To fulfill the technical requirements for fibers in the SMA mixture, several treatments such as de-esterification, scattering, separation, and drying were performed in the production of waste cellulose diacetate fiber (Figure 1). In detail, a de-esterification solution consisting of 1:1:1 (v/v) ethanol (99.5 vol%), tetrachloroethylene, and NaOH solution (1 mol/L) was used to remove the glycerol triacetate from waste cellulose diacetate. A certain amount of waste cellulose diacetate was split into clusters by a low-speed shear device and put into the de-esterification solution with magnetic stirring for 2 h at room temperature. After the reaction, the mixtures were taken out and beaten for 10 min within a Valley beater after being washed to neutralize them. It was noted that this procedure assured the mixtures were broken up by the blades in the machine and impacted to a net-structure shape instead of small pieces. Then the mixtures were removed from the solution and dehydrated for 3 min within a tubular dehydrator. After being dried for 2 h at 80°C, the recycled cellulose diacetate fiber was stored in a sealed container for the subsequent experiments.

General properties such as diameter, density, tenacity, and elongation at break of rCDA were determined according to the relative Chinese standard, and are listed in Table 1. The surface structure of recycled cellulose diacetate fiber was analyzed by scanning electron microscopy (SEMS-520, Hitachi, Japan) as well as its morphology in asphalt mastics. Functional cellulose groups were determined by Fourier transform infrared (FTIR) spectra (IRTracer-100, Shimadzu, Japan) with the wavenumber range 4,000–400 cm-1 at room temperature. Thermal gravimetric analysis was performed with an integrated thermal analyzer (Netzsch, STA 449 F3, Germany). The heating rate was set as 10°C/min. The input gas was composed of 21% oxygen and 79% nitrogen by volume, and the gas flow rate was 100 ml/min which was high enough to sweep the gaseous products to keep the experimental gas environment constant.

Asphalt mortar containing various fibers was prepared by a high-speed shearing device. Figure 2 presents the aggregate gradation with a nominal maximum aggregate size of 13.2 mm. The content of the mineral filler was 10% by weight of the total mixture. Except specimens for the drain down and particle loss tests, similar asphalt content (6.0%) and aggregate gradation were used for preparation of all specimens in order to evaluate the effect of fibers on mixtures. The content of rCDA accounted for 0.3, 0.4, and 0.5% by weight of the asphalt mixtures. The rCDA was directly used in all test samples. The content of polyester and polyacrylonitrile was 0.3% by weight of the asphalt mixtures by reviewing the literature of SMA mixture practices. The SMA mixture was designed and prepared based on the Marshall methodology in accordance with the Standard Test Methods of Bitumen and Bituminous Mixtures for High Way Engineering (JTG E20-2011). The recycled cellulose diacetate, polyester, and polyacrylonitrile fiber are referred to as rCDA, PET, and PAN in brief. Test samples were recorded as 0.3% rCDA, 0.4% rCDA, 0.5% rCDA, 0.3% PET, and 0.3% PAN according to the content and fiber type used in the mixtures.

To evaluate the high-temperature performance of the asphalt mortar containing fibers, a dynamic shear rheological test (JTG E20 T0628) and a binder leakage stability test were conducted. The method for the latter was defined as follows. The asphalt binder was heated to 140°C, then fibers were added and mixed for 3 min at 150°C. A certain amount of asphalt mortar was put into a net-shaped basket and weighted at 130°C. The weight loss was recorded. The fibers to asphalt binder ratio was 1:10 for these two tests.

The drain down test was performed to evaluate the free asphalt binder in the mixtures according to JTG E20 T0732. The Cantabro loss particle test was carried out to assess the mechanical property of mixtures containing fiber according to JTG E20 T0733. To evaluate the pavement performance of SMA mixtures, tests for Marshall stability, rutting, low-temperature bending, moisture susceptibility, and fatigue were carried out in accordance with JTG E20 T0709, T0719, T0715, T0729, and T0739, respectively.

Figures 3A,B show the typical morphologies of the surface structure of rCDA fibers. A curly and corrugated structure with approximate rough texture which contributed to the adhesive property between fibers and asphalt could be observed from the SEM images. Compared with the fresh cellulose acetate, the irregular structures might be due to the effect of erosion by the de-esterification solution during removal of the glycerol triacetate from waste cellulose diacetate. This difference would lead to a higher asphalt absorption rate when designing and preparing the asphalt mixtures containing rCDA (Slebi-Acevedo et al., 2019). It concluded that no visible cracks, holes, and hollows in the structure were found, indicating that the morphologies of rCDA were similar to synthetic fibers. Besides, no cross-section structure was observed, which makes it very different to lignin fibers, resulting in a reduction of the absorbing and retaining ability for the asphalt binder (Li et al., 2020).

The infrared spectra for the rCDA fiber are shown in Figure 3C. A strong peak at 3,485 cm⁻¹ was classified as O–H stretching vibration and the hydrogen bond of the hydroxyl. It may be due to the merging of the atmospheric water and hydroxyl groups, and the first stage of the TG technique proved the existence of water. The peak at 2,945 cm⁻¹ was associated with the stretching vibration of the C–H bond from aliphatic CH and CH₂. The peak at 1755 cm⁻¹ was related to the stretching vibration of carbonyl C=O from carboxylate ester compounds. The peak at 1,639 cm⁻¹ was related to the C=C bond. The peaks at 1,375 cm⁻¹ and 1,047 cm⁻¹ were contributed to the bending vibration of the C–H or C–O bond from aromatic compounds. The peak at 1,236 cm⁻¹ corresponded to the stretching vibration of the C–O bond of ester linkage of the acetyl group.

The thermogravimetric (TG) and differential thermogravimetric (DTG) analysis curves of rCDA are presented in Figure 3D. Three stages can be observed from the TG curve. The first stage occurs from 89.3°C to 294.1°C with a weight loss of 14.47%, which was related to the water loss of the rCDA and the release of small molecule volatile organic matters (Lambat et al., 2019; Zhou et al., 2019; Khedkar et al., 2020; Chopra et al., 2021). The second stage whose temperature is below 376.0°C had a weight loss of 62.48%, which was related to the decomposition and degradation of fixed carbon. The final stage had a weight loss of 23.05% which was mainly due to the thermal condensation reaction in rCDA (Chaudhary et al., 2016). As seen from the DTG curve, the peak of the maximum weight loss rate of the rCDA fiber appeared in this temperature range.

The rutting and fatigue factors of the asphalt mortar with different fibers are presented in Figure 4. The rutting factor was used to evaluate the high-temperature performance of asphalt mortar. Data of complex modules and phase angle at four temperature points were collected and calculated for determination of G*/Sinδ. It was concluded that better high-temperature performance was related to a higher value of G*/Sinδ (Tao et al., 2019). In this work, PET mortar exhibited the highest rutting factor at different temperatures. The rutting factors of rCDA mortar at different temperatures were higher than that of asphalt without fibers, indicating that the addition of rCDA increased the resistance ability of asphalt at higher temperatures. It was concluded that rheological properties can be improved with the addition of fibers in the asphalt binder. This might be due to the absorption of the light components of the binder into fibers, resulting in an increase of the viscosity and reduction in the cone sink depth during the DSR test (Slebi-Acevedo et al., 2019). However, the rutting factor of asphalt mortar with rCDA was obviously lower than that of PET and PAN under similar fiber content, which might indicate that fibers with higher tensile strength, such as PET and PAN, can lead to higher viscosities and shear resistances with respect to binders with rCDA fibers (Chen and Xu, 2010). On the other hand, the rCDA exhibited a rather curly antenna tip, which limited the generation of a bond between the asphalt and the fiber and led to the propagation of cracks when compared to PET and PAN (Mohammed et al., 2018).

In order to investigate the macroscopic high-temperature stability of asphalt with fibers. The test for the stability of asphalt binder leakage at 130°C was conducted and results are listed in Table 2. Due to a higher content of fibers in the asphalt mortar, the mass loss after the test was lower than 30%. The mass loss of rCDA asphalt mortar accounted for 28.2%, which was significantly higher than that of the PET and PAN mortars. This might be due to the heterogeneous characteristic of rCDA and weak adsorptive or/and absorptive ability of asphalt components, leading to a lower cohesion bond and higher fluidity of the mortar, which consequently resulted in a higher mass loss rate. Although the rheological properties and high-temperature stability of the rCDA mortar were inferior to PET and PAN mortars. Addition of rCDA had a positive effect on the modification of the original asphalt binder. It was necessary to make sure that the effect of dosage of rCDA improved the volumetric and pavement performance of asphalt mixtures.

Bulk density, percentage of air voids in asphalt mixtures (VV), percentage of voids in mineral aggregate in asphalt mixtures (VMA), and Marshall stability were measured, and the test results are listed in Table 3. It indicated that 4.0 of VV was obtained in 0.3% PET indicated that 6.0% is the optimum asphalt content for the SMA mixture with PET. Under similar asphalt content, VV slightly increased in the order of 0.3% PET, 0.3% PAN, and 0.3% rCDA. Increasing the content of rCDA from 0.3 to 0.5%, the bulk density and VFA of specimens decreased, while VV increased. All test specimens were prepared with similar asphalt content. High specific density materials such as aggregates were substituted by fibers, resulting in a decrease of the bulk density and increase of VV and VMA. Additionally, fibers with different length, shape, and orientation had effects on the volumetric properties of mixtures. The Marshall stability of 0.3% rCDA accounted for 8.5 kN, which was lower than that of 0.3% PET and 0.3% PAN. Increasing the content of rCDA, the Marshall stability reached 10.5 kN, then dramatically decreased to 7.8 kN when the content of rCDA was 0.5%. It was noted that the stability of 0.4% rCDA was higher than that of 0.3% PAN, indicating that the rCDA had optimum content in preparation of SMA mixtures due to improvement of the ability to resist the deformation under vertical load. Increasing the content of rCDA led to an increase of VV due to excess fibers, resulting in a decrease of Marshall stability by lower resistance to deformation. In this work, 0.4% rCDA in mixtures led to a greater bearing capacity of specimens by preventing relative slippage between the joint surfaces of the aggregate and rCDA-reinforced asphalt binder (Yin and Wu, 2018). It indicated that both the higher fluidity of 0.3% rCDA (bleeding of binders) and the worse compaction characteristics of 0.5% rCDA (lacking of binders) led to the reduction of Marshall stability. However, these volumetric properties were measured under similar asphalt content, indicating the indexes of mixtures with rCDA could be improved by its optimum asphalt content. Consequently, the asphalt content would significantly increase, resulting in an increase of costs.

The drain down test results for the asphalt mixtures are listed in Table 4. It was evident that almost all the fibers provided significant stabilization of the mixtures due to the drain down of mixtures less than 0.3%, which proved all fibers were performing their function as stabilizing additives in the asphalt mixtures. The drain down value of rCDA asphalt mixtures decreased from 0.18 to 0.12% when the asphalt content was 6.0%. A high rCDA fiber to binder ratio restricted the fluidity of mixtures by combining with more asphalt binder when asphalt content was at a lower level. For comparison, mixtures with different fibers at their respective optimum asphalt content were prepared. Except for the 0.3% PET mixture, all drain down values increased due to an increase of asphalt content. The drain down of rCDA mixtures increased from 0.15 to 0.35% with increasing rCDA content from 0.3 to 0.5%. The drain down of 0.3% rCDA decreased as its optimum asphalt content was less than 6.0%. The drain down of 0.5% rCDA significantly increased, indicating that the effect of increasing asphalt content on drain down was of great importance compared to increasing rCDA content. The drain down increment of 0.4% rCDA was 26.7%, which was significantly lower than that of 0.5% rCDA, implying that higher content of rCDA might result in early damage of the asphalt mixtures. On the other hand, although the drain down of 0.3% rCDA decreased, a smaller asphalt content might influence the durable performance, such as low-temperature cracking resistance. Hence, 0.4% rCDA was moderate in the SMA mixture at the optimum asphalt content.

Table 4 presents the results concerning the particle loss of specimens subjected to Los Angeles abrasion under dry conditions. Similarly, the particle (mass) loss of rCDA specimens was higher than that of 0.3% PET and 0.3% PAN specimens at both the constant asphalt content and optimum asphalt content. When the asphalt content was 6.0%, the mass loss of rCDA specimens significantly increased from 11.5 to 15.9% with increasing fiber content from 0.3 to 0.5%. When prepared with its optimum asphalt content, the mass loss increased from 10.6 to 11.8%. This might contribute to the better impact resistance ability of specimens at the optimum asphalt content. Specimens at 6.0% asphalt content were found to be loose and less compacted, resulting in higher mass loss. On the other hand, increasing asphalt content led to lower VV of specimens, and aggregates tightly bonded with the fiber-reinforced asphalt binder when mixtures were cooling down. Hence, the particle loss of specimens decreased when it was subjected to impact.

The high-temperature stability of asphalt mixtures is usually evaluated by dynamic stability and rutting depth. The rutting test results are presented in Figure 5A. The 0.3% PET exhibited the highest dynamic stability and the smallest rut depth followed by the 0.4% rCDA, 0.3% PAN, 0.3% rCDA, and 0.5% rCDA, respectively. It is worth noting that the dynamic stability of 0.4% rCDA (4,105 cycles/mm) was slightly higher than that of 0.3% PAN (3,985 cycles/mm). The rutting depth of 0.4% rCDA accounted for 2.6 mm, which was lower than that of 0.3% PAN (2.8 mm). Although higher content of rCDA fiber led to higher VV of the Marshall specimen, the rutting test specimen was far bigger and made the fiber well-separated in mixtures for formation of stiffer mixtures, resulting in better stabilization of the asphalt binder on the specimen’s surface making it able to resist permanent deformation at a temperature of 60°C. Besides, increasing the content of rCDA fiber from 0.3 to 0.4% enhanced the skeleton structure to resist the shear force and reduce the fluidity by limiting aggregates from moving in the asphalt matrix (Chen and Xu, 2010; Yin and Wu, 2018). When the content of the rCDA fiber reached 0.5% in mixtures at the same asphalt content, the excess fiber exhibited a negative effect on the resistance of deformation due to weak bonding between the asphalt binder and aggregates. By the effect of wheel track, the dynamic stability dramatically decreased with an increase of rutting depth. However, it is necessary to understand the effect of rCDA fiber on the anti-deformation ability of mixtures by adjusting the asphalt content and aggregate gradation in further investigations.

In this work, low-temperature cracking resistance of the fiber-reinforced asphalt mixture was evaluated according to a three-point flexural test at the temperature of 10°C below zero. The ultimate flexural strength and flexural stiffness modulus are presented in Figure 5B. The changing rule of ability to low-temperature cracking resistance of SMA mixtures with different fibers were like its high-temperature deformation resistance ability. The ultimate flexural strength and flexural stiffness modulus of 0.4% rCDA accounted for 3,722 MPa and 9.7 MPa, which were slightly higher than that of 0.3% PAN. It was concluded that the improvement of low-temperature cracking resistance of fiber-reinforced asphalt mixtures depended on the retarding effect of fibers on the crack propagation in the asphalt matrix. Figure 1 shows that the shape of the rCDA fiber was like a net structure. It was found that the waste cellulose diacetate was extracted by regeneration solvent and recycled as separate fibers, instead of “thinning” the cigarette filter. The rCDA fibers still consisted of tows which enhanced the tensile strength due to combination of lots of single fibers in a small area. This might contribute to the improvement of ability in low-temperature cracking resistance of rCDA mixtures. Hence, the crack propagation was gradually weakened and terminated until the expansion energy was depleted due to cumulated fracture energy of net-structure and bonded rCDA fiber. When the content of rCDA was 0.5%, it led to damage on the surface of specimen due to the excess unbonded fibers. The crack propagation along this damage continuously expanded, resulting in the lower ultimate flexural strength and flexural stiffness modulus of specimens.

Freezing-thaw splitting and the immersion Marshall test were performed to evaluate the moisture sensibility of SMA mixtures. The results of tensile strength ratio (TSR) and residual Marshall stability (RS) are presented in Figure 5C. The value of TSR and RS of 0.3% PET reached 85.2 and 83.1%, respectively. Compared with 0.3% PAN, 0.4% rCDA slightly decreased the TSR and RS from 81.4 to 80.2%–78.1 and 75.3%, respectively. Concurrently, the TSR and RS of 0.3% rCDA and 0.5% rCDA were below 75%, indicating that all rCDA mixtures exhibited a worse anti-moisture damage ability. It is known that loss of cohesion of the asphalt film and the failure of the adhesion between the aggregate and asphalt binder are the main mechanisms to interpret the moisture damage of asphalt mixtures (Luo et al., 2019). In this work, the rCDA fiber bonded with the sticky binder on its surface and decreased the asphalt film thickness around the aggregate. Moisture easily penetrated rCDA asphalt mixtures through the cracks and stripped particles on the specimen surface, which led to most of the distress and the failure of mixtures when rCDA asphalt mixtures were subjected to freezing-thaw or moisture immersion conditions.

The fatigue life of fiber-reinforced asphalt mixtures was evaluated by a three-point bending beam fatigue test at a temperature of 15°C with a loading frequency of 10 Hz. The fatigue life under stress control mode was calculated by Eq. 1 as follows: N f = K ( 1 σ 0 ) n (1) Where   N f is the amount of fatigue life of repeated loads at the time of failure;   σ 0 is the fatigue life for the initial flexural tensile stress (MPa); K and n are constants of material performance.

The relationship between N f and stress is presented as single logarithm plots and analyzed by its linear regression in Figure 5D. In this work, five stress levels ( σ / σ 0   of 0.3, 0.4, 0.5, 0.6, and 0.7) were chosen to perform statistical analysis for the fatigue equation. It was evident that with increasing stress level the fatigue life of all specimens substantially decreased. The 0.3% PET exhibited the highest fatigue life for all stress levels. While the repeated loading cycles of 0.4% rCDA at all stress levels were lower than that of 0.3% PAN, indicating the great influence of rCDA fiber in the fatigue life of the mixtures. Besides, the difference between the repeated loading cycles of 0.4% rCDA and the other two mixtures decreased at lower stress levels. Therefore, the fatigue life of 0.4% rCDA was technically like that of 0.3% PAN and 0.3% PET at lower stress levels when they were used in pavement construction. The worst fatigue performance of 0.5% rCDA was obtained at a higher stress level. The obtained results also indicated that the content of rCDA fiber should be rather decreased by increasing the asphalt content when using this type of asphalt mixture in the pavement construction.