute value of the magnitude of the binding energy characterizes the stability of intermolecular interactions, the stability of the interactions between CWTB and various molecules of asphalt is ranked as follows: NAPHPN > NADOCHN > SHopane > Ssqualane > PATIRE > PATMBO > PAPNH > PAQNP > AThiophene > APyrrole > APhenol > PABBTP. Overall, CWTB interacted most stably with aromatics, followed by saturates, then resins, and least stably with asphaltenes. The binding energy of CWTB with aromatics ranged from 1150 to 1350 kcal/mol, and the binding energy of CWTB with saturates ranged from 740 to 830 kcal/mol, while the binding energy of CWTB with resins was between 90 and 330 kcal/mol. The interaction between CWTB and asphaltenes was the least stable with a binding energy of 100-140 kcal/mol. This phenomenon suggests that CWTB binds easily to light components in asphaltene matrix. In other words, when CWTB is added to the asphalt matrix, it tends to bind with the lighter components (especially the aromatics), which is the fundamental reason for the volume expansion of CWTB in SEM images. However, since CWTB does not react chemically with asphalt (FTIR results), CWTB does not dissolve in asphalt, which is the mechanism of formation of CWTB modified asphalt two-phase structure in fluorescence microscopy.
From the composition of the binding energy, the bond energy between CWTB and all asphalt molecules is 0 kcal/mol, indicating that the interaction between CWTB and asphalt is physical, which corroborates the FTIR results. In addition, the coulomb energy contributes little to the binding energy, while the vdW energy is the vast majority of the binding energy. In other words, the main type of intermolecular interactions in the CWTB-modified asphalt system is vdW interactions rather than electrostatic interactions.

Effect of CWTB on molecular motion of asphalt
The MSD-time results in this study are shown in Fig. 14, where a linear fit was performed for each molecule, and all the coefficients of determination were greater than 0.95, which means that the diffusion coefficients obtained in this study can characterize the motion of the asphalt molecules. The results of the diffusion coefficients are shown in Fig. 15 the larger the diffusion coefficient, the faster the movement of the molecules is implied.
It can be found that the diffusion coefficients of different asphalt molecules are different, and the diffusion coefficients of different asphalt systems are also different. In the virgin asphalt system, the overall diffusion coefficients from largest to smallest are aromatics > saturates > resins > asphaltenes. Coincidentally, the diffusion coefficients of various molecules in CWTB-modified asphalt system are aromatics > saturates > resins > asphaltenes. This phenomenon indicates that the motion of asphalt molecules is significantly related to the molecular mass, the larger the molecular mass, the slower the Fig. 13 The binding energy between CWTB and various molecules of asphalt (a), vdW energy (b), long range correction energy (c), and electrostatic energy (d) molecular motion; the smaller the molecular mass, the blockier the molecular motion.
It can also be found that the diffusion coefficients of molecules in CWTB-modified asphalt ranged from 0.039 to 0.06 × 10 −8 m 2 s −1 , while the diffusion coefficients of molecules in virgin asphalt ranged from 0.088 to 0.13 × 10 −8 m 2 s −1 . In other words, the introduction of CWTB resulted in a significant decrease in the diffusion coefficient of asphalt molecules. The reason for this phenomenon is that CWTB dispersed in the asphalt matrix will hinder the movement of asphalt molecules; in addition, there is an interaction between CWTB on asphalt molecules, and this interaction will bind the movement of asphalt molecules.

Conclusion
The above conclusions were obtained by exploring the multi-scale mechanism of CWTB modification of asphalt using MD simulation, FTIR, SEM, fluorescence microscopy, and DMA. 1. The optimum degree of polymerization of natural rubber in MD simulations was 12, and the glass transition temperature of the asphalt model is 267 K. 2. CWTB will interact with the base asphalt and the interaction is enhanced with the development time. However, the surface of CWTB will become smooth, which will increase the tendency of CWTB separation from asphalt. 3. The modification mechanism of CWTB to asphalt is that CWTB has the strongest interaction with aromatics, followed by saturates, resins third, and the weakest interaction with asphaltenes, which causes CWTB to absorb part of saturates and aromatics and swell. The diffusion coefficient of asphalt molecules was significantly correlated with the molecular mass, and the introduction of CWTB reduced the diffusion coefficient of asphalt molecules. However, CWTB does not have peanut chemical reaction with asphalt, and the modified asphalt system shows an obvious two-phase structure due to the incompatibility between CWTB and the base asphalt. With the growth of development time, the interface between CWTB and base asphalt becomes more and more blurred, and various properties of CWTB are gradually transferred to the base asphalt, causing the growth of asphalt rutting factor. Data availability All data generated or analyzed during this study are included in this published article.
Code availability The calculations have been carried out using Materials Studio 2018.

Declarations.
Ethics approval and consent to participate The manuscript is prepared in compliance with the Ethics in Publishing Policy as described in the Guide for Authors. The manuscript is approved by all authors for publication.

Consent for publication
The consent for publication was obtained from all participants.