Investigating Degree of Blending Between Reclaimed Asphalt Pavement and Virgin Binder Using Atomic Force Microscopy and Focused Ion Beam

Binder Sample Preparation

In this part of the study, unless stated otherwise, virgin binder and extracted RAP binder refer to PG 58-28 virgin binder and extracted and recovered RAP binder following AASHTO T 319, respectively. Blended binder is a homogeneous binder mixed with 65% virgin binder and 35% extracted RAP binder at 150°C for 3 min. Blending zone means the transition area between virgin binder and extracted RAP binder.

Binder samples for AFM mapping were prepared by the following procedures: (i) heat binder at 150 ° C for one hour; (ii) apply a small drop of binder on an aluminum substrate using a syringe; (iii) heat substrates at 150°C again to melt binder drops on substrates into thin films. Virgin binder, RAP binder, and blended binder film were formed on independent substrate. To manufacture a sample with simulated blending zone, virgin binder and RAP binder were dropped on two sides of one substrate. This substrate was then heated at 150°C and went through visual check every 15 s. As soon as the two binder drops melted, spread, and began contact with each other, the substrate was removed from the heat and left to cool at room temperature. This preparation procedure was similar to those applied elsewhere ( 12 , 15 , 17 ), the only difference being the heating temperature and duration, because of difference in binder stiffness.

Peak Force Quantitative Nanomechanical Mapping and Test Parameters

The peak force tapping quantitative nanomechanical mapping (PFT QNM) of AFM measurement was carried out to measure the microstructure and mechanical properties of binders and the blending zone. Compared with the most commonly used tapping mode, PFT not only captures topographic information, but also measures quantitative nanomechanical properties simultaneously.

During mapping, the tip first approaches the sample surface, which increases the force between tip and material (attraction to repulsion). Once the tip touches the material, the cantilever that connects the tip bends slightly because of deflection, resulting in the tip deforming the material. During this process, the force between tip and material keeps increasing. As soon as the force reaches the target value (pre-defined peak force), the tip retracts from the surface of the material and overcomes the adhesion force between the tip and the material until it completely detaches from the surface and returns to the neutral point. Through this process, a complete force-distance curve is established.

An individual force-distance curve is obtained in real time for each image pixel during measuring, and then transferred into a force-separation curve (Figure 1). This curve is further used for fitting, analyzing, and finally obtaining information such as topography, elasticity (modulus), and other data ( 18 ). The separation axis (Figure 1) is the cantilever deflection. Peak force is the difference between the maximum pre-defined load and noncontact baseline. A fit of the unloading portion of the cycle using the Derjaguine-Müllere-Toporov (DMT) model is utilized to calculate the DMT modulus. More information on parameter calculation can be found elsewhere ( 15 , 19 ).

OPEN IN VIEWER

Figure 1.

Schematics showing force versus separation curve obtained during peak force tapping quantitative nanomechanical mapping (PFT QNM) operation (after Hussain et al. [ 18 ]).

A Bruker Icon AFM was employed for mapping at room temperature (~22°C). A silicon tip (Model: OLTESPA-R3) was first calibrated using reference materials through a thermal tune procedure before the actual test. Then the following parameters were used throughout all tests:

Surface Morphology and Mechanical Properties of Asphalt Binders

Surface topographic images and their corresponding DMT modulus histogram of virgin binder, extracted RAP binder, blended binder, and blending zone are presented in Figure 2. Only one measurement was performed on each of the first three pure binder samples. The image of the blending zone, however, was more difficult to obtain, because there was no reliable method to differentiate virgin and RAP binder, whether through visual observation or using optical microscope embedded in the AFM. In other words, considering the size difference between scan area and possible blending zone (20  μ m × 10  μ m versus 500  μ m × 2000  μ m), it is difficult, if not impossible, to land exactly on an ideal blending zone in one trial. Thus, five successive scans were performed and Figure 2d presents one of them.

OPEN IN VIEWER

Figure 2.

Surface topography and corresponding Derjaguine-Müllere-Toporov (DMT) modulus histogram of (a) (unaged) virgin binder, (b) recycled asphalt pavement (RAP) binder, (c) blended binder, and (d) blending zone between virgin and RAP binder.

One of the most dominant characteristics on all the topography images is the so-called bee structure ( 14 , 20 ) or catana phase ( 19 ), which are slightly elongated, elliptical shaped patterns with black and white stripes. The characteristics of these structures are possibly determined by the chemical composition of binders ( 19 ) or they may be attributed to interactions between crystallizing paraffin waxes and the remaining non-asphalt components ( 20 ).

According to the height legend beside each plot, these structures are height oscillations or surface wrinkling perpendicular to the binder surface (Figure 2). Among these bee structures, the black stripe represents the bottom spots while the white stripe sits on the peak. For virgin binder, such bee structure is relatively long; and it becomes smaller in RAP binder. It is hardly noticeable in blended binder and even more difficult to pinpoint on the blending zone. Clearly, surface topography is capable of differentiating asphalt binders based on stiffness and aging history.

However, topographic characteristics alone might not be enough to distinguish among binders because: (i) it is inconclusive whether such bee structure is a surface or a bulk phenomenon ( 21 ), and (ii) the bee structure could melt or dissolve when time and temperature change ( 20 ). In other words, these dominant patterns that distinguish binders may not exist within binder bulk; even if they do exist, they could disappear at high temperature, which is common during blending of asphalt binders. This observation implies that the bee structure may not be a fundamental material property that can be used to quantify blending between RAP binder and virgin binder.

Micro-mechanical measurement manifests its advantages as it offers information on fundamental properties of the test material such as elasticity. Modulus measurement provides two-dimensional images like Figure 2. Instead of giving height information, however, it maps DMT modulus. Large errors reported in the past ( 15 , 17 ) and observed from binder results imply that average modulus is associated with high error because of surface oscillation. For this reason, the histogram of DMT modulus (Figure 2) on the scan area of each binder, which demonstrates frequency distribution of DMT modulus, was used for evaluation. The frequency distribution plot lumps all data in the mapping area and thus avoids statistical insignificance caused by high variation. Peak value in each histogram plot represents the highest possible DMT modulus of the test binder.

DMT modulus differs significantly among binders. The highest possible DMT modulus value ranking of virgin binder, RAP binder, and blended binder match the corresponding shear moduli measured at 20°C and 10 rad/s using a dynamic shear rheometer, which were 1.850 MPa, 7.25 MPa, and 5.61 MPa, respectively. But they are not strictly proportional, because of differences in test mechanisms.

The peak DMT modulus frequency of the blending zone is closer to that of virgin binder rather than blended binder. Possible reasons for such observation are twofold: (i) virgin binder flows faster than RAP binder during sample preparation process because of lower viscosity. As a result, even though two binder drops flow toward each other simultaneously, chances are higher that the virgin binder covers the RAP binder, (ii) visual estimation is not the most reliable way to determine the measurement point. Even though multiple measurements were performed, considering the relative size difference between mapping area and possible blending zone, one may ask if the mapping areas are indeed representative of the blending zone. This observation raises questions on previous studies, all of which employed a similar approach for sample preparation and AFM measurement.

Additional questions prevented the authors from further pursuing the above approach: (i) the blending zone in reality is much narrower than the one simulated using the described drop-contact method. Reported data ( 17 ) demonstrate a transition zone as wide as 2,000 μm when adopting this approach for mapping. Obviously, such a wide range of transition zone is unrealistic considering that the binder layer around an aggregate is normally not thicker than 100 μm, (ii) the binder sample preparation method used in this and previous studies cannot simulate effect of mechanical force (blending and shearing) during mix production, and (iii) the effect of aggregate on blending is not accounted for. Because of these shortcomings and to fully explore the potential of AFM, RAP aggregates were incorporated for further investigation.

RAP Aggregate Sample Preparation: Mechanical Polishing

AFM requires an extremely smooth surface to land the tip and perform measurements. The binder sample does not require polishing because it has a sufficiently smooth surface for AFM mapping, but the same cannot be said for RAP aggregate. As a result, the first challenge of using real RAP aggregate is to find a way to expose a surface that contains layered structure (virgin binder-blending zone-RAP binder-aggregate) without jeopardizing their interaction, then polish that surface as smooth as possible. For an irregularly shaped object like RAP aggregate, one approach to tackle this challenge is through mechanical polishing: first embed RAP aggregate into epoxy, then polish the solid bulk sample down to the desired depth with an exposed surface that contains the complete layered structure, finally fine polish the surface to the target roughness.

To prepare samples for polishing, 300 g of air-dried RAP aggregate retained on 9.5 mm sieve was mixed with 6 g of PG58-28 binder at room temperature without applying pre-mix heating to RAP aggregates. Individual aggregate was then placed into plastic containers and soaked with fast curing viscosity epoxy system (Figure 3a). Although only a limited number of samples were subjected to the polishing process, 300 g of RAP aggregate were mixed to ensure effective blending. Finally, bulk specimens were extracted from plastic containers after the epoxy was hardened, (Figure 3b left).

OPEN IN VIEWER

Figure 3.

(a) Aggregate embedded in epoxy, (b) specimen after extraction from container (left) and after mechnical polish (right), (c) cross section of a polished sample, and (d) artificial recycled asphalt pavement (RAP) right after mixing.

Bulk specimens were then exposed to mechanical polishing followed by manual polishing with the aid of sandpapers. Sandpapers with coarse grits (lower number, rougher surface) were applied first to remove materials quickly, with a goal of exposing an ideal surface, that is, a clear section showing binder covering RAP aggregate (Figure 3c). No.180 sandpapers were used at this stage. After exposing the desired surface, sandpapers with finer grits (higher number, finer surface) were employed for fine polishing. Sandpapers were applied sequentially from No.240 to No.1000 without skipping grit size because skipping sizes results in polishing the tip of larger flaws rather than removing the entire flaw. Finally, a 1  μ m diamond pad and a 0.03  μ m aluminum pad was used to perform final polishing on each specimen. Through this process, the highest surface roughness after mechanical polishing was 0.03  μ m. Water lubrication was applied during the entire polishing process. Figure 3c shows a polished sample, the black line around the aggregate (light brown color) is (layered) asphalt binder; the transparent part that looks like glass is hardened epoxy.

Preliminary Tests on Actual RAP and Artificial RAP

Figure 4 shows optical profilometer images of a portion of a polished actual RAP sample; the dark portion between epoxy and aggregate is asphalt binder (Figure 4a). It is difficult, if not impossible, visually to differentiate virgin binder and RAP binder and the blending zone between them, which makes selecting a landing zone (maximum 100  μ m in length) for AFM tip challenging. Fines covering the bulk aggregate further raise the level of difficulty. In addition, bulging of fines, up to 30  μ m, is observed (Figure 4b). Because the surface of bulk aggregate and fines are relatively smooth, such bulging is unlikely to be the result of inadequate polishing. It could, however, be attributed to the flow of asphalt binder as a result of the heat generated during the mechanical polishing process.

OPEN IN VIEWER

Figure 4.

(a) Optical profilometer image showing fines covering rock in two-dimensional (2D) image and (b) bulging of fines in three-dimensional (3D) image. Optical profilometer image of polished artificial recycled asphalt pavement (RAP) in (c) 2D and (d) 3D form.

As a result of the above observation, the authors rejected authentic RAP aggregates in favor of artificial RAP to eliminate the influence of fines covering the bulk aggregate. The procedures used to produce artificial RAP in the laboratory are as follows:

The same quantity of 300 g of materials was used for blending again to ensure uniformity. In addition, all tools used for mixing were cleaned carefully to avoid any possible contamination from fines. Selected artificial RAP aggregates (Figure 3d) were further exposed to the mechanical polishing procedure described previously. This time, however, mechanical polishing was applied at a much slower rate to avoid generating an excessive amount of heat.

Figure 4c depicts a much cleaner boundary among the three phases without any fines interfering in between. At first glance, one would suspect the white band in the center of the binder layer to be the blending zone. Unfortunately, this was not the case. Under optical microscope, the white band is a reflection of light from the imaging system because of depth difference. The 3D optical profilometer image in Figure 4d demonstrates such explanation in detail. A depth difference of 140  μ m was formed between binder and aggregate/epoxy surface. Once again, although an extremely smooth surface from the aggregate and epoxy proved the promising results of mechanical polishing, the excessive amount of heat generated during the polishing process caused the binder to flow, and in this case, to be dragged away. The authors tried mechanical polishing of more than 30 samples, all of which either showed similar indentation, or binder hump above the surface. Under no circumstances could an AFM tip land on a surface with such height difference and perform mapping without severely damaging the tip. Therefore, no further testing was pursued on mechanically polished samples.

Focused Ion Beam (FIB) Polishing and AFM Results

The approach to polishing asphalt samples must be site-specific (covering the full binder layer) without introducing extensive heat that can result in binder flow. The focused ion beam (FIB), which is an extension to a scanning electron microscope (SEM), can perform sectioning in this way. With this equipment, one can image and modify a specimen with high spatial resolution. In the polishing process, a beam of ions hits the target sample and causes local sputtering, thus removing materials in a controlled way. With the FIB, one can select precisely where and how to cut through a sample ( 22 ). Additionally, because it also has an SEM, one can simultaneously monitor and adjust the polishing process. A FEI Quanta 200 3D FIB/SEM was employed for sample preparation in this study.

Unlike mechanical polishing applied to complete aggregate, small samples must be used because of space limitation in the FIB instrument chamber. To prepare such samples, artificial RAP aggregates were first frozen to –50°C using liquid nitrogen, then crushed using a mechanical loading frame (Figure 5a). Ideal candidates for FIB sample preparation are flat and thin, with no dimension exceeding 5 mm. Selected samples were then glued to a flat silica substrate, and finally mounted to a special device that was custom made for FIB (Figure 5b). The hypothesis here is that the blended RAP aggregate (artificial RAP covered with virgin binder) has a complete binder layer system, which is the RAP binder covering RAP aggregate, then the RAP binder covered by virgin binder. In this case, and this case alone, one can be confident that the binder in the sample (Figure 5a) has virgin binder, blending zone, and RAP binder at the same time. Because of the high binder content used during two mixing processes, this hypothesis can be satisfied.

OPEN IN VIEWER

Figure 5.

(a) Artificial recycled asphalt pavement (RAP) aggregate after crushing (black edge line is binder) and (b) crushed sample embedded on a silicon substrate and a mounting device.

Figure 6 demonstrates SEM images of a sample before and after FIB polishing. The goal was to create a flat plateau that: (i) long enough to cover full binder layer from edge to interface with aggregate, and (ii) wide enough for AFM tip to land and perform measurement. The first goal is to make sure the full range of asphalt binder—from virgin binder to blending zone to RAP binder—is covered, thus requiring the ion beam to polish from top to bottom (Figure 6b). The second goal is to guarantee that the AFM tip could land and perform measurement. Because the tip height is less than 2  μ m, excessive height oscillation could compromise mapping or, in the worst-case scenario, damage the tip. Although there were undesirable stripes (Figure 6c) on the polished area, AFM measurement was carried out regardless.

OPEN IN VIEWER

Figure 6.

(a) Sample before focused ion beam (FIB) polish, (b) sample after FIB polish, (c) cropped and enlarged image of the FIB polished area, (d) atomic force microscopy (AFM) height images at A position, and (e) AFM height images at B position.

The aforementioned AFM test parameters were carried out to perform the measurements. Results on point A (Figure 6c), which represents virgin binder close to the edge, and point B, which represents RAP binder close to the aggregate, are shown in Figure 6, d and e. Unfortunately, the topography results were not encouraging. At first glance, it looks like there are notable topographic differences from point A to B. However, because these measurements were taken after FIB polishing, surface characteristics were nothing more than polished patterns; that is, the stripes (Figure 6c) or invisible (because of image resolution) obstructions. Coefficient of variation on height and DMT modulus was around 40%, which means that, although the average DMT modulus of point A differs from point B, high standard deviation prevents such difference from being statistically significant.

Several issues were encountered that prevented the authors from fully embracing the great potential of AFM and FIB in this research. First, only one-direction polishing was applied, yet FIB polish in two directions was needed. Specifically, the stripes shown in Figure 6c compromised AFM mapping. To make a sample free of stripes or any kind of imperfection, one must first clear obtrusions close to the FIB gun. In other words, one needs to smooth out the edge part near point A vertically before creating a plateau analogous to the one shown in Figure 6c. This maneuver requires the FIB device to be able to rotate more than 90° and polish two surfaces with an angle of 90°, which was beyond the capacity of the instrument employed in this study. Secondly, asphalt binder is soft and compliant at room temperature, constant tapping by the stiff AFM tip means chances are high that residual binder could stick to the tip and affect the subsequent measurements. This process implies that, as the tip deforms the material during mapping, it also retracts from the binder in every cycle. Because of the high viscosity of binder at room temperature and high oscillation on the binder surface (bee structure and other imperfection), residual binder could be retained on the tip, thus affecting subsequent measurements. No effective method to prevent such error is available yet. And finally, because the ion beam polishes in one direction from top to bottom, it is not clear whether smearing between materials—that is, when materials close to the beam tip move and cover materials at the far end of the beam tip—happened or not. Even though it is easy to navigate the AFM tip to land on the position of interest, there is no guarantee that the measuring spot, for example, point B in Figure 6c, is the real RAP binder. It could be virgin or blended binder because of the possible material smearing during the FIB polishing process. One possible solution for this concern is to use more powerful FIB equipment with high voltage output and longer processing time, both of which are luxuries this study did not have at the time.