Construction and Building Materials 425 (2024) 135929

Contents lists available at ScienceDirect

Construction and Building Materials

journal homepage: www.elsevier.com/locate/conbuildmat

Phenomenological modeling of aging and rejuvenation on asphalt binder

fatigue characteristics
Saqib Gulzar a, 1, Lei (Gabriel) Xue a, *, 2, Jaime Preciado a, 3, Andrew Fried a, 4,
Cassie Castorena a, 5, Benjamin Shane Underwood a, 6, Jhony Habbouche b, 7, Ilker Boz b, 8
a
North Carolina State University, 915 Partners Way, Raleigh, NC 27695, USA
b
Virginia Transportation Research Council, 530 Edgemont Road, Charlottesville, VA 22903, USA

A R T I C L E I N F O A B S T R A C T

Keywords: Recycling agents (RAs) have been explored as additives to enhance the properties of recycled asphalt binders.
Asphalt Binder Despite extensive research on RAs, a critical knowledge gap exists regarding their impact on fatigue cracking of
Fatigue asphalt binders, especially considering the aging processes. This study aims to comprehensively investigate the
Linear Amplitude Sweep
effects of recycling agents on the fatigue properties of asphalt binders, with a specific focus on the influence of
Performance Index
aging and rejuvenation. The current Linear Amplitude Sweep (LAS) test analysis procedure is utilized to assess
Recycling Agent
Aging the fatigue properties, and the efficacy of existing LAS test-based indices is evaluated. Furthermore, a novel index
Rejuvenation based on the simplified viscoelastic continuum damage (S-VECD) theory framework is proposed to monitor
changes in fatigue properties with aging and rejuvenation. A total of 26 recycled binder blends are studied and
the results demonstrate that the proposed index consistently tracks the effects of aging and rejuvenation in
recycled binder blends containing recycling agents. Additionally, cross-scale assessments are conducted by
correlating the proposed index with mixture performance, specifically the Indirect Tensile-Cracking Test
parameter (CTindex). The findings indicate that the proposed index exhibits a correlation with the CTindex after 1-
day Long Term Aging testing and effectively captures the aging sensitivity across different scales.

1. Introduction improve the properties of recycled asphalt binders. Although extensive

research has been conducted on RAs, there is a critical knowledge gap
The growing utilization of reclaimed asphalt pavement (RAP) in regarding the effects of these agents on fatigue cracking of asphalt
asphalt mixtures has gained significant attention due to its positive binders, especially in relation to aging processes [3].
environmental and economic impacts. Several efforts are being made to Fatigue cracking poses a significant challenge to asphalt pavements
increase the amount of recycled asphalt in new pavements [1,2]. as it negatively impacts ride quality, compromises structural integrity,
However, incorporating high amounts of recycled materials raises con­ and necessitates costly maintenance and rehabilitation efforts. The fa­
cerns about the durability and performance of asphalt pavements, tigue behavior of asphalt mixtures is influenced by various factors,
particularly regarding fatigue cracking and reduced serviceability [3,4]. including loading conditions, environmental factors, pavement struc­
To address these challenges, researchers have turned their focus to ture, and the properties of both the binder and aggregates. Among these
recycling agents (RAs), which are additives or modifiers designed to factors, the binder plays a critical role in determining the fatigue

* Corresponding author.
E-mail address: [email protected] (L.(G. Xue).
1
ORCID: 0000-0002-8098-1528
2
ORCID: 0000-0002-4417-1747
3
ORCID: 0000-0002-0097-9701
4
ORCID: 0000-0002-8996-0184
5
ORCID: 0000-0002-5915-0084
6
ORCID: 0000-0002-7223-3968
7
ORCID: 0000-0002-6216-3134
8
ORCID: 0000-0002-8534-6996

https://doi.org/10.1016/j.conbuildmat.2024.135929
Received 16 January 2024; Received in revised form 18 March 2024; Accepted 20 March 2024
Available online 1 April 2024
0950-0618/© 2024 Elsevier Ltd. All rights reserved.
S. Gulzar et al. Construction and Building Materials 425 (2024) 135929

resistance of asphalt mixtures. This fatigue resistance deteriorates as restore the low-temperature performance grade (PGL) of the recycled
binder ages. In fact, aging negatively influences the long-term perfor­ binder blend to − 22◦ C. The supplier of RA1 did not provide a recom­
mance of asphalt pavements, and understanding how RAs interact with mended dosage, therefore, 10% by weight of total binder was considered
binders under different aging conditions is crucial for effective for the blends prepared with RA1. The samples were prepared by pre­
implementation. heating the virgin and RAP binders to 140◦ C and 165◦ C respectively to
The Linear Amplitude Sweep (LAS) test has been proposed as an make them sufficiently fluid to be blended. The blending was done using
efficient test that relies on simplified viscoelastic continuum damage (S- a power drill with a paddle attachment for one minute at 1800 rpm and
VECD) theory to characterize fatigue properties of asphalt binders. the RA was incorporated using a pre-weighted syringe. The detail pro­
While the LAS test has emerged as a commonly used method for eval­ cedure describing the recycled binder blend preparation can be found
uating the fatigue properties of asphalt binders, recent studies have elsewhere [12]. The corresponding mixtures for three blends namely
revealed limitations associated with the current performance parame­ B1R1RA1, B1R1RA2, and B2R1RA5 were tested for three mixture aging
ters derived from this test [5]. These limitations especially include conditions such as STA, 1-day LTA, and 3-day LTA as discussed in the
assessing the impact of RA dosage on binder fatigue behavior and the methods subsection. Details about the mixture design and testing can be
effect of aging on the long-term performance of binders. Additionally, found elsewhere [13].
the fatigue properties of asphalt binders have been found to improve
with increased binder aging, contrary to initial expectations [6]. 2.2. Methods
While many indices based on LAS test have been proposed [7–9],
there is still no index that works universally in terms of capturing the The binders were subjected to short-term aging in the RTFO in
evolution of fatigue resistance with aging. Particularly very few studies accordance with AASHTO T 240–13 followed by the standard long-term
[10,11] have looked into capturing the effect of RAs on fatigue prop­ aging (referred to as PAV-20) conditioning as per AASHTO R 28–12. The
erties of asphalt binders. In light of these challenges, this study aims to extended long-term aging (referred to as PAV-40) used the same pro­
comprehensively investigate the effects of recycling agents on fatigue tocol that the PAV-20 conditioning with the difference that the total
properties of asphalt binders, particularly focusing on the influence of time in the PAV was 40 hours. The fatigue characteristics of the binders
aging and rejuvenation by employing a new LAS test-based index. By were characterized using linear amplitude sweep (LAS) test following
addressing the identified knowledge gaps, this research aims to the procedure of Wang et al. [14]. The LAS test includes two compo­
contribute to enhancing the sustainability and performance of asphalt nents. Firstly, a linear viscoelastic fingerprint test is run at very low
pavements incorporating recycled materials and RAs. strains and secondly, a linear amplitude sweep is run which is a strain
This study aims to use the current LAS test to evaluate the fatigue sweep from 0% to 30% at a frequency of 10 Hz. The LAS test tempera­
properties of asphalt binders containing RAP and RAs. Specifically, this ture was set to 16◦ C to ensure that the linear viscoelastic |G*| values at
paper; [1] analyzes fatigue properties using the current LAS test analysis the test frequency of 10 Hz fell within the range of 12–60 MPa to ensure
procedure, [2] evaluates the efficacy of existing LAS test-based indices, cohesive failure, as advised by Safaei and Castorena [15]. After each
and [3] proposes a novel index based on S-VECD theory framework to test, visual confirmation was conducted to ascertain the absence of ad­
track changes in fatigue properties with aging and upon rejuvenation. hesive failure between the binder and parallel plates, or any material
flow phenomena. Additionally, close attention was given to identify and
2. Materials and Methods monitor any experimental artifacts arising from such failures.
The mixture testing was performed for three aging conditions, (i)
2.1. Materials short-term aging the loose mixture for four hours at the compaction
temperature of 140.6◦ C (STA), (ii) long-term aging in the loose mixture
The experimental program for this study consisted of a partial condition for one day (1-day LTA) at 95◦ C after STA corresponding to
factorial arrangement involving three asphalt binder sources, three RAP roughly 4 years of field aging in VA as per NCHRP 09–54, and (iii) long
binder sources, and six RAs. Two PG 64S-22 virgin binders referred to term aging for three days (3-day LTA) in the loose mixture condition at
herein as B1 and B2, and one PG 58–28, named B3 have been used. The 95◦ C after STA corresponding to roughly 8 years of field aging in VA as
RAP binders extracted and recovered from three different sources per NCHRP 09–54 [16]. The STA aging was done according to the
referred to as R1 (PG 94–04), R2 (PG 106–04), and R3 (PG 94–10) were method described in AASHTO R 30–19, while the LTA conditioning
used. A total of six RAs (RA1 to RA6) were used to prepare the recycled followed the procedures described in NCHRP Project 09–54 where the
binder blends as shown in Table 1. RA1 was a paraffinic oil, RA2 was an mixtures are spread approximately one nominal maximum aggregate
aromatic extract, RA3 was a tall oil, RA4, RA5, and RA6 were tri­ size thick on a pan and the pan is rotated inside the oven daily. The
glycerides and fatty acids. A total of 26 recycled binder blends were indirect tensile cracking test (IDT-CT) as per ASTM D8225–19 was used
chosen for this study and the composition including RA dosage of these to evaluate the cracking resistance of mixtures where in the CTindex was
blends are shown in Fig. 1 along with the corresponding continuous high calculated using the Eq. [1].
temperature PG. It should be noted that the RA dosage was provided by ( ) ( )
Gf l75 t
the respective RA manufacturers and these dosages were selected to CTindex = × × (1)
|m75 | D 62
⃒ ⃒
⃒p85 − p65 ⃒
Table 1 m75 = ⃒⃒ ⃒ (2)
Properties of recycling agents used. l85 − l65 ⃒
Recycling Type Boiling Specific
where Gf is the total area under the load-displacement curve divided by
Agent point gravity
the product of the specimen, thickness [t] and diameter [D], m75 is the
RA1 Paraffinic oil 426◦ C 0.960 slope of interest expressed in Equation (2), p85 is the 85% of the peak
RA2 Aromatic extract 303–479◦ C 0.960
RA3 Tall oil 360◦ C 0.943
load (Pmax) at the post-peak stage, p75 is the 75% of Pmax at the post-peak
RA4 Triglycerides and fatty NA 0.930 stage, p65 is the 65% of Pmax at the post-peak stage, l85 is the displace­
acids ment corresponding to p85, l75 is the displacement corresponding to p75,
RA5 Triglycerides and fatty NA 0.950 l65 is the displacement corresponding to p65, D is the specimen diameter,
acids
and t is the specimen thickness.
RA6 Triglycerides and fatty NA 0.918
acids

2
S. Gulzar et al. Construction and Building Materials 425 (2024) 135929

Fig. 1. Binder blends containing RAs (a) RA dosage (%)1 and (b) continuous high temperature PG (◦ C). 1. The RBR for Source#1, Source#2, and Source #3 materials
were 0.35, 0.29, and 0.30 respectively.

3. Fatigue Life Estimation Using LAS Test where GR is an arbitrary reference modulus taken as 1 and γp is the peak
shear strain for a given cycle, |G*| for DMR are complex shear modulus
The damage evolution within the S-VECD model framework is based from fingerprint frequency sweep test of LAS and linear viscoelastic
on Schapery’s work potential theory [17] where the damage evolution modulus at that frequency and temperature. It can be seen from Eq. [6]
rate is expressed as shown in Eq. [3] that γ Rp is essentially equivalent to the linear viscoelastic stress response
dS
(
∂W R
)α for a given loading history.
= − (3) The fatigue life (i.e., Nf) at different strain levels for each binder
dt ∂S
blend is estimated as per Eq. [8] based on the S-VECD model and
where S the internal state variable (ISV) representing damage, α is the compared for both aging levels. The failure definition of maximum peak
material-dependent constant in its undamaged state, WR is the work store pseudo-strain energy as proposed by Wang et al. [14] was used for
performed, and t is the time. WR is expressed as shown in Eq. [4] fatigue life estimation.
( )− B
1 ( )2 Nf = A × γ p (8)
W R = C(S) γ R (4)
2
where Nf is estimated fatigue life or number of cycles to fatigue failure,
where C(S) is pseudo-stiffness expressed as shown in Eq. [5] γ p is the strain amplitude, A and B are modeling parameters based on the
τp S-VECD model and are given as shown in Eqs. (9) and (10)
C(S) = (5)
γRp × DMR ( )1− αC2 +α
f ⋅2α ⋅ Sf
A= ( )2α (9)
where τp is the peak shear stress and γRp is the peak pseudo-strain for a (1 − αC2 + α)(C1 C2 )α |G∗ |LVE
given cycle expressed as shown in Eq. [6], and DMR is the dynamic
modulus ratio expressed as shown in Eq. [7] B = − 2α (10)
1 where f is the loading frequency, equal to 10 Hz, α is the undamaged
γ Rp = γ × |G∗ |LVE (6)
GR p material-dependent constant, equal to 1+1/m where m is the slope of
LVE master curve, Sf is the damage at failure, C1 and C2 are best fit model
|G∗ |fingerprint
DMR = (7) parameters of damage characteristic curve, and |G*|LVE is the LVE dy­
|G∗ |LVE namic shear modulus.

3
S. Gulzar et al. Construction and Building Materials 425 (2024) 135929

While the recycled binder blends showed either similar or better through the LVE aging index property may not necessarily have poorer
fatigue life compared to the corresponding virgin binders, the Nf at a cracking performance compared to the binder that shows less sensi­
given strain level such as 5% was found to be higher at PAV-40 age level tivity. The reason for this occurrence may be that the failure mechanism
compared to PAV-20 age level for most cases. This observation implies of fatigue is not captured by LVE property alone and hence, the corre­
better fatigue resistance at the higher age level, which is counterintui­ lation cannot be established. However, it can be an effective index in
tive and averse to the collective understanding of the pavement com­ tracking the changes in LVE properties of the binder with aging and
munity. However, if a higher strain level such as 15% was chosen, the rejuvenation. From an aging perspective, the higher value in the LVE
expected trend was observed in some cases only. Chen et al. (2021) also aging index property implies a higher degree of aging as captured by
observed the same trends with age level in their study and recommend complex shear modulus and phase angle. From a rejuvenation
using higher strains to evaluate changes in fatigue properties with aging perspective, a blend with certain RA can show lower value compared to
[6]. Fig. 2 shows the fatigue life estimations at the two strain levels for reference binder, for example, B1R2RA4 as compared to B1. This result
both aging conditions. The RA1 binder blends (not shown in Fig. 2) due implies less sensitivity of this binder blend to aging, which may be
to very high Nf also showed opposite trend with aging. attributed to the presence of RA4 in addition to other properties such as
RAP binder properties, dosage level, %age of RAP binder used, etc. One
4. Aging Index Analysis way to look at the aging sensitivity is to compare the LVE aging index
parameter at PAV-20 and PAV-40 as per Eq. [11].
4.1. Linear viscoelasticity
[log(|G∗|/sin δ)]PAV− 40 − [log(|G∗|/sin δ)]PAV−
LVEagingsensitivity = 20
(11)
[log(|G∗|/sin δ)]PAV− 20
The complex shear modulus and phase data from the LVE fingerprint
test can be used to calculate the LVE aging index parameter, log(|G*|/ Fig. 4 shows the LVE aging sensitivity of the study binders and it can
sinδ), proposed by Elwardany et al. [18] and Mensching et al. [19]. This be seen that most the RA blends have a higher aging sensitivity in terms
LVE aging index parameter is calculated at the test temperature of 16◦ C of their LVE properties compared to the control binders. In the case of B1
and at a frequency of 10 Hz for both aging levels and shows consistent and B2, which have the same PG but are from different sources, the
trends with aging. The complex shear modulus is expected to increase aging sensitivity differs greatly. The RA blends generally have a higher
while phase angle is expected to decrease with aging, and LVE aging aging sensitivity than the virgin binders except B1R1RA4, B2R1RA5,
index parameter perfectly captures it as it increases with aging. Fig. 3 and B2R3RA3. RA3 blends show higher aging sensitivity as part of
shows the LVE aging index for all blends for both aging levels. B1R2RA3 and B2R1RA3, however, show much lower aging sensitivity in
The LVE aging index property is a reasonable parameter to capture case of B2R3RA3. This observation implies that several other factors
the changes in linear viscoelastic properties and yields consistent results, such as source of virgin binder, nature of RAP binder in addition to RA
meaning a higher log(|G*|/sinδ) for a more severe aging level. However, type and dosage play a part in determining the overall aging sensitivity
it must be noted that a binder that shows higher sensitivity to aging

Fig. 2. Fatigue life estimation for the studied binders at different strain levels of (a) 5% and (b) 15%.

4
S. Gulzar et al. Construction and Building Materials 425 (2024) 135929

Fig. 3. LVE aging index parameter for studied binders for PAV-20 and PAV-40 aging levels.

Fig. 4. LVE aging sensitivity of the studied binders.

of the binder blends. Zhang et al. [7] argued that a higher IR is preferred as it implies that
the material has higher ability to resist a loss in material integrity before
ultimate failure. As a result, it can be inferred that a stiffer or severely
4.2. Existing fatigue performance indices
aged binder will have a lower IR compared to a less-aged binder. It must
be noted that the failure definition used by the authors is defined at peak
Several aging indices have been proposed in the literature to capture
stress and it is used to calculate IR. Fig. 5 shows the schematic for the
the fatigue properties of asphalt binders using this linear amplitude
calculation of IR.
strain sweep. The failure definitions used in the below studies vary so
The binders in this study were evaluated for IR and it can be seen
does the point of failure where the performance indices are calculated.
been from Fig. 6 that the blends mostly have a higher IR compared to the
The focus of the current study is to evaluate the indices while
virgin binders at either PAV-20 or PAV-40. This finding suggests a
acknowledging the potential interaction effects that might occur from
higher fatigue resistance of these blends; however, this resistance de­
varying failure definitions in an LAS test. The study binders were eval­
creases with aging as expected and is captured by IR except for B2R1RA5
uated using the below indices and an attempt was made to evaluate their
where it seems to increase. This observation can be attributed to the way
trend with respect to aging.
the IR parameter is calculated to capture the fatigue resistance with
aging. It is interesting to note that B2R1RA5 has the highest dosage of all
4.2.1. Average reduction in integrity up to failure (IR)
RAs as shown in Fig. 1(a), which may affect its trends in fatigue per­
Zhang et al. [7] proposed a parameter, IR to evaluate the fatigue
formance with aging.
properties of different asphalt binders using the insights from mixture
failure criterion, DR proposed by Wang and Kim [20]. IR is defined as the
4.2.2. Strain tolerance up to failure (εT) and strain energy tolerance (Ef)
average reduction in integrity up to failure and is a measure of the
Zhang et al. [7] after having proposed IR argued that this parameter
degradation of fatigue resistance and is defined by Eq. [12].
did not consider the post-peak behavior of the stress-strain curve as data
∫ Nf
(1 − C)dN up to failure point (defined as peak stress) only was used in the calcu­
IR = 0 (12) lation of IR. They observed that the binder sample retains some load
Nf
carrying capacity after peak stress and noted flatter and longer post-peak
where Nf is the number of load cycles till failure and C is the pseudo- stress strain curves for polymer modified binders over unmodified
stiffness. binders which could be attributed to the cross-linkages of the polymer

5
S. Gulzar et al. Construction and Building Materials 425 (2024) 135929

Fig. 5. Schematic for the calculation of IR.

Fig. 6. Average reduction in integrity up to failure (IR) for PAV-20 and PAV-40 aging conditions.

networks even after peak stress point. In order to capture this post-peak low variability and the ability to capture the effect of different modifiers.
behavior, the authors proposed strain εT as a parameter and defined it as Additionally, the authors defined strain energy tolerance (Ef) as the area
the strain tolerance up to failure. The end point for calculation of the under the stress-strain curve up to the end point for fatigue tolerance
strain tolerance was taken as 25% of the peak stress post-peak given its limit, that is, 25% of the peak stress. Both the fatigue tolerance up to

Fig. 7. Schematic for the calculation of εT and Ef.

6
S. Gulzar et al. Construction and Building Materials 425 (2024) 135929

failure and strain energy tolerance are schematically shown in Fig. 7. A modulus at 50% peak shear stress often tends to lead to higher suscep­
higher strain tolerance corresponds to higher fatigue cracking resistance tibility to cracking. The components of FREI are shown schematically in
and this parameter is expected to decrease with aging. Similarly, a Fig. 10. In general, a higher FREI is considered to correspond to better
higher value of Ef is expected to yield better fatigue resistance and it is cracking resistance and it is expected to decrease with aging.
expected to decrease with increasing aging level. FREI was calculated for the study binders for both PAV-20 and PAV-
The strain tolerance was calculated for the study binders, and it was 40 aging levels as per Eq. [13] and the results are shown in Fig. 11. It can
found that this parameter did not follow the expected trend with respect be seen that B1- and B2-based binder blends with RAs show higher FREI
to age level for all binders except B1R2RA2. This observation is evident compared to the two control binders B1 and B2 as opposed to B3-based
in Fig. 8 where the parameter is calculated and plotted at PAV-40 versus binder blends. Furthermore, FREI decreases with aging as expected for
PAV-20 aging conditions. It can be seen only one blend, which in this all binders except B2R1RA5 where it increases slightly. This single case
case is B1R2RA2 shows a decrease in strain tolerance (post-peak) with outlier can again be viewed from a dosage perspective as B2R1RA5 has
aging. the highest dosage of an RA across all binder blends. Thus, in addition to
The study binders were evaluated for Ef and it was found that this the factors such as type of RAP binder, type of RA, source of virgin
parameter, similar to the strain tolerance may not give a consistent trend binder, etc. FREI seems to be affected by RA dosage especially higher
with aging. Most of the binder blends did not follow the expected trend dosages close to 10%, which was recommended by NCHRP Project
except B1R2RA2, B1R2RA4, and B2R1RA5 can be seen in Fig. 9. It has 09–58 as the upper limit for RA dosage in USA [4, 12, 21].
been seen that the binder response for the post-peak behavior can show
different ranges in terms of stress-strain curve and calculation of the 4.2.4. Fatigue factor of binder (FFB)
inflection or end points as a percentage of peak load or peak stress can In 2014, a large database of asphalt binder data was collected by
occur at significantly different percentiles. For example, in case of Petrobras Research Center in Brazil and after analyzing the fatigue
B1R1RA1, strain tolerance (post peak) could not be even calculated behavior of asphalt mixtures and binders, a significant correlation was
because the test stopped before the stress response could reach 25% of observed by means of the so-called fatigue factor of the mixture (FFM)
peak stress. and fatigue factor of the binder (FFB) [22]. Later in 2021, Possebon [23]
classified the asphalt binders based on FFB, however, the range of FFB
4.2.3. Fatigue resistance energy index used appears to be very narrow and may not encompass all the binders.
Fatigue resistance energy index (FREI) is an index parameter pro­ Overall, FFB is an empirical index and is calculated as shown in Eq. [14],
posed by Zhou et al. [9] using the pre-peak stress-strain curve of the LAS [ ( ) ( )]
log Nf ,1.25% + log Nf ,2.50%
test. The components of the parameter rely solely on the second part of FFB = × (log(0.025) − log(0.0125)) (14)
LAS test, so the LVE fingerprint test of the LAS test is not needed to 2
obtain this index. It is based on fracture mechanics principles as opposed
where Nf,1.25% is the fatigue life estimated at 1.25% strain and Nf,2.5% is
to continuum damage mechanics and the authors use phenomenological
the fatigue life estimated at 2.5% strain. The failure definition used in
insights from a time sweep test at a constant shear strain amplitude to
the test to estimate Nf is the peak stress condition.
develop FREI for LAS test. FREI is calculated using Eq. [13]
A higher value of FFB is related to better fatigue performance and it is
Jf − τmax ( )2 expected to decrease with aging [24, 25]. In this study, FFB was calcu­
FREI = γ (13)
G0.5τmax 0.5τmax lated and analyzed for the study binders for the two aging conditions. It
was found that all the binders showed excellent fatigue performance
where Jf-τmax is the shear fracture energy calculated until the peak shear (FFB > 1.57) as per Possebon’s classification, however, the FFB did not
stress, G0.5τmax is the apparent shear modulus calculated at 50% of peak show the expected poorer performance at the PAV-40 compared to
shear stress in the pre-peak response, and γ 0.5τmax is the shear strain at PAV-20 age level for almost all blends. It can be observed from Fig. 12
50% of peak shear stress. that FFB for PAV-40 is higher than that for PAV-20 for all the blends
The phenomenological understanding of FREI can be had by except B1R1RA2 and B1R2RA4 similar to Ef aging sensitivity. The pri­
observing three aspects of binder response under the strain sweep part of mary reason for a counter-intuitive trend of FFB with aging can be
the LAS test. Firstly, a material with higher fracture energy is normally attributed to the use of fatigue life (Nf) estimations at comparatively
expected to have better fatigue resistance. Secondly, larger shear strain strain levels such 1.25% and 2.5% which are known not to follow the
at 50% of peak shear stress means better flexibility and relaxation ca­ expected trend with aging as discussed above in fatigue life estimation
pacity when the binder reaches half its maximum shear loading capac­ section.
ity. And thirdly, a larger shear modulus, in this case, the apparent shear

Fig. 8. Strain tolerance up to failure (post peak) for PAV-20 and PAV-40 aging conditions.

7
S. Gulzar et al. Construction and Building Materials 425 (2024) 135929

Fig. 9. Strain energy tolerance for PAV-20 and PAV-40 aging conditions.

Fig. 10. Schematic for the calculation of FREI.

Fig. 11. Fatigue energy resistance index (FREI) for PAV-20 and PAV-40 aging conditions.

4.2.5. Reduction in shear modulus until peak shear stress (Δ|G*|peak τ) shear modulus for the second part of the LAS test and tracked it for the
Garcia Mainieri et al. [8] proposed an alternative method to analyze entire strain sweep. The authors argued that the reduction in material
the LAS test results to derive a practical parameter that relates more integrity from the start of the test to the peak stress condition relates to
directly to the loss of material integrity. They used an apparent dynamic the binder’s ability to tolerate damage without losing its capacity in

8
S. Gulzar et al. Construction and Building Materials 425 (2024) 135929

Fig. 12. Comparison of FFB for the studied binders at PAV-20 and PAV-40 aging conditions.

terms of shear resistance. In order to calculate Δ|G*|peak τ, only pre-peak

Table 2
stress-strain response is used, and it is calculated using Eq. [15]. A
Summary of LAS test-based indices.
higher value of Δ|G*|peak τ is implies better ability of the binder to
tolerate greater loss in material integrity before its capacity to resist LAS test-based Definition/Equation Reference
Index
deformation decreases post peak shear stress. With aging, Δ|G*|peak τ is
expected to decrease [26]. Average reduction ∫ Nf
0 (1 − C)dN [27, 28]
in integrity up to IR =
[ ∗ ] Nf
|G |start − |G∗ |peakτ failure (IR)
Δ|G∗ |peakτ = ∗
× 100 (15) Strain tolerance up Strain level corresponding with the 25% [27, 28]
|G |start
to failure (εT) maximum stress point
The studied binders were evaluated using Δ|G*|peak τ for both PAV- Strain energy Area under the stress–strain curve up to the [27, 28]
tolerance (Ef) 25% maximum stress point
20 and PAV-40 aging conditions as per Eq. [15]. It can be seen from Jf− τmax
Fatigue resistance
Fig. 13 that most of binder blends have a higher or similar Δ|G*|peak τ for FREI = (γ )2 [9]
energy index G0.5τmax 0.5τmax
both aging conditions except for B3-based binder blends which show (FREI)
[ ]
lower Δ|G*|peak τ compared to B3. Further, all binders show a decrease in Fatigue factor of
FFB =
log(Nf,1.25% ) + log(Nf,2.50% )
[22, 23]
binder (FFB) 2
Δ|G*|peak τ with aging except B2R1RA5 which is similar to the obser­
vations made from previously presented indices such as FREI and IR. The
× (log(0.025) − log(0.0125))
Δ|G*|peak Percent reduction in |G*| measured from the [8, 26]
highest aging sensitivity in terms of Δ|G*|peak τ was shown by B1R2RA2
τ
start of the test until the peak shear-stress
while the least was shown by B1 binder. condition

Note: Refer to cited references for variable definitions.

4.2.6. Summary of existing indices
The indices discussed above have been summarized in Table 2 and
5. Novel Fatigue Performance Index Based on Pseudo-Strain
their significance with respect to fatigue performance, aging, and reju­
Energy Principles
venation is presented in Table 3.

Most indices evaluated in the previous section consider different

aspects of the stress-strain response under a strain-controlled linear

Fig. 13. Δ|G*|peak τ for PAV-20 and PAV-40 aging conditions.

9
S. Gulzar et al. Construction and Building Materials 425 (2024) 135929

Table 3 can yield fundamental relationship for full fatigue characterization of an

Significance of LAS test-based indices. asphalt binder. This finding has further been established by Wang and
LAS test-based Index Significance w.r.t. Kim [20] who proposed pseudo-strain energy-based fatigue failure cri­
terion for asphalt mixtures that required less overall testing for full
Fatigue Aging Rejuvenation
characterization. It must be noted that even a PSE-based failure defini­
Average reduction in The higher Should Should increase upon tion may yield counterintuitive trend in terms of fatigue life estimation
integrity up to failure the better decrease with rejuvenation
(IR) aging
(Nf) with aging. Thus, highlighting the need for an alternate measure of
Strain tolerance up to The higher Should Should increase upon fatigue resistance. In the next section, PSE-based analysis is used to
failure (εT) the better decrease with rejuvenation evaluate a novel index that can capture fatigue performance across
aging different binders for different aging conditions.
Strain energy tolerance The higher Should Should increase upon
(Ef) the better decrease with rejuvenation
aging 5.1. Pseudo-strain energy (PSE)-based analysis
Fatigue resistance The higher Should Should increase upon
energy index (FREI) the better decrease with rejuvenation
aging Fig. 14 shows a schematic of a typical LAS test that also defines the
Fatigue factor of binder The higher Should Should increase upon PSE related terms. The undamaged line presents the linear viscoelastic
(FFB) the better decrease with rejuvenation response of the binder, which serves as a reference for the condition
aging where no damage takes place with increasing loading. The blue dotted
The higher Should Should increase upon
data series represents a typical measured response, where the measured
Δ|G*|peak τ
the better decrease with rejuvenation
aging response deviates from the ‘undamaged line’ as loading progresses and
damage occurs.
From the LAS test, the stored and released PSE can be computed for
amplitude sweep. These indices can be broadly grouped into five cate­ each cycle as shown in Fig. 14.The stored PSE, WRs , represents PSE that is
gories. The first category consists of those indices that consider some stored in each loading cycle and is calculated using Eq. [16].
representation of material integrity and its degradation such as IR and Δ|
( )2
G*|peak τ, where both of these consider peak stress as occurrence of W RS =
(1/2) × τp × γ Rp 1
= × C × γRp (16)
failure. The second category consists of those indices that are based on DMR 2
energy principles such as Ef. The third category includes the indices that
where the variable definitions are the same as those defined earlier in
are based on strain tolerance such as εT. The fourth category consists of
Section 3.
indices such as FREI that integrate the energy principles, stiffness, and
An increasing WRs means that the material retains the ability to store
ductility aspects in a single parameter. Finally, there are empirical
additional energy as the loading amplitude (consequently the energy
indices such FFB that are based on observations and correlations from
input) increases. On the other hand, a decrease in WRs means that the
large datasets employing fatigue life estimations across binders and
material is losing its ability to store PSE as the loading input increases,
mixtures. Regardless of the category considered, the existing indices do
thus showing that failure has occurred. Based on this energy-based un­
not yield a consistent trend with respect to aging level. This observation
derstanding, Wang et al. [14] proposed peak of stored PSE as the
highlights the need for an alternate parameter.
occurrence of failure in an LAS test instead of peak stress. They verified
In this work, viscoelastic continuum damage (VECD) theory has been
the fatigue life based on this failure definition and drop in phase angle
used as a basis to develop a new fatigue performance index. The VECD
data and found a good agreement indicating peak in stored PSE is a
theory uses pseudo variables such as pseudo strain, pseudo strain en­
reliable indicator of defining fatigue failure in an LAS test. The current
ergy, etc. and elastic-viscoelastic correspondence principle with damage
study builds on this energy-based failure definition to propose a novel
evolution law to model fatigue performance. The failure criterion used
index that captures the ability of binders to withstand fatigue cracking in
along with VECD theory to predict failure has evolved over time (for
terms of stress level and strain tolerance at peak pseudo-strain energy.
both asphalt binders and mixtures). With respect to asphalt binders,
The experimental results demonstrate that the stored PSE shows a
Wang et al. [14] showed that pseudo-strain energy (PSE) based analysis
peak within the current LAS test framework [12]. This peak occurs after
yields more consistent fatigue failure across binders and mixtures, and
the peak stress has occurred as shown in Fig. 15 indicating that the

Fig. 14. LAS-based PSE definitions.

10
S. Gulzar et al. Construction and Building Materials 425 (2024) 135929

Fig. 15. Schematic for; (a) peak stored PSE and (b) the corresponding shear stress.

binder continues to retain its ability to store additional energy after peak dissipation, offering insights into the asphalt binder’s resistance to fa­
stress condition. This behaviour may also imply an apparent loss in tigue, aging, and rejuvenation. Another important aspect to note is that
material integrity up to peak stress is due to the nonlinear viscoelasticity SL index integrates the shear stress at the peak stored pseudo-strain
while damage becomes dominant after the peak stress. energy, which serves as a representation of the material’s shear capac­
As can be seen from Eqs. (5), (6), and [16], the stored PSE is ity until failure. A higher SL value indicates an elevated shear capacity,
dependent on shear stress along with time, which is specific for a signifying superior resistance to fatigue failure. In other words, SL acts as
particular loading scheme. So, the shear stress at the peak stored PSE can a metric for the material’s capability to continue resisting fatigue failure
be considered as metric to compare the shear capacity until failure. In even after reaching conditions of peak stress. As discussed before, SL is
other words, the ability to the binder to continue resisting fatigue failure expected to decrease with aging, mirroring alterations in the material’s
even after the peak stress is reached. In order to obtain a normalized properties over time. Aging may lead to a decrease in the material’s
parameter, the ratio of stress at peak stored PSE to the peak stress termed capacity to withstand fatigue, potentially attributed to factors like
as stress level at peak stored PSE (SL) is taken as an index parameter to increased stiffness, embrittlement, or other aging-related effects. SL al­
characterize the fatigue resistance of asphalt binders and is calculated lows to monitor the impact of aging on the material’s integrity and
using Eq. [17]. A higher value SL corresponds to higher fatigue resis­ performance during cyclic loading.
tance capacity. It is expected to decrease with aging. In this paper, SL has been used to evaluate different binder blends for
( ) different aging conditions, recycling agents, etc. In general, binders with
τpeakstoredPSE
SL = × 100 (17) higher SL are inferred to possess superior fatigue resistance, whereas
τpeakstress
those with lower SL may be more prone to fatigue cracking. And changes
The SL index offers some valuable insights into the fatigue perfor­ in the relative rankings of binders under different aging conditions may
mance of asphalt binders especially with regards to energy dissipation offer insights into the binders’ sensitivity to aging effects, including the
and damage accumulation. Since it originates from the concept of stored intricate interactions among various constituents, including base
pseudo-strain energy a diminishing SL value suggests a reduction in the binders, RAP binders, and RA types and dosages.
material’s ability to store pseudo-strain energy, implying heightened The study binders were evaluated using SL for both aging PAV-20 and
energy dissipation and potential accumulation of damage. SL implicitly PAV-40 aging conditions, and it was found that SL for the virgin binders
captures the delicate equilibrium between energy storage and B1, B2, and B3 is higher than those of most of the corresponding binder

11
S. Gulzar et al. Construction and Building Materials 425 (2024) 135929

blends containing RAs. The RA5 blends which generally have a higher owning to a softer grade ranks high in terms of SL.
RA dosage show higher SL as compared to other binder blends. RA1 The relative ranking of the studied binders in terms of SL is tabulated
which is reported to have negative impact of cracking performance in Table 4 for PAV-20 and PAV-40 aging conditions. It can be seen that
shows low SL across aging levels. SL seems to capture the impact of aging relative ranking changes from PAV-20 to PAV-40 aging condition,
consistently across all study binders and decreases with aging. Amongst however, as discussed above, most of the best and the worst binder
the RA blends, RA1 blends show highest sensitivity to aging with SL blends in terms of fatigue performance retain their relative rankings.
decreasing by about 14% and 28% for B1R1RA1 and B2R3RA1 blends The change in the rankings can be attributed to varying sensitivity to
respectively. RA1 is REOB-based which has been reported to increase aging of the base binders, RAP binders, RA types, source of respective
aging and cracking susceptibility [29]. binders, the interaction of these three constituents in the binder blends,
In terms of ranking the binders at respective age levels, RA1 blends RA dosage, etc. among several other factors.
show worst fatigue performance for both PAV-20 and PAV-40 aging An important aspect of the comparison between PAV-20 and PAV-40
level while most of RA5 blends show the best fatigue performance for was examining if the two age levels provide similar or distinct insight
both PAV-20 and PAV-40 aging levels in terms of SL. Binder B3 which is regarding the relative performance of the study binders and blends in
a PG 58–28 which is expected to show better fatigue performance terms of SL. The was achieved by performing the Spearman’s rank

Table 4
Relative ranking of binder blends based on SL for PAV-20 and PAV-40 aging conditions.

Note: The color palette corresponds to the relative ranking of the binder blends based on SL, with green corresponding to the highest rank [1] and red corresponding to
the lowest rank [28]. The color of each cell corresponds to the rank entry in that cell.

12
S. Gulzar et al. Construction and Building Materials 425 (2024) 135929

correlation analysis. In this analysis, the rank for each blend was Department of Transportations (DOTs) including Virginia DOT employs
compared to between the two aging conditions using Eq. [18] as shown PG 64S-22 asphalt binder as a standard when recycled asphalt pavement
below. The Spearman’s rank correlation coefficient ranges from nega­ (RAP) is not utilized. As an example, the blend B2R1RA5 was compared
tive one (-1) to one (+1) with negative one (-1) implying a negative to the binder B1, where B1 served as the control binder for evaluation.
relationship, one (+1) implying a positive relationship, and values Additionally, two reference blends that consisted of a softer binder and
closer to zero implying no correlation. This ranking analysis was per­ RAP binder without RA which is again a typical practice in several DOTs
formed to evaluate if the two aging conditions provide similar infor­ that do not use RAs. These reference blends consisted of a PG 58–28
mation and evaluate if either is potentially redundant while also virgin binder and a recycled binder devoid of RA. The objective was to
providing additional insights to determine how additives are affecting determine the level of concordance or divergence between the recycled
the performance of a given binder blend. binder blends containing RAs and both the PG 64S-22 virgin binders (B1
∑ and B2) and the prevailing practices observed in several state DOTs that
6 d2i
ρ = 1− (18) do not use RA but a softer PG binder, as exemplified by the reference
n(n2 − 1) blends (B3R1 and B3R3). Table 5 shows the two sets of binders that were
evaluated for statistical similarity in terms of SL.
where ρ is the Spearman’s rank correlation coefficient, di is the differ­
The Set 1 binders did not pass Bartlett’s test for equal variance for
ence between the two ranks of each observation, and n is the number of
PAV-20 aging condition and the binders with standard deviation were
observations.
removed which entailed removing B1 binder as well. As a result, the
The Spearman’s rank correlation coefficient was found to be 0.76
statistical comparison could not be made for PAV-20 aging condition for
indicating that SL results are positively correlated at the two age levels.
Set 1 binders. However, Set 1 passed Bartlett’s test for equal variance for
This result essentially means that the two aging conditions generally
PAV-40 aging condition while Set 2 passed Bartlett’s test for equal
provide mostly similar insight regarding the relative performance of the
variance for both PAV-20 and PAV-40 aging conditions. Subsequently,
study binders and blends in terms of SL. However, the ranking order is
Dunnett’s test was conducted on these datasets that passed Bartlett’s test
not absolutely similar as some RA blends showed better while some
for equal variance. Color coding with the graphical presentation of re­
showed worse performance with increased aging condition of PAV-40
sults was used to convey Dunnett’s test results. Yellow indicates that the
than PAV-20.
binder blend has statistically similar performance as that of control
binder in terms of SL. Red or green indicate worse or better performance
5.2. Statistical analysis to assess rejuvenation effectiveness of RAs than the control blend respectively. A binder blend is considered better
if it has a statistically different result based on Dunnett’s test and is in the
The statistical analysis for the SL index involved the use of Bartlett’s desirable (higher) direction of the expected trend of SL. Similarly, a
test for verification of equal variance assumption and Dunnett’s test for binder blend is considered better if it has a statistically different result
subsequent comparison. Dunnett’s test is employed to compare multiple based on Dunnett’s test and is in the undesirable (lower) direction of the
treatment groups to a control group or reference group. In this analysis, expected trend of SL.
Dunnett’s test was used to determine if a binder blend containing RA Fig. 17, Fig. 18, and Fig. 19 show the Dunnett’s test results for Set 1
(treatment group) was statistically equal to the target binders B1 and B2 PAV-40 aging condition, Set 2 PAV-20 age condition, and Set 2 PAV-40
(control group) at their respective age levels. The goal was to assess the aging condition, respectively. It can be seen that all of the RAs in either
performance similarity of the binder blends in comparison with the B1 set are able to restore the fatigue performance of the recycled binder
or B2 binders to determine how effective RAs were in terms of SL. For the blends to that of the control binder across PAV-20 and PAV-40 aging
purpose of this analysis, it was assumed that the data followed a normal conditions except RA1 and RA2 in some cases. It is interesting to note
distribution. This assumption is generally valid when the sample size that the reference blend (without RA) B3R1 performs worse across both
exceeds 30, which was the case in this analysis considering replicates in sets and aging levels while B3R3 performs worse in case of PAV-20 for
the dataset. Dunnett’s test assumes equal variance which needs to be Set 2 binders compared to the control binders. This observation is an
verified first before running this test. This assumption of equal variance important finding, which suggests that it can be favorable to use RAs
was verified by performing Bartlett’s test at a specific age level. Bar­ since they can achieve better fatigue characteristics over using a softer
tlett’s test examines whether the variances of the groups being binder without an RA to restore the recycled binder system.
compared are equal. The significance level for all tests were taken as Additionally, RA1 blends such as B1R1RA1 and B2R3RA1 perform
0.05. If the corresponding p-value from Bartlett’s test was greater than
0.05, it meant that the variances could be assumed as equal. In such
Table 5
cases, Dunnet’s test was applied. If Bartlett’s test resulted in a p-value Dataset used for statistical analysis.
less than 0.05, it meant the variances were not equal among the binder
Category Set 1 Set 2
blends for that specific age level. To address this, the standard deviations
of the blends were compared. The blends with the highest standard Control B1 B2
Recycled binder blends containing RA B1R1RA1 B2R1RA3
deviation (indicating more variability) were removed. This step aimed
B1R1RA2 B2R1RA5
to improve Bartlett’s test score and make it above 0.05, thus allowing B1R1RA3 B2R1RA6
the assumption of equal variances for subsequent Dunnett’s test. Finally, B1R1RA4 B2R2RA4
based on the means obtained for each binder blend in Dunnett’s test, the B1R2RA2 B2R2RA5
binders were classified as equal, better, or worse performing in relation B1R2RA3 B2R3RA1
B1R2RA4 B2R3RA2
to their respective target binders for each age level. This classification B1R2RA5 B2R3RA3
helped determine how each binder blends compared to the target B1R2RA6 B2R1RA3
binders as well as how effective RAs were in restoring the fatigue B1R3RA2 B2R1RA5
characteristics of the recycled binder blend at each age level. B1R3RA3 B2R1RA6
B1R3RA4
It must be noted that the recycled binder blends were subjected to a
B1R3RA5
comparative analysis with respect to the PG 64S-22 virgin binder B1R3RA6
component (B1 or B2) present within the respective blends. The inclu­ B1
sion of the PG 64S-22 virgin binder within each specific reclaimed B1R1RA1
asphalt (RA) blend served as a control to represent the desired binder B1R1RA2
Reference blends (without RA) B3R1 and B3R3 B3R1 and B3R3
properties. This selection was based on the fact that the several

13
S. Gulzar et al. Construction and Building Materials 425 (2024) 135929

Fig. 16. SL(%) for PAV-20 and PAV-40 aging conditions.

Fig. 17. Statistical comparison of Set 1 binders for PAV-40 aging condition. Colors convey Dunnett’s test results: green, yellow, and red indicate better, equal, and
worse results relative to B1 respectively.

Fig. 18. Statistical comparison of Set 2 binders for PAV-20 aging condition. Colors convey Dunnett’s test results: green, yellow, and red indicate better, equal, and
worse results relative to B2 respectively.

14
S. Gulzar et al. Construction and Building Materials 425 (2024) 135929

Fig. 19. Statistical comparison of Set 2 binders for PAV-40 aging condition. Colors convey Dunnett’s test results: green, yellow, and red indicate better, equal, and
worse results relative to B2 respectively.

worse than the control in both sets. RA1 is a REOB-based recycling agent PG of − 22◦ C for all blends and perhaps also yields similar fatigue
which is known to have problematic performance in terms of cracking characteristics. Additionally, this could imply that the SL may fail to
and this observation of worse performance of RA1 blends across the two capture the sensitivity of the binder blends to different RAs. This
aging levels in terms of SL provides additional evidence to it. This observation warrants further research to evaluate SL for blends with
observation also provides evidence to the efficacy of SL parameter to different RAs at different dosages and/or from different dosage selection
screen worse performing RAs in terms of fatigue performance. RA2 procedures.
blends such as B1R2RA2 and B1R3RA2 also show worse performance
than the control B1 binder. As can be seen from Fig. 1(a), RA2 has the 5.3. Comparison with LVE aging index parameter and validation with
lowest dosages amongst all RAs and the poorer fatigue performance of mixture aging
RA2 binder blends can be attributed to low RA dosage. Additionally,
RA5 blends which have ranked highest in terms of fatigue performance As was shown previously, the LVE aging index parameter (log(|G*|/
as showcased in the previous sections, B2R2RA5 shows better perfor­ sinδ)) shows a positive trend with aging and has proven useful in
mance than the control for Set 2 binders for PAV-20 condition as well. In assessing the aging sensitivity of binders in terms of their linear visco­
summary, it can be noted that recycled binder blends containing RAs are elastic properties. The SL index, which is a fatigue performance index
able to achieve similar fatigue performance as that of the corresponding based on pseudo-strain energy principles also shows a consistent decay
control binder showcasing rejuvenation effectiveness of RAs except with aging. Fig. 20 shows a two-dimensional plot which considers SL
RA1, which shows worse performance. Further, reference blends index along with log(|G*|/sinδ) for the two aging conditions of PAV-20
without RAs are also not able to reach similar fatigue performance levels and PAV-40. It can be seen that SL index tracks consistently with log(|
as that of control binders in some cases. It must be noted that most of the G*|/sinδ) with aging. A higher SL corresponds to a lower log(|G*|/sinδ)
binder blends were deemed to have statistically equal SL results to the at PAV-20 compared to PAV-40 where a lower SL corresponds to a higher
respective control binders. This result can be due to the dosage selection log(|G*|/sinδ). This observation provides supportive evidence for SL
procedure of RAs which was done to achieve a target lower temperature index to be used as fatigue performance index based on LAS test.

Fig. 20. Comparison of SL with LVE aging index parameter for all evaluated binders under PAV-20 and PAV-40 aging conditions.

15
S. Gulzar et al. Construction and Building Materials 425 (2024) 135929

A subset of the binder blends was shortlisted, and corresponding

mixtures were fabricated and tested for IDT-CT test. The details of the
mixture testing can be found elsewhere [13]. In this section, the focus is
on three mixtures namely B1R1RA1, B1R1RA2, and B2R1RA5, that were
tested for three mixture aging conditions such as STA, 1-day LTA, and
3-day LTA as discussed in the methods section. The CTindex of these three
mixtures was compared across the three aging conditions and an aging
sensitivity in terms of CTindex was calculated using the Eqs. (13) and (14).
[ ]
(CTindex )STA − (CTindex )1DLTA
(CTindex )1DSTA
agingsensitivity = × 100 (19)
(CTindex )STA

In terms of binder fatigue performance index, SL was compared

across the two aging conditions of PAV-20 and PAV-40, and a similar
aging sensitivity was calculated using Eq. [15].
[ ]
(SL )PAV− 20 − (SL )PAV− 40
(SL )agingsensitivity = × 100 (20) Fig. 22. Aging sensitivity across binder and mixture scales for B1R1RA1,
(SL )PAV− 20
B1R1RA2, and B2R1RA5 in terms of LVE aging index parameter and CTindex.
As noted from the aging assessment of mixtures, the CTindex shows a
significant discrimination of mixtures after 1 day LTA over 3 day LTA, 6. Conclusions
and the aging sensitivity for 1 day LTA captures that change in the
CTindex as shown in Eq. [13]. The aging sensitivity in terms of SL for the On the basis of the results described in this paper, the following
corresponding binder blends was calculated and compared with mixture conclusions are reached:
aging sensitivity after 1-day LTA. The basic idea was to verify if the SL
index is able to capture the degradation in cracking performance with 1. While several existing LAS based indices were evaluated in this
aging as reflected from the mixture testing. Fig. 21 shows the compar­ study, none of the these was able to consistently track across the two
ison across binder and mixture scales and it can be seen that there is a aging levels of PAV-20 and PAV-40 for all the evaluated binders.
good agreement between the binder and mixture aging sensitivities as Some of the indices yielded significant counter-intuitive results while
captured by SL and CTindex respectively. A similar comparison was done few yielded consistent trends with one or two outliers.
between the CTindex and LVE aging sensitivities as shown in Fig. 22. It 2. A novel pseudo-strain energy-based index parameter, SL was pro­
can be seen that even though LVE aging index parameter correlates well posed as an index to capture the fatigue characteristics of recycled
with SL, it is not able to capture the mixture aging sensitivity in terms of binder blends. It was shown that SL tracks consistently with aging
CTindex. This observation further motivates the need of the SL parameter. and provides insights about the aging and rejuvenation impacts of
Further, if these three binder blends were to be ranked in terms of SL using RAs in recycled binder blends.
for PAV-20 and PAV-40 aging conditions, the order would be 3. The cross-scale assessments carried out with mixture performance in
B2R1RA5>B1R1RA2>B1R1RA1. It must be noted that B1R1RA1 was terms of IDT-CT found that SL parameter correlates with CTindex after
shown to be the worst across all tested binders and B1R1RA2 also ranked 1-day LTA and is able to capture the aging sensitivity across the two
very low, while B2R1RA5 ranked at or close to the top among the best scales. However, it should be noted that a very limited number of
performing binder blends. On the other hand, at mixture scale in terms mixtures were used to compare the fatigue performance across the
of CTindex, the relative ranking at STA and 3 day LTA is two scales, so the efficacy of the SL index still needs to be validated
B2R1RA5>B1R1RA1>B1R1RA2 while the order at 1D LTA is same as with a larger set of mixture testing.
observed from binder SL assessment across PAV-20 and PAV-40 aging
conditions. This observation also suggests looking more closely at SL as a While SL was able to consistently track with aging and capture the
possible index for assessing fatigue performance of binder blends that best and worst performing binders, its sensitivity to assess the rejuve­
may translate consistently to the mixture scale in terms of IDT-CT nation capacity of different types of RAs was lower. Future work with a
testing. larger pool of binders and corresponding mixtures is needed to propose
any screening of RAs, detecting modifiers, or capturing the fatigue
performance improvement due to polymers based on SL index.

CRediT authorship contribution statement

Ilker Boz: Writing – review & editing, Project administration,

Funding acquisition. Jhony Habbouche: Writing – review & editing,
Project administration, Funding acquisition. Saqib Gulzar: Writing –
review & editing, Writing – original draft, Visualization, Methodology,
Investigation, Formal analysis, Data curation, Conceptualization. Lei
Xue: Writing – review & editing, Visualization, Methodology, Investi­
gation, Formal analysis. Cassie Castorena: Writing – review & editing,
Supervision, Methodology, Investigation, Conceptualization. Shane
Underwood: Writing – review & editing, Supervision, Methodology,
Investigation. Jaime Preciado: Writing – review & editing, Resources.
Andrew Fried: Writing – review & editing, Resources.

Declaration of Competing Interest

Fig. 21. Aging sensitivity across binder and mixture scales for B1R1RA1,
B1R1RA2, and B2R1RA5 in terms of SL and CTindex. The authors declare that they have no known competing financial

16
S. Gulzar et al. Construction and Building Materials 425 (2024) 135929

interests or personal relationships that could have appeared to influence the high-temperature climatic performance grade, Transp. Res. Record: J. Transp.
Res. Board 2676 (6) (2022) 302–314, https://doi.org/10.1177/
the work reported in this paper.
03611981221074361.
[13] J. Preciado, S. Gulzar, C. Castorena, S. Underwood, J. Habbouche, and, I. Boz,
Data Availability Effect of recycling agents on asphalt mixture performance (2024).
[14] C. Wang, C. Castorena, J. Zhang, and, Y.R. Kim, Unified failure criterion for asphalt
binder under cyclic fatigue loading, Asphalt Paving Technology: Association of
Data will be made available on request. Asphalt Paving Technologists-Proceedings of the Technical Sessions 84 (2015)
269–300, https://doi.org/10.1080/14680629.2015.1077010.
Acknowledgements [15] F. Safaei, and, C. Castorena, Temperature effects of linear amplitude sweep testing
and analysis, Transp. Res. Record: J. Transp. Res. 2574 (2016) 92–100, https://doi.
org/10.3141/2574-10.
This research was supported by the Virginia Transportation Research [16] Y.R. Kim, C. Castorena, N.F. Saleh, E. Braswell, M. Elwardany, and, F.Y. Rad, Long-
Council and Virginia Department of Transportation under Grant UPC Term Aging of Asphalt Mixtures for Performance Testing and Prediction: Phase III
Results, Transportation Research Board, 2021.
117566, http://vtrc.virginiadot.org/ProjDetails.aspx?id=710. [17] R.A. Schapery, Correspondence principles and a generalized J Integral for Large
Deformation and Fracture Analysis of Viscoelastic Media, Int. J. Fracture 25 (3)
References (1984) 195–223, https://doi.org/10.1007/BF01140837.
[18] M. Elwardany, V. Veginati, A. Andriescu, and, D. Mensching, Understanding the
sensitivity of mixture cracking performance tests to binder properties and long-
[1] M. Rathore, M. Zaumanis, and, V. Haritonovs, Asphalt Recycling Technologies: A
term aging, 036119812311706, Transp. Res. Record: J. Transp. Res. Board (2023),
Review on Limitations and Benefits, IOP Conf. Series: Mater. Sci. Eng. 660 (1)
https://doi.org/10.1177/03611981231170629.
(2019) 012046, https://doi.org/10.1088/1757-899X/660/1/012046.
[19] D.J. Mensching, M.D. Elwardany, and, V. Veginati, Evaluating the sensitivity of
[2] M. Zaumanis, and, R.B. Mallick, Review of very high-content reclaimed asphalt use
intermediate temperature performance tests to multiple loose mixture aging
in plant-produced pavements: state of the art, Int. J. Pavem. Eng. 16 (1) (2015)
conditions using the FHWA accelerated loading facility’s RAP/RAS experiment,
39–55, https://doi.org/10.1080/10298436.2014.893331.
Transp. Res. Rec. 2676 (10) (2022) 474–485, https://doi.org/10.1177/
[3] S. Gulzar, A. Fried, J. Preciado, C. Castorena, S. Underwood, J. Habbouche, and,
03611981221090237/ASSET/IMAGES/LARGE/10.1177_03611981221090237-
I. Boz, Towards sustainable roads: a state-of-the-art review on the use of recycling
FIG6.JPEG.
agents in recycled asphalt mixtures, .Cleaner Prod. 406 (2023) 136994, https://
[20] Y. Wang, and, Y. Richard Kim, Development of a pseudo strain energy-based
doi.org/10.1016/J.JCLEPRO.2023.136994.
fatigue failure criterion for asphalt mixtures, Int. J. Pavem. Eng. 20 (10) (2019)
[4] J. Habbouche, I. Boz, B.S. Underwood, C. Castorena, S. Gulzar, A. Fried, and,
1182–1192, https://doi.org/10.1080/10298436.2017.1394100.
J. Preciado, Review from multiple perspectives for the state of the practice on the
[21] A. Epps-Martin, F. Kaseer, E. Arámbula-Mercado, A. Bajaj, L.G. Cucalon, F. Yin,
use of recycled asphalt materials and recycling agents in asphalt concrete surface
A. Chowdhury, J. Epps, C. Glover, E.Y. Hajj, N. Morian, J.S. Daniel, M. Oshone,
mixtures, Transp. Res. Record: J. Transp. Res. Board 2676 (No. 4) (2022) 407–420,
R. Rahbar-Rastegar, C. Ogbo, and, G. King, Evaluating the effects of recycling
https://doi.org/10.1177/03611981211061130.
agents on asphalt mixtures with High RAS and RAP binder ratios, NCHRP (2020)
[5] W. Cao, and, C. Wang, A new comprehensive analysis framework for fatigue
9–58.
characterization of asphalt binder using the linear amplitude sweep test, Construct.
[22] A.T. Martins, Contribuição Para a Validação Do Ensaio de Resistência Ao Dano Por
Build. Mater. 171 (2018) 1–12, https://doi.org/10.1016/J.
Fadiga Para Ligantes Asfálticos (2014).
CONBUILDMAT.2018.03.125.
[23] Possebon, É.P. Caracterização Avançada de Ligantes e Misturas Asfálticas
[6] H. Chen, and, H.U. Bahia, Modelling effects of aging on asphalt binder fatigue
Brasileiras. Universidade Federal de Santa Maria, 2021.
using complex modulus and the LAS test, Int. J. Fatig. 146 (2021) 106150, https://
[24] P.H. Osmari, L.F.M. Leite, F.T.S. Aragão, M.C.C. Cravo, L.N. Dantas, and, T.
doi.org/10.1016/J.IJFATIGUE.2021.106150.
F. Macedo, Cracking resistance evaluation of asphalt binders subjected to different
[7] R. Zhang, J.E. Sias, and, E.V. Dave, Development of new performance indices to
laboratory and field aging conditions, Road Mater. Pavem. Design 20 (sup2) (2019)
evaluate the fatigue properties of asphalt binders with ageing, Road Mater. Pav.
S663–S677, https://doi.org/10.1080/14680629.2019.1618530.
Design 23 (2) (2022) 377–396, https://doi.org/10.1080/
[25] P.G. Tavares Marinho Filho, L.C. De Figueirêdo Lopes Lucena, and, M. Costa Lopes,
14680629.2020.1826349.
Investigation of titanium nanoparticles-reinforced asphalt composites using
[8] J.J.G. Mainieri, P. Singhvi, H. Ozer, B.K. Sharma, and, I.L. Al-Qadi, Fatigue
rheological tests, Transportes 30 (2) (2022) 2614, https://doi.org/10.14295/
tolerance of aged asphalt binders modified with softeners, Transp. Res. Rec. 2675
transportes.v30i2.2614.
(11) (2021) 1229–1244, https://doi.org/10.1177/03611981211025510.
[26] P. Singhvi, J.J. García Mainieri, H. Ozer, B. Sharma, Rheology-chemical based
[9] F. Zhou, P. Karki, and, S. Im, Development of a simple fatigue cracking test for
procedure to evaluate additives/modifiers used in asphalt binders for performance
asphalt binders, Transp. Res. Rec. 2632 (1) (2017) 79–87, https://doi.org/
enhancements, Phase 2 (2021).
10.3141/2632-09.
[27] R. Zhang, J.E. Sias, and, E.V. Dave, Comparison and correlation of asphalt binder
[10] B. Asadi, N. Tabatabaee, and, R. Hajj, Use of linear amplitude sweep test as a
and mixture cracking parameters incorporating the aging effect, Constr. Build.
damage tolerance or fracture test to determine the optimum content of asphalt
Mater. 301 (2021) 124075, https://doi.org/10.1016/J.
rejuvenator, Constr. Build. Mater. 300 (2021) 123983, https://doi.org/10.1016/j.
CONBUILDMAT.2021.124075.
conbuildmat.2021.123983.
[28] R. Zhang, J.E. Sias, and, E.V. Dave, Development of new performance indices to
[11] P. Lin, X. Liu, P. Apostolidis, S. Erkens, S. Ren, S. Xu, T. Scarpas, and, W. Huang, On
evaluate the fatigue properties of asphalt binders with ageing, Road Mater. Pavem.
the rejuvenator dosage optimization for aged SBS modified bitumen, Constr. Build.
Design (2020), https://doi.org/10.1080/14680629.2020.1826349.
Mater. 271 (2021) 121913, https://doi.org/10.1016/J.
[29] Asphalt Institute, IS-235 State-Of-The-Knowledge, The Use of REOB/VTAE in
CONBUILDMAT.2020.121913.
Asphalt, Auburn, 2016.
[12] A. Fried, L.G. Xue, S. Gulzar, J. Preciado, B.S. Underwood, and, C. Castorena,
Evaluation of recycling agent and extender dosage selection procedures to restore

17