Construction and Building Materials 68 (2014) 172–182

Contents lists available at ScienceDirect

Construction and Building Materials

journal homepage: www.elsevier .com/locate /conbui ldmat

Evolution and locational variation of asphalt binder aging in long-lifehot-mix asphalt pavements

http://dx.doi.org/10.1016/j.conbuildmat.2014.05.0910950-0618/� 2014 Published by Elsevier Ltd.

⇑ Corresponding author.E-mail address: ceyhwang@polyu.edu.hk (P.E. Yuhong Wang).

P.E. Yuhong Wang a,⇑, Yong Wen a, Kecheng Zhao a, Dan Chong a, Alvin S.T. Wong b

a Department of Civil & Env. Engr., The HK Polytech. Univ., Hong Kongb R&D Division, Highways Department, Ho Man Tin, Hong Kong

h i g h l i g h t s

� Asphalt binder continuously ages throughout the long-life HMA pavement structure.� The type of asphalt mixture significantly affects binder aging rate.� Aging rate decreases with pavement depth for the same type of asphalt mixture.� Cross-sectional binder aging variation is insignificant.� Asphalt at the bottom of pavement has access to oxygen at a reduced concentration.

a r t i c l e i n f o

Article history:Received 16 March 2014Received in revised form 25 May 2014Accepted 28 May 2014

Keywords:AsphaltLong-life pavementPerpetual pavementAgingMechanistic–empirical pavement design

a b s t r a c t

Understanding asphalt binder aging in long-life hot-mix asphalt (HMA) pavements is critically importantfor rational pavement design and construction practice. Using binder test data obtained at various timesfrom a heavily trafficked 36-year-old HMA road pavement, the study examined the evolution of binderaging as well as variations in aging severity with pavement depth and cross-sectional location. Asphaltconsistency, ductility, and temperature sensitivity were the parameters used to assess the status andthe effects of binder aging. The influence of binder aging on the dynamic modulus (E⁄) of asphalt concrete(AC) was also examined. It was found that asphalt binders continuously and severely age over time, irre-spective of location in the pavement structure. This finding differs from the long-held HMA pavementdesign assumption. Test data also revealed that mixture type and pavement depth have statistically sig-nificant effects on binder aging, but the effect of cross-sectional location (wheel path vs. non-wheel path)is insignificant. Binder aging variations with pavement depth can be attributed to the temperature andoxygen content variations in the pavement structure. The age hardening of asphalt binder leads to anincrease in E⁄. This inevitably affects the load-induced response of the pavement and its durability.The findings are expected to assist in improving the design of long-life flexible pavements based onmechanistic–empirical principles.

� 2014 Published by Elsevier Ltd.

1. Introduction

Reconstruction of highway pavements at the end of their ser-vice lives is not only costly, but also always creates negative socialand environmental impacts. For heavily trafficked roads, the socialand environmental costs of road closure for reconstructionpurposes are so high that pavements with long service lives areusually desired. For instance, long-life pavements that last for atleast 40 years are found to be the most economical solution fortrunk roads in the United Kingdom [1]. The counterpart in the

United States, known as a ‘‘perpetual pavement’’, is expected tolast longer than 50 years without structural damage [2]. In HongKong, life-cycle cost analysis (LCCA) by the highway agency favorsthe use of long-life pavements on all carriageways because of thehigh traffic volume born by such roads and the high social costsassociated with road closures. As a result, a new initiative has beentaken in Hong Kong, by which all carriageway pavements are to beupgraded to the long-life standard (P40 years of design life). It isanticipated that the load-carrying layers of these pavements donot suffer cumulative structural damage or that damage wouldbe so small that noticeable pavement distress within the design lifewould not result. Long-life HMA pavements can bring significantbenefits to society if the promised performance can be delivered.

P.E. Yuhong Wang et al. / Construction and Building Materials 68 (2014) 172–182 173

Long-life (or perpetual-life) HMA pavement design has tradi-tionally focused on both the control of load-induced tensile strainat the bottom of the asphalt concrete (AC) layers and the compres-sive strain at the top of subgrade [3]. A commonly used designstandard is a tensile strain at the bottom of the AC of less than70 ue and a compressive strain at the top of the subgrade of lessthan 200 ue [4]. Within these limits, neither bottom-up fatiguecracking nor subgrade rutting is expected to develop [4]. However,if these critical strains are exceeded, it does not necessarily implythat the pavement cannot achieve a ‘‘long life.’’ Rather, the damagecaused by traffic loads will accumulate over a period of time, aswill pavement distress. The design life will then be a function ofthe end-of-the-life pavement distress thresholds, traffic loads,and the mechanistic responses of the pavement under the trafficloads and environmental factors. It is possible that the pavementlife can be over 40 years under a combination of these factors evenif the strain limits are exceeded. The performance life can beassessed using a mechanistic–empirical pavement analysis proce-dure such as the Mechanistic–Empirical Pavement Design Guide(ME-PDG) [7].

The various mechanistic and empirical models in ME-PDG aredependent on the assumptions of material property changes withtime. Currently, it is assumed that asphalt binder aging and theresultant mixture hardening only take place at the pavement sur-face. This assumption is based on the findings of several early stud-ies [5,6]. Recent evidence in the US, however, suggests that binderaging may also take place at greater pavement depths [8–10]. Sev-eral research findings on asphalt binder aging with time are sum-marized in Table 1. As shown in the table, contradictory resultswere reported. In addition, the research findings may not be appli-cable to long-life asphalt pavements due to several reasons: (1)None of the pavements in the studies are even close to long life,(2) the pavements are much thinner than the typical long-lifepavement, (3) detailed layer-by-layer comparisons of aging statusare missing, (4) historical binder test data for longitudinal compar-ison is unavailable, and (5) statistical significance test results arenot reported. Moreover, some of the field samples used in the stud-ies were exposed in an environment without air conditioning for along time (e.g., over 10 years) [8]. The likelihood of further agingmay make the samples different than those freshly obtained fromthe field.

If binder aging is limited to the pavement surface, a long-lifepavement can be achieved with an adequate thickness designand timely maintenance activities such as resurfacing. Conversely,if severe binder aging occurs deeper within the pavement, it willchange the stiffness and fatigue resistance of those layers. This fur-ther leads to the change of mechanistic responses and performanceof the pavement structure under traffic loads. Therefore, knowl-edge of the characteristics and effects of asphalt oxidative agingin long-life HMA pavements is critically important for the develop-ment of appropriate design, construction, and preservationstrategies.

The purpose of the research is to identify asphalt aging charac-teristics in HMA pavements that are close to ‘‘long-life’’ status with

Table 1Change of binder aging with pavement depth reported in existing literature.

Aging pattern shown from field studies

Aging greatly diminished within the top 25–39 mm of pavement surface

Aging can be neglected below the top 38 mm of pavement surfacePavements oxidize at uniform rates with depthAging in colder regions is slower than that in warmer regions; Aging severity initially

with pavement depth but increases towards the pavement bottom

particular focus on: (1) the evolution of binder aging over time, (2)the variations of binder aging with pavement depth and location,(3) the effect of binder aging on the HMA’s dynamic modulus(E⁄), which is a fundamental property of the HMA mixture, and(4) factors that potentially affect binder aging in a pavement struc-ture. The answers to these questions are also important for thedesign of conventional HMA pavements.

2. Research method

2.1. Pavement samples used for analysis

The HMA pavement sample cores used in this study wereobtained from Tuen Mun Road in Hong Kong. Tuen Mun Road, witha total of six lanes, is one of the most heavily trafficked express-ways in Hong Kong. The first phase (3 lanes Eastbound) wasopened to traffic in 1977, with the second phase opened a fewyears later. The asphalt layers consist of a 30 mm open-graded fric-tion course (OGFC), 40 mm wearing course (WC), 60 mm basecourse (BC), and 150 mm roadbase (RB). Since the initial construc-tion, several resurfacing operations have occurred in various loca-tions, but the BC and RB layers have never been replaced. The roadis currently under reconstruction mainly for realignment purposesand the addition of shoulders (Fig. 1). Although fatigue crackinghas started to appear at certain locations, pavement conditionsfor the majority of the road are still good, as shown in Fig. 1. Thereis confidence that the pavement can last for another 4 years in agenerally good condition and hence literally meet the ‘‘long-life’’definition. Therefore, the road was selected for the study of binderaging characteristics.

2.2. Analysis of historical data

In addition to its long service life, a further advantage of thisroad for analysis is the availability of test data of the original bin-der as well as the extracted binder after 8 years of road usage. Thedata, resulting from collaborative research with the Hong Konghighway agency in the 1980s, was published in a landmark paperby Mcleod [11]. Although the test methods and parameters usedat that time are different from those used today, established equa-tions are available for converting the test results obtained at differ-ent times to equivalent values for comparison purposes. Hence, theevolution of the asphalt binder during the whole lifecycle can beexamined.

2.3. Asphalt sample preparation

Six sample cores of 100 mm diameter were taken from theroad for analysis. The sampling locations are shown in Fig. 2. Ateach longitudinal location, samples were taken at wheel pathand non-wheel path locations, respectively. The cores wereimmediately placed in glass jars, which were subsequently filledwith nitrogen gas and sealed to preclude further contact withair. The cores were then cut into six slices, including the OGFC,

Data source Existing literature

Material master databasefrom various studies

Coons and Wright [5]

Georgia, US Mirza and Witczak [6]Texas, US Glover et al. [8] and Al-Azri et al. [9]

decreases Texas and Minnesota, US Woo et al. [10]

Fig. 3. Five slices of specimens stored in a glass jar filled with nitrogen.

Fig. 4. Partial FTIR spectrum of one of the extracted binder specimens.

Fig. 2. Sampling locations of the cores.

Fig. 1. Reconstruction of the 36 year-old road pavement (Note: the left part of thephotograph shows the old pavements).

174 P.E. Yuhong Wang et al. / Construction and Building Materials 68 (2014) 172–182

WC, BC, and three slices of RB. The separated specimens, exceptfor the OGFC layer, were replaced in the glass jars with silicagel to absorb moisture, refilled with nitrogen, sealed and storedin an air-conditioned environment before binder extraction(Fig. 3).

After most of the moisture in the sliced cores had been removed(indicated by the color change of silica gels), the cores were takenout and broken into small pieces by hammer. Asphalt from thesesmall pieces was extracted using an extraction unit bowl accordingto the ASTM standard D2172/D2172M-11 [12]. The solvent usedfor extraction was dichloromethane. To ensure that the fine parti-cles were thoroughly removed, the extracted bitumen solution wasclarified in a sample tube centrifuge machine in accordance withthe European Standard EN 12697-3:2005 [13]. During the centri-fuging process, a small amount of silica gel that is finer than0.063 mm was added to the solution to further remove the residualwater. Asphalt binder was recovered from the clarified solution byusing a rotary evaporator, following the European Standard EN12697-3:2005. To ensure that the solvent was thoroughly removedduring the rotary evaporation, the extracted bitumen was exam-ined using Fourier Transform Infrared Spectroscopy (FTIR). Theabsence of a peak at the infrared band of 1265 cm�1 in the FTIRspectrum indicates total evaporation of the solvent (Fig. 4) [14].After the bitumen had been recovered from the rotary evaporation,it was stored in a jar, refilled with nitrogen, and placed in a refrig-erator for further analysis.

2.4. Binder test method

Tests using a dynamic shear rheometer (DSR) were conductedon the extracted binders based on the controlled stress model, withstrain levels chosen to be in the linear viscoelastic range of thebinder. An 8 mm plate with a 2 mm gap was used to run the tests.The test temperatures ranged from 10 �C to 82 �C with equalincrements of 6 �C, and frequency ranged from 0.1 Hz to 30 Hz.Parameters obtained from the binder test were compared for tem-perature consistency, durability, and temperature susceptibility.The detailed parameters and comparison process are discussedbelow.

3. Evolution of asphalt binder properties

3.1. Change of temperature consistency

The physical properties of asphalt binder are sensitive to tem-perature changes. Various approaches have been used to measure

P.E. Yuhong Wang et al. / Construction and Building Materials 68 (2014) 172–182 175

asphalt consistency at different temperatures, including penetra-tion, softening point, viscosity, and shear modulus measured by aDSR. The original binder used for the Tuen Mun Rd. in the 1970swas a Pen 80/100 binder, with penetration value of 88 at 25 �C[11]. The original test data also includes the softening point andkinematic viscosities at 135 �C, 146 �C, and 157 �C. The originalbinder test results after thin film oven (TFO) treatment are alsoavailable, including penetration values at 25 �C, softening points,and viscosities at 135 �C. These values represent the conditions ofthe binder just after short-term aging in the mix plant and con-struction. The test results of the extracted binders after 8 years ofservice include penetration values at 25 �C and Penetration–Vis-cosity Numbers (PVN). For the most recently extracted binders,the test data included complex shear moduli and phase anglestested at different temperatures and frequencies using the DSR.To make consistent comparisons between the binders of differentages, derived viscosity at 25 �C was used.

Two equations developed in the Texas Department of Transpor-tation [15] were used to convert penetration to viscosity. Eq. (1)was used for penetration values greater than 54 and Eq. (2) wasused for values less than or equal to 54.

l ¼1:559719� 109 � ln 0:0275

0:0005�0:0114488P

� �

P2 ð1Þ

l ¼1:559719� 109 � ln 0:0275

3:94�10�6�Pþ0:000075

� �P2 ð2Þ

where:l = viscosity in poise;P = penetration in penetration units.

The field test conducted in the 1980s [11] involved five sam-pling locations. At each location, pairwise comparisons were madebetween the surface layer (wearing course) and the base layer(including base course and roadbase). The log viscosities convertedfrom penetration values for the specimens at the 5 sample loca-tions are shown in Fig. 5. The average penetration was 22.6 forthe surface layer and 22.7 for the base layers, corresponding toan average log viscosity (poise) of 7.22 and 7.19, respectively.The paired-t test revealed that the small differences in penetrationand log viscosity are not statistically significant. It can be con-cluded, therefore, that the asphalt binders in the base and surfacelayers had aged to similar extents after 8 years of road use. The vis-cosity data obtained from the base layers was used for comparingthe extent of binder aging at different times along the length of theroad.

The recent DSR measurements were converted to viscosity at25 �C in three steps for comparison purposes. Firstly, viscosities

6.9

6.95

7

7.05

7.1

7.15

7.2

7.25

7.3

7.35

7.4

1 2 3 4 5

Surface

Base

Fig. 5. Log viscosities of the extracted asphalt specimens after 8 years of road use(converted from penetration values, data source: [9]).

measured at different temperatures and the frequency of 13.58(rad/s) were converted to viscosity using the following equation[16]:

g ¼ jG�bjx

� �1

sindb

� �a0þa1xþa2x2

ð3Þ

where:g = viscosity, P (Note: The viscosity unit may be mistakenlywritten as cP in the original equation. It has been changed toPoise in this paper);jG�bj = binder shear modulus, Pa;db = binder phase angle, degree;x = angular frequency, rad/s;a0, a1, a2 = fitting parameters, 3.639, 0.1314, and �0.0009,respectively.

Within the range of the DSR measurements, it was found thatthe log viscosity (Poise) and temperature (�C) are well depictedby a linear relationship, as shown in Fig. 6. Therefore, in the secondstep, linear regressions were conducted between log viscosity andtemperature for all extracted binders. As the third step, the log vis-cosities at 25 �C were calculated from the linear regression equa-tions. The viscosity data obtained from the bottom lift (50 mm)of the roadbase layer (RB3) was chosen for making comparisonsof viscosity changes over time.

The derived viscosity data at the four time points are summa-rized and shown in Fig. 7. The figure shows that binder viscositycontinuously increased over the different stages, but at a slowerrate in the later stage from years 8 to 36. From the original binderto that treated after TFO, the log viscosity increased by 0.40. For thenext 8 years, the average log viscosity increased by 0.85. For thenext 28 years, the average log viscosity increased by 1.39. The datafor years 8 and 36 were obtained from road base; therefore, the fig-ure also provides clear evidence that oxidative aging occurs at thedeep layer of the asphalt pavement.

3.2. Change of binder durability

The rheological data can also be used to assess the aging-relateddurability properties of asphalt binder. It has long been recognizedthat the durability of asphalt binder is strongly related to its brit-tleness, represented by its ductility at 15 �C and with an appliedelongation rate of 1 cm/min. Ruan et al. [17] found that ductilitymeasured under these conditions correlates well to the rheological

parameterG0g0

G0, namely the DSR function. In calculating the DSR func-tion, the values of g0 and G0 measured at 44.7 �C and 10 rad/s in aDSR are converted to values at 15 �C and 0.005 rad/s. Glover et al.

y = -0.065x + 9.9947R² = 0.9993

0

2

4

6

8

10

12

0 20 40 60 80 100

Log

Vis

cosi

ty (

P)

Temperature (Celsius)

BC

RB1

RB2

RB3

Linear (RB3)

Fig. 6. Relationship between the estimated log viscosity (P) and temperature (�C)within the range of DSR measurements.

y = -3.4814x + 10.397R² = 0.9997

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

2.7 2.75 2.8 2.85 2.9

log-

log

Vis

cosi

ty (

cP)

Log Temperature (Rankine)

Fig. 8. Viscosity–temperature characteristic relationship of the original binder.

Fig. 7. Change of log viscosity (P) values at 25 �C with time.

176 P.E. Yuhong Wang et al. / Construction and Building Materials 68 (2014) 172–182

[18] further recommended the use of a ductility value of 3 cm at15�Cand 1 cm/min of elongation rate (with DSR function value>0.003 MPa/s) as the threshold for binder failure, and a ductilityvalue of 5 cm (DSR function value >0.0009 MPa/s) as the thresholdfor pavement cracking. In this study, the extracted binders weretested at 44.7 �C and 10 rad/s and the parameters obtained wereused to calculate the DSR function and estimate binder ductility.

The ductility values of the original binder and that treated afterTFO, as reported by Mcleod [11], are both greater than 110 cm.However, the measurements were taken at 25 �C and hence arenot compatible with the estimated ductility at 15 �C and 1 cm/min elongation rate. The following steps were made in this studyto estimate the DSR function and ductility values related to the his-torical rheological data.

Using the well-known viscosity–temperature characteristicrelationship in ASTM D2493-85 (Eq. (4)) [19], the viscosity at dif-ferent temperature levels can be determined.

log log g ¼ Aþ VTS log TR ð4Þ

where:g = viscosity, cP;A = regression intercept;VTS = regression slop (viscosity–temperature susceptibilityparameter);TR = temperature, degree Rankine.

The viscosity of the asphalt binder at 25 �C was estimated fromthe penetration value based on Eqs. (1) and (2). The viscosity at thesoftening point is commonly assumed to be 13,000 poise [6]. Theviscosities at 135 �C, 146 �C, and 157 �C were converted from

centistokes (CTS) to centipoise. Thus, five data points wereavailable to develop the viscosity–temperature characteristic curvefor the original binder, as shown in Fig. 8.

As stated above, the test data for the asphalt after TFO treat-ment includes penetration values at 25 �C, the softening point,and viscosity at 135 �C. These three data points were used to con-struct the viscosity–temperature characteristic relationship for thebinder after short-term aging. The test results on the bindersextracted after 8 years of service include the penetration value at25 �C and the Penetration–Viscosity Number (PVN). The viscositiescorresponding to the penetration values can be calculated using Eq.(2), while the viscosities at 135 �C can be calculated from the PVNvalue. PVN has been used to assess the temperature susceptibilityof an asphalt binder and is calculated according to Eq. (5) [20]. Byknowing the penetration values and PVN, the kinematic viscosityat 135 �C can be obtained. In this fashion, the viscosity–tempera-ture characteristic curves were constructed for the original binder,the binder after TFO treatment, and after 8 years of road life.

PVN ¼ �1:5ðL� XÞL�M

ð5Þ

where:L = 4.258–0.79674(logPen@77 �F);M = 3.46829–0.61094(logPen@77 �F);X = log(Kin.Visc.@275 �F).

A procedure developed by Bari and Witczak [16] was used toestimate the DSR phase angle (db) and complex shear modulusðG�bÞ from the viscosity–temperature relationship. The processand equations for performing such estimations are brieflyintroduced as follows [14]:

(1) Development of a frequency-related viscosity–temperaturerelationship:

log log gfs ;T ¼ A0 þ VTS0 log TR ð6Þ

A0 ¼ 0:9699f�0:0527s � A ð7Þ

VTS0 ¼ 0:9668f�0:0575s � VTS ð8Þ

where:

gfs ;T = viscosity of asphalt binder as a function of both load-ing frequency (fs) and temperature (T), cP (for calculatedgfs ;T > 3� 1012 cP, use =3 � 1012 cP);fs = loading frequency in dynamic shear mode as used in theG�b testing, Hz;A = regression intercept from the conventional ASTMAi � VTSi equation (Eq. (4));VTS = slope from the conventional ASTM Ai � VTSi equation(Eq. (4));A0 = adjusted A (adjusted from loading frequency);VST0 = adjusted VTS (adjusted from loading frequency);TR = temperature in Rankine scale, �R.

(2) Estimate of gfs,T at temperature 44.7 �C and loading fre-quency of 10 rad/s based on Eqs. (6)–(8) and the identifiedA0 and VTS0.

(3) Estimate of db based on Eq. (9) and the parameters obtainedabove.

db ¼ 90þ ðb1 þ b2VTS0Þ � logðfs � gfs ;TÞ þ ðb3 þ b4VTS0Þ � log ðfs �gfs ;TÞ2

ð9Þ

where:

db = phase angle, deg (for calculated db > 90�, use db = 90�);VTS0 = adjusted VTS (from Eq. (8));fs = loading frequency in dynamic shear, Hz;

P.E. Yuhong Wang et al. / Construction and Building Materials 68 (2014) 172–182 177

gf,T = viscosity of asphalt binder as a function of both loadingfrequency (fs) and temperature (T), cP (from Eq. (6));b1, b2, b3 and b4 = fitting parameters =�7.3146, �2.6162,0.1124 and 0.2029 respectively.

(4) Estimate of jG�bj based on Eq. (10) and the parametersobtained above.

Fig. 9.min wi

jG�bj ¼ 0:0051f sgfs ;TðsindbÞ7:1542�0:4929f sþ0:0211f 2s ð10Þ

where:

Fig. 10. Change of the absolute values of VTS slope with time.

G�b = dynamic shear modulus, Pa (for calculated G�b > 1 GPa,use G�b ¼ 1 GPa);fs = dynamic shear loading frequency to be used with G�b anddb, Hz;gfs ;T = viscosity of asphalt binder as a function of both load-ing frequency (fs) and temperature (T), cP (from Eq. (6));db = phase angle, deg (from Eq. (9)).

Once the db and jG�bj are estimated, they can be easily convertedto G0 and g0 to calculate the DSR function. The derived ductility val-ues at different times are summarized and presented in Fig. 9. Asshown in the figure, the ductility values of the asphalt binder con-tinuously decline over the years. After 8 years of road use, the aver-age ductility had reduced to 13.2 cm; however, this is stillacceptable according to the criteria proposed by Glover et al.[16]. After 36 years, the ductility values of all the binders extractedat the bottom lift of the pavement reached ‘‘failure’’ E⁄ status.Therefore, aging-related brittleness may be the predominant con-cern affecting the durability of long-life HMA pavements. The fig-ure also shows that the derived ductility values are much lessvariable than the estimated log viscosity values. It is clear thatthe binder ductility and durability continuously deteriorate overtime, even at deep pavement layers.

3.3. Change of binder temperature susceptibility

The temperature susceptibility of the asphalt binders at differ-ent times was assessed using the slope of the viscosity–tempera-ture characteristic curve, i.e., the VTS. As stated above, theviscosity–temperature characteristic curve was constructed foreach binder using Eq. (4). The change of the absolute value ofVTS over time is shown in Fig. 10, which indicates that the slopedecreases significantly after long-term aging. It appears, therefore,that asphalt binder consistency becomes less sensitive to temper-ature variation as it oxidatively ages. The reduction in temperaturesensitivity is more prominent from year 8 to 36. This is eventuallyreflected in the decrease in sensitivity of the shear modulus of thebinder and the of AC to temperature change.

Change of estimated ductility values at 15 �C and elongation rate of 1 cm/th time.

4. Locational variations of binder aging

The total thickness of AC layers in a long-life HMA pavement isusually greater than that of a conventional HMA pavement. Thevariations of temperature and oxygen content in the thick AC lay-ers may lead to binder aging variations throughout the pavementdepth. In addition, traffic loads may cause further compaction atthe wheel path and/or induce micro-cracks, both of which may cre-ate cross-sectional aging variations. Therefore, the following twoquestions on the locational variations in binder aging are of inter-est in this study: (1) whether the severity of binder aging changeswith AC pavement depth after long-term field service, and (2)whether there is any difference in binder aging between samplesobtained at wheel paths and those distant from the wheel paths.These two questions were examined in relation to the consistencyand durability of the extracted binders. The log-viscosity at 22 �Ccalculated from the DSR measurements and Eq. (3) were used torepresent binder consistency, while the derived ductility valuesfrom DSR measurements at 44.7 �C and frequency of 10 rad/s wereused to represent durability. The test data for the different speci-mens are shown in Table 2.

The paired-t test was used to check for differences between thegroups of specimens. To test the difference in aging between pave-ment layers, pairwise comparisons (BC vs. RB1, BC vs. RB2, etc.)were made for specimens obtained from the same sample core.To test the aging differences between wheel path and non-wheelpath, two specimens from the same depth and at the same longitu-dinal locations (Fig. 2) were treated as a pair. The test statistics onthe effect of pavement depth, based on log viscosity, are shown inTable 3. They indicate that the average log viscosity of BC is lowerthan that of RB1 and RB2. For the three RB lifts, the log viscositydecreases with pavement depth.

The test statistics for the effect of pavement depth based on thederived ductility are shown in Table 4. The results follow a patternsimilar to that in Table 2. The derived ductility of BC is significantlyhigher than RB1 and RB2 values. For the three RB lifts, ductilitygenerally increases with pavement depth, although the differencebetween RB1 and RB2 is marginally insignificant at the 0.95 signif-icance level.

Tables 3 and 4 clearly indicate that the types of material play asignificant role in binder aging. Although the temperature of the BClayer is generally higher than that of the RB layers and highertemperature accelerates oxidative aging, the severity of aging ofthe BC layer is less than that of the two adjacent RB layers. Forthe same type of material (RB), aging severity apparently decreasewith pavement depth.

For the specimens obtained from wheel path and non-wheelpath, both the log viscosity and ductility values show no significant

Table 2Estimated ductility and log-viscosity of the extracted asphalt binders.

Sample no. Layer Location DSR function Ductility (cm) Log-viscosity (22 �C)

C1 BC Non-wheel path 0.0058 2.21 8.7675C1 RB1 Non-wheel path 0.0076 1.97 8.8952C1 RB2 Non-wheel path 0.0077 1.96 8.7985C1 RB3 Non-wheel path 0.0046 2.45 8.6014C2 BC Wheel path 0.0028 3.08 8.4089C2 RB1 Wheel path – –C2 RB2 Wheel path 0.0105 1.71 9.0946C2 RB3 Wheel path 0.0091 1.82 9.0226C3 BC Non-wheel path 0.0075 1.98 8.7923C3 RB1 Non-wheel path 0.0409 0.94 9.6382C3 RB2 Non-wheel path 0.0094 1.79 9.0368C3 RB3 Non-wheel path 0.0129 1.56 9.1122C4 BC Wheel path 0.0084 1.88 8.7976C4 RB1 Wheel path 0.0337 1.02 9.6110C4 RB2 Wheel path 0.0257 1.15 9.5420C4 RB3 Wheel path 0.0093 1.80 8.9422C5 BC Non-wheel path 0.0036 2.72 8.5573C5 RB1 Non-wheel path 0.0154 1.44 9.3784C5 RB2 Non-wheel path 0.0089 1.83 8.8973C5 RB3 Non-wheel path 0.0077 1.95 8.7714C6 BC Wheel path 0.0052 2.32 8.7726C6 RB1 Wheel path 0.0095 1.78 8.9043C6 RB2 Wheel path 0.0083 1.89 8.8178C6 RB3 Wheel path 0.0034 2.81 8.5367

Table 3Statistics for paired-t test based on log-viscosity.

Comparisons Average difference Standard dev. No. of observation t-Value Significant at 0.95 significance level (one-side)?

BC–RB1 �0.548 0.382 5 3.208 YesBC–RB2 �0.348 0.308 6 2.772 YesBC–RB3 �0.148 0.315 6 1.153 NoRB1–RB2 0.267 0.254 5 2.348 YesRB1–RB3 0.493 0.158 5 6.955 YesRB2–RB3 0.2 0.23 6 2.132 Yes

Table 4Statistics for paired-t test based on derived ductility value.

Comparisons Average difference Standard dev. No. of Observation t-Value Significant at 0.95 significance level (one-side)?

BC–RB1 0.793 0.411 5 4.318 YesBC–RB2 0.644 0.447 6 3.528 YesBC–RB3 0.302 0.647 6 1.145 NoRB1–RB2 �0.294 0.348 5 1.891 NoRB1–RB3 �0.681 0.225 5 6.766 YesRB2–RB3 �0.342 0.42 6 1.992 Yes

178 P.E. Yuhong Wang et al. / Construction and Building Materials 68 (2014) 172–182

differences at the 0.95 significance level. Therefore, the cross-sec-tional binder aging differences are minimal. It is worth to mentionthat, although differences in binder aging are statistically signifi-cant for the specimens from several different pavement layers,the magnitude of the differences is actually not large. As shownin Fig. 9, all binder specimens apparently severely aged after36 years of service.

5. Effect of binder aging on the dynamic modulus of AC

Binder aging over time and aging variations with pavementdepth manifest themselves as changes in AC stiffness. A keyparameter used in the ME-PDG is the E⁄ value of AC. Several modelshave been developed to estimate E⁄ based on aggregate gradation,the volumetric properties of the asphalt mixture, and binder stiff-ness [7]. The original model used in ME-PDG to estimate E⁄ isshown in Eq. (11):

logE� ¼3:750063þ0:02932q200�0:001767ðq200Þ2�0:002841q4

�0:058097Va�0:802208Vbeff

Vbeff þVa

� �

þ3:871977�0:0021q4þ0:003958q38�0:000017q238þ0:005470q34

1þeð�0:603313�0:313351logðf Þ�0:393532logðgÞÞ

ð11Þ

where:E⁄ = dynamic modulus, psi;g = bitumen viscosity, 106 Poise;f = loading frequency, Hz;Va = air void content, %;Vbeff = effective bitumen content, % by volume;q34 = cumulative % retained on the 3/4 in sieve;q38 = cumulative % retained on the 3/8 in sieve;q4 = cumulative % retained on the No. 4 sieve;q200 = % passing the No. 200 sieve.

Fig. 11. The change of the estimated E⁄ (25 �C) with time.

P.E. Yuhong Wang et al. / Construction and Building Materials 68 (2014) 172–182 179

This model was adopted in this paper to assess the effects ofbinder aging on E⁄ of the RB layer. The gradation and volumetricsof the RB mixtures used in Hong Kong was used and the followingparameters were employed:

Va = 4.5%;Vbeff = 7%;q34 = 27;q38 = 42;q4 = 54;q200 = 5.

Assuming that the loading frequency ranges from 0.01 to 50 Hzand using the log viscosity data as shown in Fig. 7, the E⁄ value at25 �C over a wide range of frequency was computed as shown inFig. 11.

Fig. 11 indicates that the extent of binder aging has significanteffects on the E⁄ values of the RB layer. For instance, at the fre-quency of 10 Hz, E⁄ (25 �C) of AC after 8 years increased from6050 MPa to 10,514 MPa, and was further increased to20,874 MPa at the end of 36 years. The E⁄ value jumped more thanthree times of its original value during the whole service period.Therefore, ignoring the evolution of binder aging during the servicelife of HMA pavement will result in inaccurate estimates of theload-induced pavement responses.

Fig. 12. Estimated hourly temperature (�C) distribut

6. Factors affecting binder aging in asphalt pavement

It is clear from the above discussion that asphalt aging occurs atall the positions in pavement and the effect of aging is significant.Asphalt oxidization, which is the predominant aging mechanism,has been extensively studied in laboratory settings [e.g. 21,22].Laboratory tests indicate that the oxidative age hardening ofasphalt consists of two stages: an early rapid aging stage and a sub-sequent slow aging stage with a constant hardening rate [8,23].Factors that affect the hardening rate during the second stagecan be expressed by the following equation [21,24,25]:

rg ¼ HS� rCA ð12Þ

where rg = HS � rCA is the hardening rate after the ‘‘initial jump,’’ HSis the hardening susceptibility of a particular type of asphalt, rCA isthe growth rate of the carbonyl area in the infrared spectrum of theasphalt sample, indicating the increased amount of the carbonylfunctional group (AC@O), and

rCA ¼ @CA=@t ¼ APae�E=RT ð13Þ

where A, E, and a are coefficients that are dependent on asphalttype, R is the gas constant, P is oxygen pressure, and T is the abso-lute temperature. E was also found to be a function of oxygen pres-sure [26].

As shown in Eqs. (12) and (13), the oxidative aging rate ofasphalt is dependent on oxygen pressure and absolute tempera-ture. Therefore, oxidative asphalt aging in AC in field conditionsis essentially governed by the distribution of temperature and par-tial oxygen pressure within the AC layers. These two factors havebeen further investigated in this study based on weather dataand field monitoring results.

6.1. Temperature distribution within the asphalt pavement

In this study, the temperature distribution in asphalt pave-ments in Hong Kong were computed based on thermodynamicmodels, local climate conditions and material properties, using aone-dimensional finite difference (FD) method. Heat flux on thepavement surface was calculated using Eq. (14) [27]. Hourlyweather data recorded by the Hong Kong Observatory in 2012was used to calculate the heat flux in the equation. It was foundduring a previous study that soil temperature in Hong Kong12.5–15 m below the surface approaches a constant value of26.5 �C [28]. This temperature was used as the lower boundary

ion at the different depths of AC within a year.

Fig. 13. Installation of oxygen sensors under the pavement: (a) layout of the oxygen sensors, (b) close view of the sensor, and (c) completed pavement section.

Fig. 14. Oxygen concentrations recorded by the sensors after construction.

180 P.E. Yuhong Wang et al. / Construction and Building Materials 68 (2014) 172–182

condition at the 13.23 m in the FD calculation. In a previous study,it was also found that the predicted pavement temperatures basedon the same thermodynamic models and coefficients derived from

the local climate conditions and material properties matched thefield observations on pavement temperature change quite well[29].X

_Q ¼ _Q solar � ~a � _Qsolar þ _Q abs � _Q rad � _Q conv � _Qcond ð14Þ

where:P _Q is the summary of heat flux (W m�2),_Q solar is the heat flux due to solar radiation,~a � _Q solar is the portion of the solar radiation reflected by thepavement surface and ~a is the albedo of the pavement,_Q abs is the down-welling long-wave radiation heat flux from the

atmosphere,_Q rad is the outgoing long-wave radiation heat flux from the

pavement surface,_Q conv is the convective heat flux between the pavement surface

and the air, and_Q cond is the heat conduction at the pavement surface.

The calculated temperatures at different positions in the asphaltpavement at hourly midpoints are shown in Fig. 12. Also shown inthe figure are the averages and standard deviations of thetemperatures at these positions. The figure indicates that average

P.E. Yuhong Wang et al. / Construction and Building Materials 68 (2014) 172–182 181

temperatures and temperature variations at the surface layers aregreater than those at the bottom layers. Although the differences inaverage temperatures are not large, the cumulative effect may besignificant because the temperature difference affects the slopeof the aging curve in Eq. (12). This suggests that the oxidative agingrates of the upper layers will be greater than those of the lowerones. This matches the observation that aging effects decrease withpavement depth in the three RB lifts However, temperature alonedoes not explain the lesser extent of binder aging effects in theBC layer, which is subject to higher temperatures than the RB lay-ers. It can be inferred from Eq. (13) that the binder in the BC layermust have a lower accessibility to oxygen.

6.2. Accessibility of the asphalt binder to oxygen

Partial oxygen pressure in a pavement structure has receivedlittle attention in the literature. An extensive literature review byGlover et al. [18] in 2009 suggests that the study of oxygentransport into HMA pavements and asphalt binder is ‘‘practicallynon-existent.’’ Prapaitrakul [30,31] recently developed an oxygentransport and reaction model, where oxygen was modelled to betransported via air void (AV) channels and diffused into asphaltshells. The average AV size and distance between adjacent poreswere used to predict binder oxidation. Han [32] further refinedthe model by including temperature and the number of air voidsin oxidation prediction. The model assumes that the oxygen con-tent beneath the AC slab is the same as that in the atmosphere.Under this assumption, the aging of asphalt in a long-life pavementessentially depends on the AV in the asphalt mixture and the envi-ronmental temperature. In this study, the partial oxygen presentbeneath the pavement slab was investigated using a full-scalepavement slab.

As shown in Fig. 13, two full-scale 4 m � 4 m HMA pavementsections were constructed for this study. Each section consists of50 mm of WC, 50 mm of BC and 100 mm of sand base. VirginPG64-22 asphalt binder was used in the asphalt mixture. Threeoxygen sensors were installed in the pavements, numbered as211, 212, and 213. Both the sensors 211 and 212 were installedbeneath the AC slab, with 211 being installed in the middle ofthe slab and 212 at the edge of the slab. The sensor 213 wasinstalled in sand uncovered by AC but with the same depth asthe other two sensors.

The oxygen concentrations recorded by these sensors after con-struction were monitored and are shown in Fig. 14. Several obser-vations can be made from the figure. Firstly, the oxygen contentsrecorded by all sensors were less than the atmospheric oxygencontent (21%). Secondly, the oxygen concentrations were subjectto locational variations: The sensor located in the middle of theslab (211) shows reduced concentration compared with the onenot covered by the slab (213) and the one closer to the slab edge(212). Thirdly, the oxygen contents were subject to large variationsover time, most likely due to variations in climate conditions. It canapparently be concluded that asphalt pavement at its bottom isexposed to an environment of lower oxygen content. Although,in this study, the monitored oxygen contents were only slightlylower than atmospheric levels, in a long-life pavement where thetotal thickness of AC layers is much greater than 100 mm, theoxygen content may be further reduced. The difference in oxygencontent at the pavement upper layer and bottom layer creates anoxygen pressure gradient, which also contributes to the variationsof asphalt aging in a pavement structure.

7. Conclusion and discussion

Historical data obtained from a 36-year-old highway pavementat different points in time shows that asphalt binder continuously

aged over time, regardless of the location of the binder within theroad structure. Even at the bottom of the thick AC slab, the bindersaged to ‘‘failure’’ levels based on the estimated ductility values. Thetest data also revealed that material type and pavement depthhave statistically significant effects on binder aging, but the effectof wheel loads is not statistically significant. The age hardening ofasphalt binder results in an increase in dynamic modulus (E⁄) ofAC. Based on the E⁄ estimation model developed in the Mechanis-tic–Empirical Pavement Design Guide (ME-PDG), it is estimatedthat binder aging can cause the E⁄ value of the structure-carryinglayer to increase by over two times of its original value during ser-vice life. This affects the load-induced response of the pavement aswell as durability. Therefore, it is necessary to revisit the binderaging assumption made in the ME-PDG, especially in relation tothe design of long-life flexible pavements. The longevity of long-life pavements appears to be highly dependent on aging-relateddegradation of the asphalt binder. The study also attempted toquantify the distribution of temperature in the pavement structureand monitored the oxygen content at the bottom of a full-scale ACslab. The variations in temperature and oxygen content werebelieved to contribute to binder aging variations with pavementdepth. Since binder aging has detrimental effects on the durabilityof long-life asphalt pavements, future studies need to be directedtowards the improvement of its aging resistance properties.

References

[1] UK Dept. of Transport, Highways Agency. Design manual for roads and bridges,vol. 7, section 2, part 3, HD 26/06; 2006.

[2] APA (Asphalt Pavement Alliance). Perpetual pavements: a synthesis, asphaltpavement alliance; 2002.

[3] Monismith CL, Long F. Overlay design for cracked and seated Portland cementconcrete (PCC) pavement—interstate route 710. Technical memorandum TMUCB PRC 99-3, Pavement Research Center, Institute for Transportation Studies,University of California, Berkeley; 1999.

[4] Newcomb DE, Buncher M, Huddleston IJ. Concepts of perpetual pavements.Trans. research circular number 503. Perpetual bituminous pavements, TRB,Washington, DC; 2001. p. 4–11.

[5] Coons RF, Wright PH. An investigation of the hardening of asphalt recoveredfrom pavement of various ages. Proc AAPT 1968;37:510–22.

[6] Mirza MW, Witczak MW. Development of a global aging system for short andlong term aging of asphalt cements. J AAPT 1995;64:393–430.

[7] ARA. Guide for mechanistic–empirical design of new and rehabilitatedpavement structures. Champaign (Illinois): ARA, Inc.; 2004.

[8] Glover CJ, Davison RR, Domke CH, Ruan Y, Juristyarini P, Knorr DB, et al.Development of a new method for assessing asphalt binder durability withfield validation. Report FHWA/TX-03/1872-2, Texas Transportation Institute,College Station, TX; 2005.

[9] Al-Azri NA, Jung SH, Lunsford KM, Ferry A, Bullin JA, Davison RR, et al. Binderoxidative aging in Texas pavements: hardening rates, hardeningsusceptibilities, and the impact of pavement depth. Trans Res Rec2006;1962:12–20.

[10] Woo WJ, Ofori-Abebresse E, Chowdhury A, Hilbrich J, Kraus Z, Martin AE, et al.Polymer modified asphalt durability in pavements. Report FHWA/TX-07/0-4688-1, Texas Transportation Institute, College Station, TX; 2006.

[11] Mcleod NW. Relationship of paving asphalt temperature susceptibility asmeasured by PVN to paving asphalt specifications, asphalt paving mixturedesign and asphalt pavement performance. J AAPT 1989:410–89.

[12] ASTM International. Standard test methods for quantitative extraction ofbitumen from bituminous paving mixtures. ASTM standard D2172/D2172M-11, ASTM International, West Conshohocken, PA; 2011.

[13] European Committee for Standardization. Bituminous mixtures – test methodsfor hot mix asphalt Part 3: Bitumen recovery: rotary evaporator. EN 12697-3,European Committee for Standardization, Brussels; 2013.

[14] De la Roche C, Van de Ven M, Planche J-P, Van den Bergh W, Grenfell J, Gabet T,et al. Hot recycling of bituminous mixtures. In: Partl MN, Bahia HU, CanestrariF, De la Roche C, Benedetto HD, Piber H, et al., editors. Advances ininterlaboratory testing and evaluation of bituminous materials. State-of-the-art report of the RILEM technical committee 206-ATB. Dordrecht, Heidelberg,New York, London: Springer; 2013.

[15] Texas Department of Transportation. Test procedure for calculating viscosityfrom penetration. TXDOT designation: TEX-535-C construction division,effective date: August 1999; 1999. http://ftp.txdot.gov/pub/txdot-info/cst/TMS/500-C_series/pdfs/aph535.pdf.

[16] Bari J, Witczak M. New predictive models for viscosity and complex shearmodulus of asphalt binders: for use with mechanistic–empirical pavementdesign guide. Transp Res Rec: J Transp Res Board 2007;2001(-1):9–19.

182 P.E. Yuhong Wang et al. / Construction and Building Materials 68 (2014) 172–182

[17] Ruan Y, Davison RR, Glover CJ. An investigation of asphalt durability:relationships between ductility and rheological properties for unmodifiedasphalts. Pet Sci Technol 2003;21(1–2):231–54.

[18] Glover CJ, Martin AE, Chowdhury A, Han R, Prapaitrakul N, Jin X, et al.Evaluation of binder aging and its influence in aging of hot mix asphalt:literature review and experimental design. FHWA/TX-08/0-6009-1; 2009.

[19] ASTM International. Standard viscosity–temperature chart for asphalts. ASTMstandard D2493/D2493M – 09, ASTM International, West Conshohocken, PA;2009.

[20] Mcleod NW. Asphalt cements: pen-vis number and its application to moduli ofstiffness. ASTM J Test Eval 1976;4(4).

[21] Petersen JC, Branthaver JF, Robertson RE, Harnsberger PM, Duvall JJ, Ensley EK.Effects of physicochemical factors on asphalt oxidation kinetics. Transp ResRec 1993;139:1–10.

[22] Domke CH, Davison RR, Glover CJ. Effect of oxidation pressure on asphalthardening susceptibility. Transp Res Rec 1999;1661:114–21.

[23] Petersen JC, Glaser H. Asphalt oxidation mechanisms and the role of oxidationproducts on age hardening revisited. Road Mater Pavement Des2011;12:795–819.

[24] Liu M, Chaffin JM, Davison RR, Glover CJ, Bullin JA. Changes in Corbett fractioncomposition during oxidation of asphalt fractions. Transp Res Rec1998;1638:40–6.

[25] Lau CK, Lunsford KM, Glover CJ, Davison RR, Bullin JA. Reaction rates andhardening susceptibilities as determined from POV aging of asphalts. TranspRes Rec 1992;1342:50–7.

[26] Domke CH, Davison RR, Glover CJ. Effect of oxygen pressure on asphaltoxidation kinetics. Ind Eng Chem Res 2000;39(3):592–8.

[27] Lytton RL, Tsai FL, Lee SI, Luo R, Hu S, Zhou F. Models for predicting reflectioncracking of hot mix asphalt overlays. NCHRP report, issue: 669. TransportationResearch Board, Washington, DC; 2010.

[28] Chow TT, Long H, Mok HY, Li KW. Estimating of soil temperature profile inHong Kong from climatic variables. Energy Build 2011;43(2011):3568–75.

[29] Wang Y, Zhu S, Wong A. Cooling time estimation of newly placed hot-mixasphalt pavement in different weather conditions. J Constr Eng Manage 2014.http://dx.doi.org/10.1061/(ASCE)CO.1943-7862.0000832. 04014009.

[30] Prapaitrakul N, Han R, Jin X, Glover CJ. A transport model of asphalt binderoxidation in pavements. Road Mater Pavement Des 2009(10/SI):95–113.

[31] Prapaitrakul N. Towards an improved model of asphalt binder oxidation inpavements. Ph.D. dissertation. Texas A&M, College Station, TX; 2009.

[32] Han R. Improvement to a transport model of asphalt binder oxidation inpavements: pavement temperature modeling, oxygen diffusivity in asphaltbinders and mastics, and pavement air void characterization. Ph.D thesis.Texas A&M Univ; 2011.