1-s2.0-S00162361130053097-main

Fuel 113 (2013) 454–466

Contents lists available at SciVerse ScienceDirect

Fuel

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

Coniferyl-alcohol lignin as a bio-antioxidant for petroleum asphalt:A quantum chemistry based atomistic study

0016-2361/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.fuel.2013.06.003

⇑ Tel.: +1 (202) 319 5165; fax: +1 (202) 319 6677.E-mail address: [email protected]

T

Tongyan Pan ⇑The Catholic University of America, 620 Michigan Avenue, N.E., Washington, DC 20064, United States

h i g h l i g h t s

� At 130 �C, asphalt generates alkanes and sulfoxides more rapidly than ketones.� Amine groups do not contribute much to asphalt aging.� Chain breaking, sulfoxidation, and ketonization are vital to asphalt aging.� Coniferyl-alcohol lignin shows radical-scavenging effect in bulk asphalt.� Coniferyl-alcohol lignin is most effective when temperature is below 130 �C.

ED

a r t i c l e i n f o

Article history:Received 15 April 2013Received in revised form 2 June 2013Accepted 4 June 2013Available online 19 June 2013

Keywords:Atomistic modelingAsphaltOxidative agingConiferyl alcoholLignin

a b s t r a c t

Petroleum asphalt is an important base material for many industrial applications, such as the binding andwaterproofing component in road pavements and roof shingles. Being an organic end product of petro-leum serving under the general open-to-air conditions, asphalt can lose the desired rheological propertieswith time due to oxidative hardening or aging that frequently leads to increase in viscosity, separation ofcomponents, and loss of cohesion and adhesion, and thereby becomes hardened. A common practice toalleviate asphalt aging today is using different chemical additives or modifiers as antioxidants. Thecurrent state of knowledge in asphalt oxidation and antioxidant evaluation is focused on monitoringthe degradation in asphalt’s physical properties, mainly the viscosity and ductility, which althoughsatisfying direct engineering needs does not contribute to the fundamental understanding of the agingand anti-aging mechanisms. Within this context, this study was initiated to study the anti-oxidationmechanisms of bio-based additives, using the coniferyl-alcohol lignin as an example, by developing aquantum chemistry based chemophysical environment in which the various chemical reactions amongasphalt components, anti-oxidative additive and oxygen, as well as the incurred physical changes canbe studied. The techniques of X-ray photoelectron spectroscopy (XPS) was used to prove the validityof the modified and unmodified asphalt models, from which the XPS results showed high agreementto the model predictions.

� 2013 Elsevier Ltd. All rights reserved.

TRAC

E

Being an organic product from the remains of ancient organ-isms, petroleum asphalt is an important base material for manyindustrial purposes, such as the primary binding and waterproof-ing component in road pavements and roof shingles [1]. Todayaround 95% of the roads we drive on are covered with asphaltmixtures. Asphalt in its general service conditions however is sub-ject to chemical oxidation by reactions with atmospheric oxygen,which can cause the hardening of asphalt and the sacrifice of itsdesirable physical properties. In asphalt pavements, for example,oxidative hardening is responsible for mixture embrittlement that

R

The primary factor that controls the rate of asphalt aging istemperature. In engineering practices, most aging occurs in theprocess of making asphalt-based end products, such as asphaltconcrete in the intense heat of an asphalt plant [2,3]. During theshort duration of heating, the temperature can reach up to165 �C. Long-term oxidative aging begins immediately after theend product, such as an asphalt pavement or roof, is put in service,which occurs at a much slower rate than the initial oxidation dur-ing mixing and production. However, long-term oxidative agingcan cause failures as serious as the initial oxidation, including alarge increase in stiffness and loss of ductility, and cracking ofthe end products under thermal and/or load stresses [1,4].

T. Pan / Fuel 113 (2013) 454–466 455

A

Petroleum asphalt is composed of carbon and hydrogen, andnitrogen, sulfur and oxygen of lower percentages [2,4]. Trace heavymetals such as vanadium and nickel often are also present. Theseelements combine to form the main fractions of asphalt cement:asphaltenes, saturates, naphthalene and polar aromatics (alsoknown as light and heavy resins, respectively) according to theCorbett’s method [5]. The three fractions, each providing differentproperties of asphalt, chemically and physically interact with eachother and form the complex mixture system of asphalt. Asphalt-enes and saturates are normally incompatible compounds, andare brought together by aromatics. Asphaltenes are the main con-tributors of viscosity and therefore hardening effects; and an abun-dance of aromatics and saturates decrease the ductility and henceelastic effects [4–10].

In general, oxidative hardening of asphalt is believed to becaused by the generation of oxygen-containing polar chemicalfunctionalities on asphalt molecules, which in turn can causeagglomeration among molecules due to increased chemophysicalassociations such as hydrogen bonding, van der Waals force, andCoulomb force [4–10]. In addition to oxygen-containing function-alities, oxidation also can cause aromatization of certain asphaltmolecules that facilitates further agglomeration of asphalt compo-nents in ambient conditions [5–10]. However, the systematic studyof asphalt oxidation from the chemical perspective has not beenseriously attempted due to one underlying challenge in the stateof knowledge: the complexity of chemical composition in asphalt,which has naturally led to the lack of an effective tool for studyingthe manifold chemical reactions involved in asphalt oxidation[3,8].

Chemical agents have been studied widely as antioxidants inasphalt. The presence of polymers in asphalt, among variouschemical antioxidants, can effectively decrease aging rate andreduce asphalt’s temperature susceptibility, such as the use ofstyrene–butadiene–styrene (SBS) and styrene-b-butadiene (SBR)in high-volume roadways to increase the high- and low-tempera-ture properties of asphalt and reduce the pavement’s tendency toform ruts or crack [11]. There are many polymers that have shownanti-oxidation abilities for asphalt; however, in the state of prac-tices none of them truly demonstrates satisfactory in situ anti-oxi-dative performance [12]. As such, the identification of an effectiveantioxidant of asphalt binder is of great benefit to the asphalt-re-lated industries. More information in the regard of asphalt agingcan be found in the literature per the references of [4,6,8,10–12].

Of the various polymers as antioxidants in asphalt, wood ligninhas also been researched and once shown promising performance.At a seven-percentage usage by weight of modified asphalt, woodlignin significantly slowed down the oxidation of asphalt [13]. Lig-nin is one of the most abundant organic polymers on Earth, consti-tuting 30% of non-fossil organic carbon and from a quarter to athird of the dry mass of wood, and exceeded only by cellulose[14]. As a biopolymer, lignin is a cross-linked three-dimensionalhydrophobic and aromatic molecule. The degree of polymerizationin nature is difficult to measure, since it is fragmented duringextraction and the molecule consists of various types of substruc-tures that appear to repeat in a haphazard manner. Although dif-ferent forms of lignin have been described depending on themeans of isolation, there are three monolignol monomers in theform of a benzene ring with a tail of three carbons, methoxylatedto various degrees: p-coumaryl alcohol, coniferyl alcohol, and sina-pyl alcohol [14]. Coniferyl alcohol is the main monomer unit thatmakes up the lignin in softwood. These lignols are incorporatedinto lignin in the form of the phenylpropanoids p-hydroxyphenyl,guaiacyl, and syringal, respectively [14].

The capability of lignins as a type of asphalt antioxidant is be-lieved to arise from the scavenging action of their phenolic struc-tures on oxygen containing free radicals [15]. The phenolic

RETR

structures of lignin are benzene rings with attached hydroxylgroups. Benzene rings are six carbon structures with each carbonsharing a single and double covalent bond to another carbon. Ina phenolic group, there can be one or more hydroxyl groups at-tached to the benzene ring. The ability of phenolic compounds tobe antioxidants is the functional group’s ability to neutralize freeradicals [16]. A phenol can neutralize a free radical by donatingeither a proton or an electron [15]. Because of its structure, a phe-nol is able to do both while remaining relatively stable. Lignin con-tains a large amount of phenolic groups, making it an effectiveantioxidant [16]. Many factors affect the antioxidant ability of lig-nin, of which the source of biological origin is the most prominentfactor [15]. Each plant is biologically and chemically different;therefore, the lignin obtained from them will be different. Theextraction method is also very important in determining a lignin’santioxidant ability. Lignin can be extracted from plant material bychemicals such as ethanol, acetone, acetic acid, methanol, and pro-panol, which produce lignins of different antioxidant abilities [15].

Lignins however can also get oxidized when exposed to oxygenor atmosphere, which is similar to the oxidation of asphalt binderbut at a slower pace. At raised temperature (e.g., P150 �C),oxidation of lignins can be significant. Based on a comprehensivestudy of oxidation of lignins from different sources (150 �C inoxygen for 4 h), the combined gas–liquid chromatography and asequential methylation technique indicated the presence of alarger number of products, including different acids of low molec-ular weights (e.g., formic acid, two-carbon-atom acids, and somethree-carbon-atom and four-carbon-atom acids), and variousphenolic compounds such as vanillin, syringaldehyde, andp-hydroxybenzaldehyde as the methyl ethers and carboxy-containing compounds as the methyl esters or methyl ether esters[17]. According to Bryan, oxygen in the early stages of oxidation isnecessary to break lignin polymer into smaller molecular frag-ments [18]. Notably, such a level of temperature, i.e., 150 �C, isclose to those at which many asphalt products are produced (e.g.,asphalt mixtures), and might be of significance when lignin isevaluated as an asphalt antioxidant.

Within the current state of knowledge, a research study was re-cently initiated to explore the molecular-scale mechanisms of as-phalt oxidation and anti-oxidation by different antioxidants,aiming at elucidating the chemical bases of asphalt oxidative hard-ening and identifying effective antioxidants for asphalts. This man-uscript presents the results from the study, focusing on reportingthe development of a quantum chemistry (QC) based atomisticmodel, validation of the model, and utilization of the model instudying asphalt oxidation and the coniferyl-alcohol lignin as a po-tential antioxidant.

CTED

2. Quantum chemistry perspective of asphalt oxidation andanti-oxidation

A fundamental understanding of asphalt oxidation and anti-oxi-dation by chemical additives is highly desired when dealing withthe problems related to asphalt aging, this task however cannotbe readily accomplished within the frame of traditional organicor polymer chemistry that has centered on experimenting bulk as-phalt using different devices. The past few decades has seen signif-icant advances made in using spectroscopic techniques forcharacterizing molecular and atomic phenomena that involveessentially the transfer of electrons [19–24]. Since the oxidationof asphalt is mainly a chemical process, an approach capable ofsimulating electron-transfer processes is desired. In this study,the quantum chemistry (QC) is resorted to as such an approach.In recent years, the efforts made in atomic-level description ofphenomena have been accelerated by the availability of powerful

456 T. Pan / Fuel 113 (2013) 454–466

A

computers and parallel computing methods, advances in the statis-tical mechanics and new experimental data. The QC-based meth-ods today are readily implemented using different computer-based programs, of which the dominant one is the Density Func-tional Theory (DFT) [25].

Extensive use of QC computation today however is limited tosmall atomic/molecular systems and therefore is not practical forstudying asphalt oxidation even though recent computing capacityhas significantly improved the speed of QC computation that en-ables predicting accurately the geometries and vibrational ener-gies. As such, it is necessary to have an accurate force-field basedmethod that enables quick evaluation of inter-atomic bondingand forces. Such a first principle based force-field method has theadvantage of being able to simulate chemical reactions whileobtaining fast computation speed as traditional force field methodsdo, and therefore is highly desired for simulating asphalt oxidationthat involves thousands of atoms for a realistic simulation.

3. Reactive force field for asphalt oxidation and anti-oxidation

The Reactive Force Field (ReaxFF) developed by van Duin et al.[22] is one such first-principle based force field method, whichsince advent has been parameterized and implemented to a varietyof materials and processes, including high-energy materials,hydrocarbon reactions, and transition-metal-catalyzed nanotubeformation. Recently ReaxFF has been extended to more materialsincluding various polymers, metals, ceramics, and silicon, and isnow used as a general tool for chemical simulations. The reactiveforce field of a CAHAOANAS system (carbon, hydrogen, oxygen,nitrogen, and sulfur) originally developed by van Duin et al.[22,26,27] is re-evaluated in this study to simulate the oxidationof asphalt in exposure to oxygen.

The overall energy of the CAHAOANAS ReaxFF contains a ser-ies of energy contributions per Eq. (1), all determined using QC. Theactual number of energy contributions of a ReaxFF system dependson the type of chemical species and processes to be modeled. TheReaxFF for simulating asphalt oxidation in an oxygen environmentinvolves complex chemical reactions between the five elementspecies of asphalt and external oxygen.

Esystem ¼ Ebond þ Eval þþEtors þ EH�bond þ EvdWaals þ ECoulomb ð1Þ

In Eq. (1), the bond energy Ebond describes the chemical energybetween each pair of bonded atoms. Valence angle energy Eval ac-counts for the energy contribution from valence angle; torsionrotation energy Etors ensures proper dependence of the energy oftorsion angle for bond order approaching trivial and bond ordergreater than 1; van der Waals interactions energy EvdWaals accountsfor the van der Waals interactions; and Coulomb interactions en-ergy ECouloms between all atom pairs adjust for orbital overlap be-tween atoms at close distances.

The fundamental ReaxFF assumption is that the bond order BO0ijbetween a pair of atoms is dependent on the interatomic distancerij according to Eq. (2), in which the parameter r0 is the bond radiusand the series of parameters ps describe the bond order. In calcu-lating bond orders, ReaxFF distinguishes between contributionsfrom r bonds, p bonds, and pp bonds. The bond orders BO0ij are up-dated in each time step. The energy of the system is finally deter-mined by summing up all the energy contributions per Eq. (1). Thethree pairs of parameters: pbo,1 and pbo,2, pbo,3 and pbo,4, and pbo,5

and pbo,6 in Eq. (2) correspond to the orders of the r bond, the firstp bond, and the second pp bond, respectively, of which the valuesare given in Table 1. The values of the exponential terms is unitybelow a particular interatomic distance r0 and negligible at a long-er distance. The bond energy is calculated from the bond order BO0ij.

RETR

BO0ij ¼ BOrij þ BOp

ij þ BOppij

¼ exp pbo1 �rij

rro

� �pbo2� �

þ exp pbo3 �rij

rpo

� �pbo4� �

þ exp pbo5 �rij

rppo

� �pbo6� �

ð2Þ

The connectivity related terms in Eq. (1) such as the bond en-ergy, valence angle and torsion angle energy terms are also bondorder dependent and will disappear upon bond dissociation. Thisfeature of ReaxFF ensures a smooth transition of the energy andforce from a bonded system to a non-bonded system. In additionto the valence interactions which depend on overlap, there arerepulsive interactions at short interatomic distances due to Pauliprinciple orthogonalization and attraction energies at long dis-tances due to dispersion. These interactions, comprised of vander Waals and Coulomb forces, are included for all atom pairs, thusavoiding awkward alterations in the energy description duringbond dissociation. In this respect, ReaxFF is similar in spirit tothe central valence force fields used earlier in vibrational spectosc-ropy. The following sections introduced these energy contributionterms.

3.1. Bond energy

Ebond of the CAHAOANAS system was determined according toEq. (3). De and pbe,1 and pbe,2 are bond parameters. Upon the disso-ciation of a bond, the bond order BO0ij approaches zero making thebond energy term Ebond disappear (Eq. (3)). To simulate oxidation ofasphalt, the energy contributions developed by van Duin et al.[22,26,27] are used to determine the overall system energy—hencethe ReaxFF potential function—for the atomic-level modeling of as-phalt oxidation, with the parameters given in Table 1.

Ebond ¼ �De � BO0ij � exp½pbe1ð1� ðBO0ijÞpbe2 Þ� ð3Þ

CTED

3.2. Valence angle energy

Eval was determined according to Eq. (4). Just like bond energy,it is important that the energy contribution from valence angleterm goes to zero as the bond orders in the valence angle goes tozero. Eq. (4) determines the energy associated with deviations invalence angle Hijk from its equilibrium value H0. The term f1(BO)per Eq. (4a) ensures that the valence angle energy contribution dis-appears smoothly during bond dissociation. Eq. (4b) deals with theeffects of over/under coordination in central atom j on the valenceangle energy. Atom undercoordination/overcoordination parame-ter Dj is defined for atoms as the difference between the total bondorder around the atom and the number of its bonding electronsvalence. The equilibrium angle H0 for Hijk depends on the sumof p-bond orders around the central atom j and changes fromaround 109.47� for sp3 hybridization (p-bond = 0) to 120� for sp2

(p-bond = 1) to 180� for sp (p-bond = 2) based on the geometryof the central atom j and its neighbors. Table 2 lists the parametricvalues used to determine the valence angle energy Eval.

Eval ¼ f1ðBOijÞ � f1ðBOjkÞ � f2ðDjÞ � fpval1 � pval1

� exp½�pval2ðHoðBOÞ �HijkÞ2�g ð4aÞ

f1ðBOijÞ ¼ 1� expð�BOpval3ij Þ ð4bÞ

f2ðDjÞ ¼2þ expðDjÞ

1þ expðDjÞ þ expð�pval4 � DjÞð4cÞ

Table 1ReaxFF parameters for determining bond order and Ebond of asphalt as a CAHAOANAS system.

Bond De (kcal/mol) pbe,1 pbe,2 pbo,1 pbo,2 pbo,3 pbo,4 pbo,5 pbo,6

CAC 145.4070 0.2176 �0.1940 5.9724 1.0000 8.6733 1.0000 �0.7816 0.3217CAH 167.1752 �0.4421 1.0000 8.5445 0.0000 0.0000 1.0000 0.0000 0.5969HAH 188.1606 �0.314 1.0000 5.7082 0.0000 0.0000 1.0000 0.0000 0.6816CAO 171.0470 0.36 �0.2660 5.0637 0.0000 7.4396 1.0000 �0.1696 0.3796OAO 90.2465 0.995 �0.1850 6.2396 1.0000 7.5281 1.0000 �0.2435 0.9704CAN 134.9992 0.042 �0.3161 5.4980 1.0000 7.0000 1.0000 �0.1370 0.2415OAN 127.7074 0.4561 �0.3555 7.0000 1.0000 7.0000 1.0000 �0.1481 0.2000NAN 151.9142 0.428 �0.1614 5.3056 1.0000 12.1345 1.0000 �0.1001 0.6229HAO 216.6018 �0.4201 1.0000 5.9451 0.0000 0.0000 1.0000 0.0000 0.9143HAN 223.1853 �0.4661 1.0000 6.1506 0.0000 0.0000 1.0000 0.0000 0.5178CAS 128.9942 0.1035 �0.2398 5.6731 1.0000 8.1175 1.0000 �0.5211 0.6000HAS 151.5159 �0.4721 1.0000 7.0050 1.0000 0.0000 1.0000 0.0000 0.6000OAS 100.0000 0.5563 �0.4577 7.1145 1.0000 12.7569 1.0000 �0.4038 0.6000NAS 0.0000 0.4438 �0.3153 5.6864 1.0000 9.1227 1.0000 �0.2034 0.6000SAS 96.1871 0.0955 �0.2373 6.4757 1.0000 9.7875 1.0000 �0.4781 0.6000

Table 2Parameters for determining valence angle energy in the developed CAHAOANAS system.

Angle atoms H0 (degree) pval,1 (kcal/mol) pval,2 (1/radian2) pval,3 pval,4

CACAC 70.0265 13.6338 2.1884 0.1676 1.0400CACAH 69.7786 10.3544 8.4326 0.1153 1.0400HACAH 74.6020 11.8629 2.9294 0.1367 1.0400CAHAH 0.0000 0.0000 6.0000 0.0000 1.0400CAHAC 0.0000 3.4110 7.7350 0.0000 1.0400HAHAH 0.0000 27.9213 5.8635 0.0000 1.0400CACAO 72.9588 16.7105 3.5244 1.1127 1.1880OACAO 80.0708 45.0000 2.1487 1.1127 1.1880CACAN 61.5055 45.0000 1.2242 1.1127 1.1880OACAN 71.9345 45.0000 1.5052 1.1127 1.1880NACAN 51.3604 45.0000 0.6846 1.1127 1.1880HACAO 66.6150 13.6403 3.8212 0.0755 1.0500HACAN 68.9632 16.3575 3.1449 0.0755 1.0500CAHAN 0.0000 0.0019 6.3000 0.0000 1.0400CAOAC 79.1091 45.0000 0.7067 0.6142 1.0783CAOAO 83.7151 42.6867 0.9699 0.6142 1.0783CAOAN 79.5876 45.0000 1.1761 0.6142 1.0783OAOAO 80.0108 38.3716 1.1572 0.6142 1.0783OAOAN 81.5614 19.8012 3.9968 0.6142 1.0783NAOAN 85.3564 36.5858 1.7504 0.6142 1.0783CAOAH 78.1533 44.7226 1.3136 0.1218 1.0500HAOAO 84.1057 9.6413 7.5000 0.1218 1.0500HAOAN 79.4629 44.0409 2.2959 0.1218 1.0500HAOAH 79.2954 26.3838 2.2044 0.1218 1.0500CANAC 66.1477 22.9891 1.5923 1.6777 1.0500CANAO 91.9273 38.0207 0.5387 1.6777 1.0500CANAN 92.6933 9.9708 1.6094 1.6777 1.0500OANAO 73.4749 42.7640 1.7325 1.6777 1.0500OANAN 73.9183 44.8857 1.1980 1.6777 1.0500NANAN 74.0572 15.4709 5.4220 1.6777 1.0500CANAH 72.7016 33.4153 1.0224 0.0222 1.0500HANAO 82.4368 44.1900 1.9273 0.0222 1.0500HANAN 82.6883 39.9831 1.1916 0.0222 1.0500HANAH 71.2183 14.4528 3.6870 0.0222 1.0500CAHAO 0.0000 0.0019 6.0000 0.0000 1.0400CAHAN 0.0000 0.0019 6.0000 0.0000 1.0400CAHAS 0.0000 0.0019 6.0000 0.0000 1.0400OAHAO 0.0000 0.0019 6.0000 0.0000 1.0400OAHAN 0.0000 0.0019 6.0000 0.0000 1.0400NAHAN 0.0000 0.0019 6.0000 0.0000 1.0400HAHAO 0.0000 0.0019 6.0000 0.0000 1.0400HAHAN 0.0000 0.0019 6.0000 0.0000 1.0400CACAS 74.9397 25.0560 1.8787 0.0559 1.0400CASAC 86.9521 36.9951 2.0903 0.0559 1.0400HACAS 74.9397 25.0560 1.8787 0.0000 1.0400CASAH 86.1791 36.9951 2.0903 0.0000 1.0400CASAS 85.3644 36.9951 2.0903 0.0559 1.0400HASAH 93.1959 36.9951 2.0903 0.0000 1.0400HASAS 84.3331 36.9951 2.0903 0.0000 1.0400HAHAS 0.0000 0.0019 6.0000 0.0000 1.0400

T. Pan / Fuel 113 (2013) 454–466 457

RETRACTED

Table 4Parameters for hydrogen bond interactions energy.

Hydrogen bond rhb phb1 phb2 phb3

OAHAO 2.0431 �6.6813 3.5000 1.7295OAHAN 1.674 �10.9581 3.5000 1.7295NAHAO 1.4889 �9.6465 3.5000 1.7295NAHAN 1.8324 �8.0074 3.5000 1.7295OAHAS 2.6644 �3.9547 3.5000 1.7295NAHAS 4.0476 �5.7038 3.5000 1.7295SAHAO 2.1126 �4.5790 3.5000 1.7295SAHAN 2.2066 �5.7038 3.5000 1.7295SAHAS 1.9461 �4.0000 3.5000 1.7295

Table 5Parameters for determining van der Waals interaction energy.

Atom units D (kcal/mol) r (Å) c a P

458 T. Pan / Fuel 113 (2013) 454–466

3.3. Torsion rotation energy

Etors ensures that dependence of the energy of torsion angle xijk

accounts properly as bond order approaches trivial and bond ordergreater than 1. This is given in Eqs. (5), (5a), and (5b). Hijk and Hjkl

are valence angles. D is the atom undercoordination/overcoordina-tion parameter. Parameters ptor3 and ptor4 in Eq. (5b) equal zero forthe specific types of elements in this study. Table 3 lists the valuesof parameters V1, V2 and ptor used to determine the torsion rotationenergy Etors.

Etors ¼ f3ðBOij; BOjk;BOklÞ � sin Hijk

� sin Hjkl12

V1 � expfptor � ðBOjk � 1þ f4ðDi;DkÞÞ2g�

� 1� 2 cos xijkl

�þ 1

2V2ð1þ cos 3xijklÞ

� �ð5Þ

ij vdW w ij vdW1

C 0.1818 1.8857 2.0784 9.5928 1.5591H 0.0600 1.603 4.4187 9.3951 1.5591O 0.088 1.9741 7.7719 10.219 1.5591N 0.1376 1.9324 7.8431 10.067 1.5591

D

S 0.2099 2.0677 4.9055 9.9575 1.5591

f4ðDj;DkÞ ¼2þ expð�Dj � DkÞ

1þ expð�Dj � DkÞ þ expðDj þ DkÞð5bÞ

A

3.4. Hydrogen bond interactions

Eq. (6) described the bond-order dependent hydrogen bond foran XAH—Z system as incorporated in ReaxFF. Table 4 lists the val-ues of parameters rhb, phb1, phb2, and phb3 used to determine the tor-sion rotation energy EHbond.

EHbond ¼ phb1 � ½1� expðphb2 � BOXHÞ�

� exp phb3ro

hb

rHZþ rHZ

rohb

� 2� �� �

� sin8 HXHZ

2

� �ð6Þ

3.5. Van der Waals interactions energy

EvdWaals accounts for the van der Waals interactions using a dis-tance-corrected Morse-potential given in Eq. (7). By including ashielded interaction (Eq. (7a)), excessively high repulsions be-tween bonded atoms (1–2 interactions) and atoms sharing a va-lence angle (1–3 interactions) are avoided. Dij is the basic energyterm of an atomic pair; rvdW is the van der Waal radius. The param-

TR

Table 3Parameters for determining torsion rotation energy (X represents any of elements C,H, O, N, and S).

Torsion angle V2 (kcal/mol) V3 (kcal/mol) ptor

CACACAC 23.2168 0.1811 �4.6220CACACAH 45.7984 0.3590 �5.7106HACACAH 44.6445 0.3486 �5.1725XACAHAX 0.0000 0.0000 0.0000XAHAHAX 0.0000 0.0000 0.0000XACAOAX 16.7344 0.5590 �3.0181XAHAOAX 0.1000 0.0200 �2.5415XAOAOAX 68.9706 0.8253 �28.4693XACANAX 66.2036 0.3855 �4.4414XAHANAX 0.1000 0.0200 �2.5415XAOANAX 14.8049 0.0231 �10.7175XANANAX 37.4200 0.0107 �3.5209XACACAX 0.9305 0.0000 �24.2568NACANAN 43.6430 0.0004 �11.5507XACASAX 30.3435 0.0365 �2.7171XASASAX �42.7738 0.1515 �2.2056XAHASAX 0.0000 0.0000 0.0000

RE

eter aij is set to be 1 for the elements of this study. The otherparameters for EvdWaals are given in Table 5.

EvdWaals¼Dij � exp aij � 1� f5ðrijÞrvdW

� �� ��2 �exp

12�aij � 1� f5ðrijÞ

rvdW

� �� �� �

ð7Þ

f5ðrijÞ ¼ rpvdW1ij þ 1

cw

� �pvdW1� � 1

pvdW1

ð7aÞCTE

3.6. Coulomb interactions energy ECouloms

ECouloms is considered between all atom pairs. To adjust for orbi-tal overlap between atoms at close distances a shielded Coulombpotential ECouloms is used (see Eq. (8)).

ECoulcomb ¼ C �qi � qj

½r3ij þ ð1=cijÞ

3�1=3 ð8Þ

Atomic charges are calculated using the electron equilibrationmethod [28,29]. The initial values for the electron equilibrationmethod parameters (g, v, and c) listed in Table 6 were determinedby Njo et al. [30] The optimization parameter cij in Eq. (8) is in-cluded for orbital overlap correction in the electron equilibrationmethod.

4. Atomistic modeling of asphalt oxidation and anti-oxidationby lignin

Petroleum asphalt is a molecular system of different chemicalelements: carbon (�85%), hydrogen (�11%), sulfur (1–5%), nitrogen

Table 6Parameters for determining Coulomb interactions energy.

Atom type g (eV) v (eV) c (Å)

C 6.9235 5.7254 0.8712H 9.8832 3.8196 0.7625O 7.8386 8.500 1.0804N 6.3404 6.8418 0.8596S 8.2545 6.500 1.0336

T. Pan / Fuel 113 (2013) 454–466 459

A

(0.3–1.1%), oxygen (0.2–0.8%), and some trace species such asvanadium (4–1400 ppm) and nickel (0.4–110 ppm). Bulk asphaltconsists of molecules with different structures formed by atomsof these chemical species. Like in general organic substances, struc-tures of the asphalt molecules are more important than theamount of each element present. Some heteroatoms, i.e., sulfur,nitrogen, oxygen, are attached to carbon atoms in different config-urations, forming different polar molecules or functional groupsdue to imbalance of electrochemical forces which are weaker thanthe primary chemical bond—the covalent bond that holds atomswithin asphalt molecules. Molecular interactions among the polarmolecules of asphalt are the primary mechanism of agglomera-tions that strongly influence the physical properties and macro-scopic performance of asphalt. Usually a small amount of polarmolecules can have a great effect on most engineering behaviorof asphalt, such as the resistance to binder-aggregate stripping thatin general is believed to be controlled by the adsorption of polarmolecules to the surface of mineral aggregates [7,10].

A proper balance among the different functional groups hence isimportant for producing durable and resistant asphalts againstdetrimental physical property changes during oxidative aging[31]. A classic method for separating asphalt into different func-tional groups is per their solubilities in a series of organic solvents,supplemented with chromatographic analyzes (namely, the Cor-bett method) [5,10]. To study the oxidation of asphalts in their ser-vice conditions, a typical petroleum asphalt can be divided in threerepresentative functional groups: asphaltenes (relatively largemolecules insoluble in straight-chain alkanes such as n-heptaneor n-pentane), resins (naphthene aromatics—alkane-soluble andelute in aromatic solvents such as benzene, and polar aromatics—alkane-soluble and elute in more polar solvents, such as an aro-matic/alcohol mixture), and oils or saturates (molecules that eluteimmediately in n-heptane) [5,10]. The resin and oil portions to-gether also are known as maltenes or petrolenes. Asphaltenes, withhigher molecular weight and polarity than resins and saturates, ifnot properly dispersed by the resinous components of maltenescan cause reduced asphalt compatibility and asphalt durabilitythereby [5,10].

In this study, a model asphalt system representing the threefunctional groups of general petroleum asphalt was developedaccording to the composition determined by Carbognani [32]. Suchan ‘‘average’’ asphalt model consists of a polyaromatic portion asincluded in the asphaltene and naphthene groups of asphalt, andsome straight or branched chain alkanes like those in saturategroup of asphalt. The overall mixture composition, i.e., the relativeamounts of the three ingredient functional groups used for makingthe average asphalt model was in agreement to the NMR-basedmeasurements by Storm et al. [33]. The lignin used for this studywas a commercially available one, i.e., the coniferyl alcohol asthe main monomer unit of lignins in softwood. The coniferyl-alcohollignin contains large amounts of phenol structures, giving itantioxidant ability. Two specific types of coniferyl-alcohol ligninmolecules were mixed with asphalt, with two lignin-1 moleculesand three lignin-2 molecules, reaching a 22 weight percentage inmodified asphalt.

The lignin materials was mixed with the modeled average as-phalt binder at 130 �C, and then subject to two different levels oftemperature, 130 �C and 150 �C, to evaluate its anti-oxidation ef-fect on the asphalt binder. The rational for selecting 130 �C and150 �C conditioning for 2 h are to test the survival ability of the lig-nin, based on existing literature indicating that most lignin willlose activities at above 150 �C. The 2 h is the typical time for sim-ulating asphalt aging. The coniferyl-alcohol lignin studied in thiswork showed different survival abilities at 130 �C vs. 150 �C, whichindicates that fast mixing and compacting procedure is necessaryfor hot mixes and that the lignin studied is more suitable for

RETR

warm-mix asphalts. Notably, with only one asphalt species stud-ied, this atomistic work is a deterministic instead of a stochasticprocess. Notably, lignin might get oxidized at raised temperature,the temperature of 150 �C, which is typical in producing asphaltproducts such as asphalt pavements, was selected in this studyto test the oxidation behavior of lignins at raised temperature.The ball-stick schemes of the polyaromatic portion, chain alkaneportion, and the average model are shown in the second columnof Fig. 1a. It is noteworthy that as the average asphalt moleculewas separated into the polyaromatic portion and saturate portion,hydrogen atoms were added to each portion to make the includedmolecules chemically balanced. The ball-stick schemes of the twolignin molecules are shown in the first column of Fig. 1b.

Thermodynamics and kinetics are two distinct aspects of chem-ical reactions. While thermodynamics describes the possibility anddirection of a reaction in terms of the free energy, DG, released orconsumed during a chemical reaction; kinetics concerns how fastthe chemical reaction can reach equilibrium, i.e., the rate of thereaction, as influenced by factors of reaction condition such asthe temperature and concentrations of reactants. The thermody-namics aspect of asphalt oxidation and lignin anti-oxidation isevaluated in this study by observing respectively the oxidationbehavior of the lignin molecules, and the average molecule andits polyaromatic/saturate portions in a domain of oxygen mole-cules, all under the same simulation condition. The kinetics of as-phalt oxidation and lignin anti-oxidation depends on factors suchas asphalt composition, products formed in oxidation, reactiontemperature, oxygen partial pressure, and other physicochemicaleffects of the system. Accordingly, the kinetics of the differentmolecular species was evaluated by studying the bulk asphalt oxi-dized under the same condition as used in studying the thermody-namics of the average molecule and its components, i.e., at onestandard atmospheric pressure (1 ATM) pressure (oxygen only)and 130 �C. A total of twenty such average asphalt molecules arecompacted to the density of typical asphalt materials under theambient condition, i.e., 0.98 � 103 kg/m3.

For the kinetics of asphalt oxidation, the compacted asphalt wasthen used to build a chemophysical environment, in form of anatomistic model shown in Fig. 2a, to study the oxidation behaviorof bulk asphalt under the condition of temperature equal to 130 �Cand oxygen partial pressure of 1 ATM. In the average asphalt mod-el, its component models and the bulk chemophysical model, inaddition to the asphalt components oxygen molecules are includedthat are separated at an average distance of 1.2 nm (same as theinter-molecule distance in air at 1 ATM). Such a simulationcondition was determined based on the experience of numerousasphalt-oxidation studies conducted before. The chemicalfunctionality developed in asphalt when oxidized at 130 �C wasonce found to be similar to that developed during normal pave-ment aging at ambient temperatures [34]. Also, it was shown thatoxidation of asphalts in exposure to high oxygen particle pressure(e.g., 100% oxygen) is equivalent to that in air of 300 psi pressure(such as in a Pressure Aging Vessel (PAV) condition), which is be-lieved to be the same oxidation level as typically found in asphaltsafter five or more years of pavement service [35]. The use of 1 ATMpressure for modeling is important for studying the escape of exist-ing small-molecule-weight components of asphalt or newly gener-ated ones during asphalt oxidation from bulk asphalt, which mightbe an important mechanism contributing to asphalt hardening.

As such, each ReaxFF model contains an individual or a numberof the selected lignin molecules, average asphalt molecule(s) or itscomponents, forming a domain embedded in oxygen molecules(see Fig. 1a and b for the single-molecule and molecular compo-nent models, and Fig. 2a, c and e for the bulk asphalt model). EachReaxFF model was run at a Velocity Verlet plus Berendsen ensem-ble [36,37] at a time step of 0.25 femtoseconds (fs) using a parallel

CTED

Fig. 1. Thermodynamics study of lignin molecules, and average asphalt molecule and its components, (a) average asphalt molecule and its components and (b) ligninmolecules.

460 T. Pan / Fuel 113 (2013) 454–466

RETRACTED

computing algorithm developed based on the Message-PassingInterface (MPI) Standard. The Virginia Tech’s suite of high perfor-mance computers—System X was used for the computation. Sys-tem X is a supercomputer comprising 1100 Apple PowerMac G5cluster nodes and capable of running at 12.25 Teraflops (meaning1012 FLoating point OPerations per Second). Each ReaxFF modelwas kept running until the equilibrium of oxidation is reached,i.e., no more new species observed for 1 h. The modified and

unmodified bulk asphalt models after oxidation simulation isshown in Fig. 2b, d and f.

5. Validation of the chemophysical environment

The technique X-ray photoelectron spectroscopy (XPS), a pow-erful surface method for accurately detecting the presence and

(a) (b)

(c) (d)

(e) (f)Fig. 2. Oxidation of unmodified and lignin-modified asphalts at 130 �C and 150 �C, (a) bulk asphalt without lignin before oxidation, (b) after oxidation at 130 �C for 2 h, (c)bulk asphalt with lignin before oxidation, (d) after oxidation at 130 �C for 1 h, (e) bulk asphalt with lignin before oxidation and (f) after oxidation at 150 �C for 2 h.

T. Pan / Fuel 113 (2013) 454–466 461

RETRACTED

relative quantities of chemical elements (except hydrogen and he-lium), was used in this study to validate the modified and unmod-ified bulk asphalt models by determining and comparing theamounts of generated functional groups before and after asphaltoxidation. The grade of the asphalt studied was PG 64-16, whichwas determined using the SuperPave method. It is noteworthy thatthe focus of the study is the chemical oxidation (aging) behavior.

The rheological properties of the binder are close to that of PG64-16. Three replicates were prepared and tested in the XPS stud-ies. XPS analysis can obtain information of the state and environ-ment of atoms in the sample surface (usually 1–10 nm in depthand 10 lm in width depending on the input energy), which canfurther be analyzed for information of the surface structure ofthe surface. Such powerful ability of XPS is particularly useful for

(a)

(b)Fig. 3. XPS spectra of unmodified asphalt before and after oxidation, (a) survey scanand (b) high-energy scan of carbon and sulfur zones.

462 T. Pan / Fuel 113 (2013) 454–466

ACTED

studying the oxidative aging behavior of asphalt binders [38–40],for which a sufficiently thin sample (e.g., a few lms) can be easilyprepared and oxidized with uniform oxidation throughout the en-tire depth of the sample in a relatively short period of time (e.g., 2 hat 130 �C) [41].

XPS analysis is based on the photoelectric effect stemming fromthe ejection of electrons from the surface of a sample in exposureto electromagnetic radiation of sufficient energy. Electrons emittedhave characteristic kinetic energies proportional to the energy ofthe radiation, according to Eq. (9), in which KE is the kinetic energyof the electron, h is Planck’s constant, m is the frequency of the inci-dent radiation, Eb is the ionization or binding energy, and u is thework function—a constant dependent on the X-ray photoelectronspectrometer used. In an XPS analysis, a level of energy radiationis used to expel core electrons from a sample, and the kinetic ener-gies of the resulting core electrons are measured. Using Eq. (9) withthe kinetic energy KE and known frequency m of radiation, the bind-ing energy Eb of the ejected electron can be determined. By Koop-man’s theorem, which states that binding or ionization energy isequivalent to the negative of the orbital energy, the energy of theorbital from which the electron originated can be determined.These orbital energies are characteristic of the element and itsstate.

KE ¼ ht� Eb �u ð9Þ

A Physical Electronics 5600 XPS setup was used in this study,which provides the relative frequencies of binding energies of elec-trons measured in 0.1 electron-volts (0.1 eV). The binding energieswere then used to identify the elements to which each peak corre-sponded, as elements of the same kind in different states and envi-ronments usually have different characteristic binding energies.Furthermore, comparing the areas under the peaks gave relativepercentages of the elements detected in a sample. The oxygen con-centration adopted in the XPS analysis was controlled by oxygenpressure. The 1 ATM oxygen pressure (oxygen only) used was high-er than the atmospheric oxygen proportion (0.21 ATM), which wasintended to accelerate the oxidation reaction of asphalt. To main-tain the oxygen at this relatively high concentration, the asphaltsamples were first vacuumed in an environmental chamber, andoxygen was then regulated into the chamber until the 1 ATM pres-sure was achieved. During the vacuum process at the room tem-perature, small oil molecules and low-molecular-weight alkanesfrom asphalt are released from asphalt surface. The lost portionwas counted as the short-chain alkanes, and leftovers were testedfor ketones and sulfoxides. An initial survey XPS scan was con-ducted to identify the elements present in the modified andunmodified asphalt samples (Figs. 3a and 4a), followed with a sub-sequent high-resolution scan of the peaks of interest, i.e., the car-bon and sulfur in this study, to identify ketones and sulfoxides asobserved in the atomistic simulation (Figs. 3b and 4b). As no obvi-ous generation of new nitrogen compounds in the atomistic modeland XPS survey scan, high-resolution scan was not conducted atthe nitrogen peak. For the high-resolution scan, the computer soft-ware associated with the Physical Electronics 5600 XPS device de-tects the transform of sigma carbon–carbon single bonds (beforeoxidation) to carbon–oxygen double-bonded ketones (after oxida-tion), and the transform of sulfides (before oxidation) to sulfoxides(after oxidation) at the carbon and sulfur peaks, which was alsopredicted by the atomistic models as shown in Figs. 1 and 2.

Furthermore, the generations of ketones and sulfoxides duringoxidation, as quantified by XPS for the same modified and unmod-ified asphalt materials as modeled atomistically, are shown inFig. 5, which further validates the ReaxFF-based bulk asphaltmodel by showing close results in the amounts of newly generatedketones and sulfoxides. The slight difference between theexperimental results and numerical predictions might come from

RETR

the machine or operation errors related to XPS. In preparing sam-ples for XPS analysis, the modified/unmodified asphalt sampleswere vacuumed at the room temperature following the proceduresestablished by Ruiz et al. [41]. The samples (in dry powders) werethen pressed onto a piece of thin indium foil (0.1 mm thick) as thesample substrate. The graphite tape was not used as sample sub-strate for the carbon-based asphalt to avoid peaks from the graph-ite tape, which would otherwise add to the carbon peak andpotentially skewing or overlapping the XPS spectra.

6. Simulation results and discussions

Fig. 1a also shows the thermodynamic aspect of the oxidation ofthe average asphalt and its components. Under the same oxidationcondition, different component molecules of the average asphaltmolecule, i.e., the polyaromatics and alkanes oxidize at differentrates and lead to different products. The carbon atoms in saturatealkanes are quite resistant to oxidation; instead some chains showthe sign of breaking into shorter chains. This could constitute animportant mechanism for asphalt hardening as liquid portions ofasphalt vaporize into air relatively easily under general serviceconditions. However, no sulfoxidation or ketonization was ob-served in the saturate alkanes. The carbon atoms in the polyaro-matic part of the average asphalt molecule show a partial level ofoxidation, i.e., oxidized at the benzylic position, not the carbonatoms forming benzene rings, resulting in a ketone. For the polyar-omatic, sulfoxidation occurs to the sulfur atom forming a sulfoxidethat consists of a double bond with an oxygen atom (S@O), how-ever no ketonization was observed. Different from the two compo-nent models, both sulfoxidation and ketonization were observed inthe average asphalt model and the ketonization occurs to somebenzylic carbon atoms. In the average asphalt model, sulfoxidation

(a)

(b)Fig. 4. XPS spectra of lignin-modified asphalt before and after complete oxidation (10 h), (a) survey scan and (b) high-energy scan of the carbon zone.

T. Pan / Fuel 113 (2013) 454–466 463

ACTED

seems to be easier than ketonization as the sulfoxide appears ear-lier than the ketone.

The snapshot of the oxidation of unmodified asphalt at 2 h,shown in Fig. 2b, displays two obvious phenomena: (1) generationof light-molecular-weight saturate molecules (with shorter chainsthan the two original saturates), which tend to leave the bulk as-phalt via diffusion; and (2) agglomeration of oxidized saturatesand aromatics as attracted by the oxygen-bearing functionalgroups. Some unbroken long saturates entangled with suchagglomeration due to larger molecular weight and/or electrostaticforces. These two observations are in agreement to the phenomenaof chain breaking of saturates, sulfoxidation, and ketonization ob-served in the preceding thermodynamics studies shown inFig. 1a. The two mechanisms probably contribute significantly tothe oxidative hardening of asphalt. The generated new specieswas summarized in Fig. 5a. In general, the generations of new spe-cies all tend to slow down and approach an upper bound numberof production, as more reactants were consumed. The generationrate of sulfoxides seems to exceed that of ketones in the earlystages, and gradually lags behind. The generation of shorter chainswas not very dramatic; however the generation rate of ketonesseemed to dominate the overall reaction at a rate higher than bothsulfoxide generation and saturate-chain breaking. Ketones there-fore may contribute more to the long-term agglomeration of agedasphalt.

Ketones and sulfoxides are the major oxidation productsformed in oxidative aging of the modeled average asphalt, whichis in agreement to results of existing laboratory experiments suchas by Functional Group Analyzes [5,10]. Moreover, such oxidationproducts formed are consistently observed among field asphalts

RETR

from different sources [34,35], and are in good agreement withthe general oxidation chemistry of hydrocarbon and sulfur-con-taining molecules [38–40]. Ketones are a group of organic com-pounds carrying a carbonyl group, C@O, in which the carbonatoms is bonded to two adjacent carbon atoms. The amine groups(nitrogen compounds) however do not show contribution to as-phalt hardening. Ketones are weak acids due to the a-proton adja-cent to the carbonyl group that are much more acidic compared tosimple hydrocarbons, and can be removed by common bases suchas HO� and RO�. Sulfoxides are another group of chemical com-pounds, containing a sulfinyl functional group, S@O, attached totwo carbon atoms. The generation of ketones and sulfoxides arethermodynamically and kinetically favorable in general asphaltunder the typical open-to-air conditions. Although the bond be-tween sulfur and oxygen atoms differs from conventional doublebonds like that between carbon and oxygen in ketones, sulfoxidesand ketones are both polar, containing an electronegative oxygenatom producing a dipole in interaction or association with other di-poles or induced dipoles, thus both contributing to agglomerationof oxidized asphalt. The XPS spectra clearly show these two spe-cies, which was in good agreement to the ReaxFF simulation re-sults. Also, these two species have demonstrated effects ofincreasing viscosity and molecular aggregation in aged asphalt asreported in numerous research studies conducted to link the phys-icochemical properties of the ketones and sulfoxides to viscosityand distresses of asphalt and pavements. More information in thisregards can be found in the series of SHRP I reports.’’

Fig. 1b shows the thermodynamic aspect of the oxidation of thetwo lignin molecules. At 130 �C lignin molecules did show signs ofoxidation, as can be seen from the second column of Fig. 1b.

(a)

(b)Fig. 5. Experimental vs. numerical results of oxidized unmodified/modified bulk asphalt (first 2 h), (a) generation of new functional groups at 130 �C and (b) generation ofnew functional groups at 150 �C.

464 T. Pan / Fuel 113 (2013) 454–466

RACTED

However, lignins do get oxidized when temperature is raised to150 �C at the same oxygen partial pressure, generating vanillinand glycolaldehyde for the two coniferyl-alcohol lignin molecules.More meaningfully, the addition of coniferyl-alcohol lignin signifi-cantly changes the oxidation behavior of the average asphalt mod-eled in this study, as presented in Figs. 2d and e, 4a and b, and 5aand b. At 130 �C the lignin does show effectiveness in anti-oxidation,which can be seen from the levels of agglomeration in the oxidizedasphalt models without lignin (Fig. 2b) vs. with lignin (Fig. 2d).This anti-oxidation effect is owing to the scavenging action oflignin’s phenolic structures on oxygen containing free radicals,i.e., the ketones and sulfoxides formed in asphalt oxidation [15].Phenolic structures are benzene rings with attached hydroxylgroups, which can form p-conjugation regions that promote theformation of donor–acceptor complex in modified asphalt. Suchcomplexes act as catalysts or activators for scavenging the ketonesand sulfoxides generated in asphalt oxidation [16,42]. However,when the temperature is raised to150 �C the lignin moleculesstarted to get oxidized and generated vanillin and glycolaldehyde,and their scavenging action began to decrease. Father oxidation oflignin at 150 �C will lead to glycolic acid as observed in bothatomistic simulation and XPS analysis. The increased level ofagglomeration as lignin molecules started to get oxidized can beclearly seen by comparing Fig. 2d–f.

RET

rized in Fig. 5a and b. In unmodified asphalt the generation rateof sulfoxides seems to exceed that of ketones in the early stages,and gradually slows down and lags behind. The generation rateof shorter chains shows a similar trend as sulfoxides but at higherrates; the generation rate of ketones however continues at a higherrate than both sulfoxides and shorter-chain alkanes, which con-tributes more to the agglomeration of aged asphalt in the longrun. When modified with lignin, the asphalt demonstrated a signif-icantly slower oxidation rate at 130 �C than the unmodified as-phalt. Lignins in general are more resistant to oxidation thanasphaltenes and resins in asphalt under the same conditions. At150 �C however, the generation rates of shorter-chain saturatesand ketones are higher than those at 130 �C, however a lower ratein sulfoxides can be seen. Therefore, the coniferyl-alcohol lignins,when added in asphalt and not oxidized, are capable of reducingthe aging speed of asphalt. Also, added in solid powders, ligninsusually can increase asphalt’s stiffness.

Furthermore, lignins tend to get oxidized at temperatureP150 �C and/or raised oxygen partial pressure, generating vanillinand glycolaldehyde for coniferyl-alcohol lignin. Sustained oxida-tion of the new species can lead to the generation of vanillic acid(from vanillin) and glycolic acid (from glycolaldehyde) that canbe further oxidized to oxalic acid. When oxidized to low-molecular

T. Pan / Fuel 113 (2013) 454–466 465

A

acids, methyl ethers, methyl ether esters, and methyl esters thathave smaller sizes than oxidized asphaltenes and even naphtha-lenes, these new products help reduce the viscosity (as caused byketones and sulfoxides in oxidized asphaltenes and naphthalenes)and recover the lost flowability and ductility of asphalt. Howeverthe benefit owing to the smaller-size molecules is limited as canbe seen from Fig. 5b, in which significant amounts of ketonesand sulfoxides were generated from both asphalt and lignin mole-cules when the temperature is raised. Therefore, the anti-oxidationeffect of lignins, if demonstrated, comes mainly from its scaveng-ing actions at a non-oxidative temperature.

It is noteworthy that although asphalts from different sourceshave different chemical compositions, they include approxi-mately the same set of chemical elements, i.e., carbon, hydrogen,oxygen, sulfur, and nitrogen, with some trace species. Quantum-chemistry based ReaxFF is a forcefield-based method that mod-els a molecular system composed of the chemical elementsincorporated in the force field. Hence, the present model is capa-ble of modeling any material systems composed of the same ele-ments as asphalt, i.e., carbon, hydrogen, oxygen, sulfur, and traceelements. Such material systems reasonably include organic anti-oxidants for asphalt and the more general organic materials.Moreover, the overall mixture composition, i.e., the relativeamounts of the three ingredient functional groups used for mak-ing the ‘‘average’’ asphalt model was in agreement to the threerepresentative functional groups of asphalt: asphaltenes, resins,and oils or saturates. Although there exist different species of as-phalt, their major differences lie in the relative proportions ofthese three representative functional groups. Since the oxidationbehavior of these representative functional groups each is ratherstable, the species of oxides from the XPS tests and the numer-ically simulated asphalt species are not supposed to changesmuch in different asphalt types. The amounts of such oxide spe-cies may be different, which however does not significantly af-fect the analysis and conclusion of this study.’’ It is admittedthat this work was focused on the major representative ingredi-ents of asphalt, i.e., asphaltenes, resins, and saturates. The meth-od developed is applicable to different virgin asphalt species. Asto modified or recycled asphalts, with more impurity species in-cluded in the force field, the applicability can also be reasonablyexpected. Regarding the performance of asphalt pavements, afield aging model of asphalt can be developed based on theatomistic model presented in this study, which can be furtherincorporated into the mixture and pavement design guides. Sucha model will significantly improve the accuracy for predictingthe service lives of asphalt binders and pavements. Admittedly,this work presents the early results of a study on asphalt aging,in which only an average asphalt model was studied. More com-plex systems such as recycled asphalt pavement will signifi-cantly add to the difficulty for model building, simulation, andidentifying functional species. Anyhow, with the major and rep-resentative ingredients of asphalt, i.e., asphaltenes, resins, andsaturates included, the aging behavior of general asphalt is cap-tured. More desirably, more lignin species can be evaluated toeventually form a database that can be used to direct the useof bio-based materials as asphalt antioxidants.

RETR

7. Summary and conclusions

The phenomena of asphalt oxidation concerns petroleum andmany manufacture industries. Based on a QC-based reactive forcefield re-evaluated for the CAHAOANAS system, the thermody-namics and kinetics of asphalt oxidation and the anti-oxidationmechanism of coniferyl-alcohol lignin are studied, aided withXPS analyzes. Important conclusions are made as follows.

1. Two distinct stages of asphalt aging can be identified based onthe composition and structure of generated chemical speciesand their generation speeds. Asphalt molecules exhibit achain-breaking trend and a fairly high reactivity with oxygen,causing a rapid spurt of light-molecular-weight alkanes andsulfoxides, and a relatively low production of ketones. Thisspurt is followed by a slower rate of asphalt oxidation andhardening.

2. Different oxidation mechanisms appear operative during thesetwo periods. During the initial spurt, sulfoxides are the majoroxidation product that controls viscosity increase. Followingthe spurt, ketones become the major product. The ratio ofketones to sulfoxides formed depends on sulfur content inasphalt. The amine groups however do not show obvious con-tribution to asphalt hardening.

3. The oxidation of asphalt and lignin involves many stable radi-cals, and a number of intermediate chemicals that can only beobserved at the atomic scale due to the high chemical instabil-ities of such intermediate products. Significant decompositionof saturates and evaporation of small-molecular-weight hydro-carbon were observed in the simulation. The sulfur atoms andbenzylic carbon atoms account for most of the oxidative reac-tions in the modeled asphalt. Therefore for oxidative hardeningof asphalt in typical service conditions, chain breaking of satu-rates, sulfoxidation, and ketonization could be the majormechanisms.

4. The coniferyl-alcohol lignin can be used as antioxidant forpetroleum asphalt, with the maximum radical-scavengingeffectiveness achieved in a non-oxidative condition of the lignin(e.g., <130 �C under the equivalent oxygen particle pressure of1 ATM). This means that when added in asphalt the mixingtemperature must be controlled to maximize the anti-oxidationeffect of lignins.

References

[1] McNichol D. Paving the way: asphalt in America. 1st ed. National AsphaltPavement Association; 2005.

[2] Petersen JC, Branthaver JF, Robertson RE, Harnsberger PM, Duvall JJ, Ensley EK.Effects of physicochemical factors on asphalt oxidation kinetics. In:Transportation Research Record 1391, TRB, National Research Council,Washington, D.C; 1993. p. 1–10.

[3] Davis TC, Petersen JC. An inverse glc study of asphalts used in the zaca-wigmore experimental test road. Proc, Assoc Asphalt Pav Technol1967;36:1–15.

[4] Plancher H, Green EL, Petersen JC. Reduction of oxidative hardening of asphaltsby treatment with hydrated lime: a mechanistic study. Proc Assoc Asphalt PavTechnol 1976;45:1–24.

[5] Corbett LW. Composition of asphalt based on generic fractionation usingsolvent deasphalteneing, elution-adsorption chromatography anddensiometric characterization. Anal Chem 1969;41:576–9.

[6] Petersen JC. A dual sequential mechanism for the oxidation of asphalts. PetrolSci Technol 1998;16(9–10):1023–59.

[7] Branthaver JF, Petersen JC, Robertson RE, Duvall JJ, Kim SS, Harnsberger PM,Mill T, Ensley EK, Barbour FA, Schabron JF. Binder characterization andevaluation, Chemistry, vol. 2. In: Report No. SHRP-A-368, SHRP, NationalResearch Council, Washington, D.C; 1993.

[8] Petersen JC. Asphalt aging: a dual oxidation mechanism and itsinterrelationships with asphalt composition and oxidative age hardening.Transportation Research Record 1638, TRB, National Research Council,Washington, D.C; 1998. p. 47–55.

[9] Hveem FN, Zube E, Skog J. Progress report on the zaca-wigmore experimentalasphalt test project, vol. 277. American Society for Testing Materials; 1959. p.1–45.

[10] Petersen JC. Chemical composition of asphalt as related to asphalt durability:state of the art. Transportation Research Record 999, TRB, National ResearchCouncil, Washington, D.C; 1984. p. 13–30.

[11] Ruan Y, Davison RR, Glover CJ. Oxidation and viscosity hardening of polymer-modified asphalts. Energy Fuels 2003;17:991–8.

[12] Lucena MC, Soares SA, Soares JB. Characterization of thermal behavior ofpolymer-modified asphalt. Mater Res 2004;7:529–34.

[13] Bishara SW, Robertson RE, Mohoney D. Lignin as an antioxidant: a limitedstudy on asphalts frequency used on kansas roads. In: The 42nd PetersonAsphalt research conference, Cheyenne, WY; 2005.

CTED

466 T. Pan / Fuel 113 (2013) 454–466

[14] Boerjan W, Ralph J, Baucher M. Lignin bios. Annu Rev Plant Biol2003;54(1):519–49.

[15] Dizhbite T, Telysheva G, Jurkjane V, Viesturs U. Characterization of the radicalscavenging activity of lignins-natural antioxidants. Bioresour Technol2004;95:309–17.

[16] Boeriu CG, Bravo D, Gosselink RJ, Van Dam JE. Characterization of structure-dependent functional properties of lignin with infrared spectroscopy. IndCrops Prod 2004;20:205–18.

[17] Brink DL, Bicho JG, Merriman MM. Oxidation degradation of wood III, lignin,structure and reactions. Adv Chem Ser 1966;177(59).

[18] Bryan CC. Manufacture of vanillin from lignin. United States Patent 2692291;October 1954.

[19] Pan T, van Duin ACT. Steel surface passivation at a typical ambient condition:atomistic modeling and X-ray diffraction/reflectivity analyses. Electrocatalysis2011;2(4):307–16.

[20] Pan T, van Duin ACT. Passivation of steel surface: an atomistic modelingapproach aided with X-ray analyses. Mater Lett 2011;65(21–22):3223–6.

[21] Pan T. Quantum chemistry-based study of iron oxidation at the iron-waterinterface: an X-ray analysis aided study. Chem Phys Lett 2011;511(4–6):315–21.

[22] van Duin ACT, Dasgupta S, Lorant F, Goddard WA. ReaxFF: a reactive force fieldfor hydrocarbons. J Phys Chem A 2001;105.

[23] Pan T, Xi Y. Physicochemical nature of iron oxidation in a damp atmosphericcondition. Acta Metall Sin 2011;24(6):415–22.

[24] Pan T, Lu Y. Quantum-chemistry based studying of rebar passivation inalkaline concrete environment. Int J Electrochem Sci 2011;2(8).

[25] Parr RG, Yang W. Density-functional theory of atoms and molecules. NewYork: Oxford University Press; 1989. ISBN 0-19-504279-4.

[26] Chenoweth K, van Duin ACT, Goddard WA. ReaxFF reactive force field formolecular dynamics simulations of hydrocarbon oxidation. J Phys Chem A2008;112:1040–53.

[27] Keten S, Chou C-C, van Duin ACT, Buehler MJ. Tunable nanomechanics ofprotein disulphide bond begets weakening in reducing and stabilization inoxidizing chemical microenvironments. J Mech Behav Biomed Mater2012;5:32–40.

[28] Mortier WJ, Ghosh SK, Shankar S. Electronegativity-equalization method forthe calculation of atomic charges in molecules. J Am Chem Soc 1986;108:4315.

RETRA

[29] Janssens GOA, Baekelandt BG, Toufar H, Mortier WJ, Schoonheydt RA.Comparison of cluster and infinite crystal calculations on zeolites with theelectronegativity equalization method (EEM). J Phys Chem 1995;99:3251.

[30] Njo SL, Fan J, van de Graaf B. Extending and simplifying the electronegativityequalization method. J Mol Catal A 1998;134:79.

[31] White RM, Mitten WR, Skog JB. Fractional components of asphalt:compatibility and interchangeability of fractions produced from differentasphalts. Proc Assoc Asphalt Pav Technol 1970;39:498–531.

[32] Carbognani L. Molecular structure of asphaltene proposed for 510c residue ofvenezuelan crude. In: INTEVEP S.A. Tech. Report; 1992.

[33] Storm DA, Edwards JC, DeCanio SJ, Sheu EY. Molecular representations ofratawi and alaska north slope asphaltenes based on liquid- and solid-stateNMR. Energy Fuels 1994;8:561–6.

[34] Petersen JC, Plancher H, Miyake G. Chemical reactivity and flow properties ofasphalts modified by metal complex-induced reaction with atmosphericoxygen. Proc Assoc Asphalt Pav Technol 1983;32:32–60.

[35] Griffen RL, Simpson WC, Miles TK. Influence of composition of paving asphaltson viscosity, viscosity-temperature susceptibility, and durability. J Chem EngData 1959;4:349–54.

[36] Verlet L. Computer ‘experiments’ on classical fluids: I. Thermodynamicalproperties of Lennard–Jones molecules. Phys Rev 1967;159(1):98–103.

[37] Berendsen HJC, Postma JPM, van Gunsteren W, DiNola A, Haak JR. Moleculardynamics with coupling to an external bath. J Chem Phys 1984;81.

[38] Gelius U, Heden RF, Hedman J, Lindberg BJ, Manne R, Nordberg R, et al.Molecular spectroscopy by means of ESCA III: carbon compounds. Phys Scripta1970;2(1–2):70–80.

[39] Kelemen SR, George GN, Gorbaty ML. Direct determination and quantificationof sulfur forms in heavy petroleum and coals: 1. the X-ray photoelectronspectroscopy (XPS) approach. Fuel 1990;69:939–44.

[40] Lindberg BJ, Hamrin K, Johansson G, Gellius U, Fahlman A, Nordllng C, et al.Molecular spectroscopy by means of ESCA II: sulfur compounds, correlation ofelectron binding energy with structure. Phys Scripta 1970;1(5–6):288–98.

[41] Ruiz JM, Carden BM, Lena LJ, Vincent EJ, Escaller JC. Determination of sulfur inasphalts by selective oxidation and photometric spectroscopy for chemicalanalysis. Anal Chem 1982;54:688–91.

[42] Kuzina SI, Demidov SV, Brezgunov AY, Poluektiv OG, Grinberg OY, DubinskiAA, et al. Study of free-radical centers in lignin with 20 mm band ESRspectroscopy. J Polym Sci 1993;A35(7):798–802.

TED

C