lable at ScienceDirect

Composites Part B 131 (2017) 165e172

Contents lists avai

Composites Part B

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

Molecular dynamics simulations on adhesion of epoxy-silica interfacein salt environment

Yohannes L. Yaphary a, b, Zechuan Yu a, Raymond H.W. Lam b, David Hui c, Denvid Lau a, d, *

a Department of Architecture and Civil Engineering, City University of Hong Kong, Hong Kong, Chinab Department of Mechanical and Biomedical Engineering, City University of Hong Kong, Hong Kong, Chinac Department of Mechanical Engineering, University of New Orleans, New Orleans, LA 70148, USAd Department of Civil and Environmental Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, USA

a r t i c l e i n f o

Article history:Received 10 March 2017Received in revised form21 May 2017Accepted 28 July 2017Available online 31 July 2017

Keywords:ResinsAdhesionEnvironmental degradationComputational modeling

* Corresponding author. Department of ArchitecturUniversity of Hong Kong, Hong Kong, China.

E-mail address: denvid@mit.edu (D. Lau).

http://dx.doi.org/10.1016/j.compositesb.2017.07.0381359-8368/© 2017 Elsevier Ltd. All rights reserved.

a b s t r a c t

Epoxy-silica interface exists in various types of newly built as well as aging concrete structures whichinclude flooring and anchoring systems. Sea environment and deicing salt often let the concrete struc-tures be exposed to sodium chloride solution. It is observed from the laboratory evaluation that suchexposure often leads to the poor bond durability of epoxy-concrete interface. Yet, there is still a lack ofsuch fundamental understanding as the existing investigations are merely limited to macroscale ormesoscale. Indeed, bond at the interface is governed by the interactions between dissimilar materialslying at the atomistic scale, where atomistic modeling and molecular dynamics simulations can effec-tively reflect the mechanical behaviors of the bilayer interface. Herein, molecular dynamics simulationstogether with Bell's model are employed to evaluate the effect of sodium chloride solution on theadhesion at interface between epoxy and silica. It is shown that sodium chloride solution weakens theadhesion significantly. This finding indicates that the bond deterioration at epoxy-concrete interface iscritical in the presence of salt solution, which must be considered in the engineering design strategy foroffshore and marine structures.

© 2017 Elsevier Ltd. All rights reserved.

1. Introduction

Epoxy-silica interface plays the important adhesion systemappearing in concrete, which dominates the construction materialin building and infrastructure. Aggregates and gluing cement hy-drate which are the component of hardened concrete, all containssilica [1e3]. Meanwhile, epoxy is the polymer with excellentadhesion to concrete. As compared to concrete, epoxy has bettermechanical performance and is relatively inert to wide array ofchemicals [4e7]. In dry condition, the failure of concrete-epoxybonded system mostly occurs at concrete [8]. It means that theadhesive force at the interface is higher than the cohesive force ofhydrated cement in concrete. Apart from dry condition, there areobserved conditions in which the adhesive force at the interface ismore critical for the integrity of epoxy-concrete bonded system [9].One of these conditions is where the sodium chloride (NaCl)

e and Civil Engineering, City

solution is present [10,11]. This condition is common exposure toconcrete structure based on the numerous existence of marine (i.e.coastal and offshore) structures and the application of salt baseddeicer [12,13]. The loss of adhesive force at the interface can causeserious issue to the level of catastrophic structural failure leading tofatality. The popular representative case is the failure of concretestrengthened by bonded fiber reinforced polymer (FRP) which canbe seen as structural system of multi-layer material consisting ofFRP, epoxy and concrete [14e20].

Macroscale mechanical tests were used in many studies toassess the effect of salt environment on the durability characteristicof FRP-bonded concrete system [21]. It is observed that the me-chanical properties of such composites are degenerated by thepresence of salt with the failure mode mostly observed due todetachment of FRP strengthening system from concrete surfacecaused by the interfacial debonding between epoxy and concrete[22]. These reported studies confirm the macro- to meso-scale ef-fect of salt to the durability of adhesion at epoxy-concrete interface.However, the effect of salt at atomistic level such as on epoxy-silicainterface as the subsystem of epoxy-concrete interface is not yetunderstood. The investigation at atomistic level can give more

Y.L. Yaphary et al. / Composites Part B 131 (2017) 165e172166

information that is closer to the fundamental origin of debonding atthe interface which is governed by adhesion of interatomic inter-action. Atomistic simulations have been used as the powerful toolfor the investigation at atomistic level to obtain adhesion energythat is well correlated with existing meso-scale experimental data[23,24]. Model has been used to study the adhesion energy be-tween epoxy-silica interface in the presence of moisture imple-menting molecular dynamics (MD) simulations [25,26]. Thisexample demonstrates that MD simulations can be used to studythe durability of epoxy-concrete interface through an under-standing of how epoxy-silica interacts with the environment atatomistic level.

The objective of this study is to probe the adhesion energy ofsilica-epoxy interface in presence of NaCl solution using MD sim-ulations. NaCl is the predominant dissolved salt in sea water andone of commonly used deicers [27]. The model consists of epoxyattached to silica substrate immersed in NaCl solution. Subse-quently, steered molecular dynamics (SMD) simulations are per-formed to apply an external detaching force to epoxy, whilekeeping silica substrate fixed. Then, modified Bell's model is used tocalculate adhesion energy that can be accomplished by imple-menting the maximum force obtained from series of SMD simula-tions at various speeds. The calculated adhesion energy from SMDsimulations of NaCl system is comparedwith those of dry andmoistsystem and the trend is linked with the observation in bonddurability at macroscopic level of the three systems reported inexisting literature. Our finding reflects the effect fromNaCl solutionon the deterioration of adhesion energy at epoxy-silica interface,which can further imply the bond durability of epoxy-concreteinterface.

2. Atomistic simulations

Three systems of models named dry, water and NaCl system aredeveloped to respectively mimic the epoxy-silica bonded system indry, moist and NaCl condition. In this study, we are focusing on theeffect from saturated state of NaCl solution on deterioration ofinterfacial adhesion energy between epoxy and silica. Naturally, ahigher NaCl concentration leads to a worse deterioration, andhence the saturated state is chosen for studying the maximumimpact provided by NaCl. This section provides the detaileddescription on forcefield, model, MD simulations and technique forthe quantification of adhesion energy.

2.1. Forcefield

The Consistent Valence Forcefield (CVFF) has been parameter-ized against a wide range of experimental observations for water,free ions and variety of other functional groups, as well as someinorganic materials including silica [28,29]. Herein, the CVFF po-tential energy function as shown in Eq. (1) is used with the pa-rameters are presented in Appendix [30].

E ¼ Pbond

Kbðb� b0Þ2 þXangle

Kqðq� q0Þ2 þX

torsion

Kf½1þ cosðnf� f0

Pout�of�plane

Kc½1þ cosðnc� c0Þ � þXL�J

" Aij

rij

!12

� Bijrij

!6 #þ

XCoulomb

Kb, Kq, Kf and Kc is the force constant; b0, q0, f0 and c0 is theequilibrium bond length, bond angle, torsion angle and out-of-plane (improper) angle; b, q, f and c is bond length, bond angle,torsion angle and out-of-plane angle; n is the periodicity param-eter; Aij and Bij is the square root of the product between Ai with Aj

and Bi with Bj; C and ε is energy-conversion constant and permit-tivity; rij is the distance between the atom i and j with charges qiand qj, respectively. In CVFF, partial charge is estimated by usingbond incrementmethod [31]. The bond increment dij is described asthe partial charge contributed from atom j to atom i. This methodassigns dij with equal magnitude and opposite sign to each pair ofbonded atoms i and j. For atom i, the charge is calculated by thesummation of dij as given in Eq. (2).

qi ¼Xj

dij (2)

where j runs over all the atoms which are bonded to atom i directly.The parameters of CVFF force field are assigned to the component ofthe simulated system using Forcite tool in BIOVIA Materials Studiosoftware from Accelrys [32].

The density of water and NaCl solution from simulation andexperiment is compared to verify the applicability of CVFF gov-erning the interaction between Naþ and Cl� ion with water mole-cule. The density is chosen as this value is widely confirmed and itssimple measuring procedure minimize the experimental error. Thedensity of simulated water and NaCl solution is obtained from theequilibrated condition of water and NaCl box under NPT 300 K and1 atm. The water box consists 3000 water molecules. The NaClsolution is in saturated condition with solubility 360 g per 1000 gwater [33]. This corresponds to NaCl box solution consisting of 2715water molecules and 294 Naþ and Cl� ions. In water and NaClsystem, the explicit model of water and spherical Naþ and Cl� ionsare employed. The water and NaCl solution box is generatedimplementing Packmol package [34]. The molecular dynamicssimulations are performed using LAMMPS with the data is gener-ated implementing msi2lmp tool [35].

2.2. Model

In the dry, water and NaCl systems, themodel of epoxy and silicarefers to the model used in study by Büyük€oztürk et al. [25].Crystalline silica (SiO2) used is a-quartz with structure shown inFig. 1a with unit cell dimensions a ¼ 4.916 Å, b ¼ 4.916 Å andc¼ 5.405 Åwith a¼ b¼ g¼ 90� [36]. This unit cell contains 3 Si and6 O atoms. A bulk silica substrate is constructed from a unit cell ofcrystalline silica shown in Fig. 1a by repeating the unit cell in x-, y-and z-directions. The bulk silica is then cut to obtain the silica blockwith the dimension a ¼ 50 Å, b ¼ 50 Å and c ¼ 20 Å witha¼ b¼ g¼ 90�. Afterwards the created open bond is terminated byhydrogen atomwhich means the non-periodic boundary conditionis applied to the silica substrate. All the atoms in silica substrate arecovalently bond to each other by harmonic bond. The epoxy put in

Þ �

Cqiqjεrij

(1)

Fig. 1. (a) Substrate model of SiO2 composed with reproduction of quartz unit cell and (b) Diglycidyl ether of bisphenol A (DGEBA) with five (n ¼ 5) repeating units. Yellow, white,grey and red color represents silicon, hydrogen, carbon and oxygen atom, respectively. (For interpretation of the references to color in this figure legend, the reader is referred to theweb version of this article.)

Y.L. Yaphary et al. / Composites Part B 131 (2017) 165e172 167

the model is the chain of diglycidyl ether of bisphenol A (DGBEA)with five repeating units as shown in Fig. 1b, which hereinafter willbe stated as epoxy. The model of the silica substrate and the epoxyis developed using BIOVIA Materials Studio.

The developed silica substrate and epoxy then are incorporatedinto the simulation box of the three systems. The size of simulationbox of the three systems is ~58� 58� 78 Åwith silica substrate andepoxy at the bottom. The size of simulation box is sufficiently largeto avoid the interaction between periodic image of silica substrate[37,38]. The amount of H2O in simulation box of water system is6541 molecules and that in NaCl system is 5819 molecules. Addi-tionally, there are 630 Naþ and Cl� ions in the simulation box ofNaCl system. In designing the amount of water molecule and Naþ

and Cl� ions, the density of water used is 0.998 g/cm3 at 300 K andthe density of saturated NaCl is 1.1998 g/cm3 [33,39]. The sameexplicit model of water and spherical Naþ and Cl� ions as used inthe simulation to obtain the density are employed. The water andNaCl solution box is generated implementing Packmol package.Afterwards, the components of each system are assembled togetherusing BIOVIA Materials Studio.

Fig. 2. Illustration on equality of parameterization used in the method of (a) SMDsimulations and (b) AFM experiment. vconstant is pulling speed. In SMD, k is springconstant and R and R0 is initial and deformed length of spring. Respectively, thisparameter is the stiffness and the initial and deformed distance between the top andtip of cantilever in AFM. The generated force (F) is calculated following Hooke's lawbased on the elasticity characteristic of SMD spring and AFM stiffness.

2.3. Molecular dynamics simulations

The molecular dynamics simulations are performed usingLAMMPS with the data is generated implementing msi2lmp tool[35]. The simulation scheme for the three systems are presented inTable 1. The silica substrate and epoxy system is equilibrated for 1ns in NVTensemble at 300 K and periodic boundary that within thisprocess the epoxy attach to silica substrate. Afterwards, in case ofwater and NaCl system, minimization is performed on water mol-ecules and Naþ and Cl� ions in NVT ensemble. During this mini-mization the atom coordinates of silica and epoxy are freezed. Then,

Table 1MD simulations scheme for dry, water and NaCl system.

Step System

Dry Water

1 Eq., NVT, 1ns Eq. (silicNVT, 1 n

2 SMD Min. (W3 Eq., NVT4 SMD

the whole systems are equilibrated for 1 ns. Following the mini-mization and equilibration process, steered molecular dynamics(SMD) are employed in order to pull off the epoxy from silicasubstrate in dry, water and NaCl system [25]. The basic idea of SMDis to stimulate a filament-substrate system to explore differentstates, and then free energy difference can be calculated between

NaCl

a þ epoxy),s

Eq. (silica þ epoxy),NVT, 1 ns

ater), NVE, 50 ps Min. (Water þ NaCl), NVE, 50 ps, 1 ns Eq., NVT, 1 ns

SMD

Y.L. Yaphary et al. / Composites Part B 131 (2017) 165e172168

these states by using Bell's model. In our case, the free energydifference between attached and detached states equals theadhesion energy between silica and epoxy. With proper settingsand standard manipulations, SMD canmeasure the adhesionwith ahigh accuracy [40].

As an experimental technique, atomic force microscopy (AFM)also adopts similar concept as illustrated in Fig. 2 and measures theadhesion energy [41]. Therefore, we employ SMD approach andrefer to AFM parameterization to evaluate adhesion energy be-tween silica and epoxy. By relating the SMD method with AFM inour study, we can put the parameter of simulation (i.e. springconstant) closer to reality. We employ the spring constant of65mN/m,which is the stifness of AFM cantilever commonly used tostudy the nano- and non-covalent adhesion force [42]. In this study,the middle carbon atom is tethered and it is vertically moved up to50 Å that makes the epoxy is pulled off from the silica base. Thismiddle carbon atom is chosen to minimize the height of simulationbox for complete detachment of epoxy, which consequently de-creases the amount of water molecules and Naþ and Cl� ionsincluded in the simulation hence simulation cost.

2.4. Adhesion energy

The adhesion energy is calculated using modified Bell's model[43]. Bell's model provides Eq. (3) wherein f is the external appliedforce, kb is Boltzmann constant (1.38064852 � 10�23 J K�1), T is the

Fig. 3. Density of water and NaCl from simulation and experiment.

Fig. 4. Equilibrated states of (a) dry system, (b) water system and (c) NaCl system. Yellow, wions, respectively. (For interpretation of the references to color in this figure legend, the re

temperature and v0 is the natural bond breaking speed as defined inEq. (4) [44]. A natural vibration frequency u0 z 1 � 1013 s�1 [45].

v ¼ v0 exp�f ,xbkb,T

�(3)

v0 ¼ u0,xb,exp�� Ebkb,T

�(4)

Substitution of Eq. (4) to Eq. (3) with further arrangement re-sults in Eq. (5). It appears from Eq. (5), by employing the slope andy-intercept in linear regression equation of the plot between f(maximum pulling force) against ln (v/v*), the energy barrier Eb(representing the adhesion energy between epoxy and silica atatomistic scale [25]) and transition state xb can be quantified. The v*is equal to 1 m/s used for normalization purpose. Various pullingspeed ranging from 0.2 m/s to 5000 m/s is applied to obtain theadhesion energy landscape.

f ¼�kb,Txb

�ln v� kb,T,ln v0

xb¼ A,ln vþ B (5)

It is noted that the spring constant influences the calculatedadhesion energy from Bell's model [46]. Bell's model accounts forthe frequency of forming and breaking the non-covalent bonds,which means the bonds are reversible [47]. Specifically, the lowerspring constant tends to increase the calculated adhesion energy.This is because the lower spring constant leads to slower incrementin pulling force (symbolized by F in Fig. 2). In this case, more time isavailable for the formation of the reversible bonds, leading to ahigher occurrence of these bonds. Consequently, a higher adhesionenergy is measured. However, this effect is not investigated furtherhere. Instead, we use a same spring constant in the study of dry,water, and NaCl system to make the calculated adhesion energy ofthe three systems is comparable.

3. Results and discussions

This section presents the result from the atomistic simulationsdescribed in previous section and the discussion regarding theadhesion energy at epoxy-silica interface in NaCl system ascompared to dry and water system.

Fig. 3 shows the density of water and saturated NaCl simulationbox equilibrated under NPT 300 K and 1 atm. The obtained densityfor water and saturated NaCl respectively is 0.971 g/cm3 and1.179 g/cm3, which is close with the respective values (0.998 g/cm3

and 1.200 g/cm3) from experiment [33,39].The equilibrated simulation box of the three systems is shown in

Fig. 4. In NaCl system, the ions tend to be attracted to the silicasurface that this observation complies to the physical phenomena

hite, grey, red, blue and green represents silicon, hydrogen, carbon, oxygen, Naþ and Cl�

ader is referred to the web version of this article.)

Y.L. Yaphary et al. / Composites Part B 131 (2017) 165e172 169

that the object tends to attract ions in ionic solution.The plot between the movement of tethered atom and its

respective SMD pulling force is presented in Fig. 5. The influencefrom the associated surrounding environment (i.e. water and NaCl)can be recognized from the different in evolution of the slope of thegraph toward the complete detachment of epoxy from silica sub-strate among the three system. The slope is the ratio on the change

Fig. 5. Pulling force with respect to the movement of tethered middle atom of epoxy atdifferent speed of (a) dry, (b) water and (c) NaCl system. The x-axis presents atomcoordinate along pulling direction.

between SMD pulling force and themovement of tethered atom. Itsdeviation is the manifestation from the variation in adhesion force(non-covalent bond to silica) and harmonic covalent force charac-teristic along the epoxy. Such deviation can also be interpreted asthe equality between the pulling force and the total force from theadhesion and harmonic covalent force exerted by epoxy interactingwith silica substrate. In the following discussion, the lattermentioned total force is called epoxy-silica force. The helical-shapeslope is related to the dynamic movement of the atom. The verticalslope means that the generated pulling force and epoxy-silica forceis equal. The horizontal slope can be resulted from the consecutivesimilar epoxy-silica force (its value equal to pulling force) is ach-ieved simultaneously. Similarly, the slope with positive tangent canbe due to the consecutive epoxy-silica force that is relativelydifferent is achieved successively. The pulling force overpasses theepoxy-silica force is indicated by the slope with negative tangent.The complete detachment of epoxy from silica substrate is indi-cated by sudden vertically downward drop of the slope or pullingforce closing null.

The recordedmaximum forces then are plotted against ln (v/v*)in Fig. 6. The parameters of linear regression equation areanalyzed to obtain the adhesion energy Eb. The linear relationshipcan be observed between the natural logarithmic speed of pullingspeed and the associated maximum pulling force as proposed tofollow Eq. (5). In all three systems, two distinct regimes isobserved, each of which follows a linear logarithmic dependenceof the maximum SMD pulling force with respect to the pullingspeed. The first regime is in the range of pulling speed 0.2e100 m/s and the second regime is in the range 100e5000 m/s. Thecalculated adhesion energy from the linear function in the firstregime for dry, water and NaCl system is 12.00, 7.09 and 5.11 Kcal/mole-Å and that obtained from the function in the second regimeis 0.55, 0.08 and 0.08 Kcal/mole-Å, respectively. The adhesionenergy for dry and water system corresponding to the first regimeis similar to that previously reported [25]. In fact, the lower pullingspeed as captured in the first regime gives a more detailed dy-namic evolution of the system (e.g. reversible bond) to occur inresponse to the applied SMD pulling force. As the result, a moreaccurate intrinsic adhesion energy can be obtained. Theoretically,the intrinsic properties should be independent from the pullingspeed.

Fig. 7 presents the comparison of the calculated adhesion en-ergy from the three systems. The adhesion energy of water systemis lower than that of dry system, which confirm the finding in theprevious reported study [25]. As compared to the other two sys-tems, NaCl system has the lowest adhesion energy. The adhesionenergy in NaCl system is 5.11 Kcal/mole-Å, which is 57% and 28%lower from that in dry (12.00 Kcal/mole-Å) and water (7.09 Kcal/mole-Å) system, respectively. This trend in decreasing adhesionenergy from dry to NaCl system is consistent with macroscaleexperimental result that indicates the more severe effect of NaClsolution than water only on the performance of FRP-bondedconcrete system [48]. Therefore, the finding in deterioration ofadhesion energy from MD simulations explains the generalobserved behavior that salt condition leads to failure mode in thedelamination at epoxy-concrete interface with load carrying ca-pacity lower than that in dry and moisture condition [49,50]. Thisbehavior has been observed in wide range of epoxy based adhe-sive. It is suggested that the decline in the macroscale perfor-mance of such system is a manifestation of deterioration at itssubsystem at lower scale (e.g. adhesion energy of epoxy-silicainterface).

Furthermore, this finding suggests the more detailed consider-ation upon engineering design process towards the improved bonddurability of epoxy-silica interface in salt environment. When it is

Fig. 6. Maximum detaching pulling force as a function of speed variation for (a) dry,(b) water and (c) NaCl system. The slow pulling mode (SPM) and the fast fulling mode(FPM) correspond to the pulling speed ranging from 0.2 to 100 m/s and from 100 to5000 m/s respectively.

Fig. 7. Adhesion energy of epoxy-silica interface in dry, water and NaCl condition. Theadhesion energy in NaCl condition is lower than that in dry and water condition.

Y.L. Yaphary et al. / Composites Part B 131 (2017) 165e172170

applied as strengthening system in concrete structure, suchimprovement is expected to give the significant impact to servicelife that is vulnerable to salt environment. The severe influence ofsalt observed at nano- (adhesion energy of epoxy-silica interface) tomacro- (epoxy delamination) scale implies the potential forimproved durability through the multiscale design approach.Within the molecular level, the adhesion energy can be used as aparameter to indicate bond durability at the interface. In this case,MD simulations can be employed as the powerful tool to design themolecular structure of epoxy that is suitable for salt environment.At macroscale level, it is entailed that the design approach devel-oped for the specific condition (i.e. dry and moisture condition)may need further modification to account the weakening effect ofsalt. The design tool such as finite element method should able toinclude the information from MD simulations to give the morecomplete influencing parameters in designing epoxy-bonded sys-tem for concrete structure in salt environment. The application ofthis engineering design process for improved bond durability ofepoxy-silica interface in salt environment can benefit the designstrategy to lengthen the service life of concrete offshore structurethrough the following recommendation.

The source of salt influencing the adhesion of epoxy-silicainterface may come not only from the surrounding offshore envi-ronment but also from inside the concrete. The latter source mayhappen largely due to the nature diffusion of salt solution throughthe pore of concrete. Based on the existence of salt influencing theadhesion of epoxy-silica interface, the refurbishment strategy tolengthen the offshore concrete structure implementing FRP-concrete bonded system where epoxy is used as bonding agentcan be grouped as follows:

1. The case for concrete is free from salt solution. This conditioncan be achieved by the application of concrete protectivecoating/insulation system since the initial stage of constructionto prevent salt diffusion into the concrete. The engineeringdesign of FRP-bonded concrete system following the guidancefor dry and moisture condition should be sufficient if the pro-tective coating/insulation system is reapplied appropriately af-ter refurbishment.

2. The case for concrete contains salt solution. In this condition, theengineering design of FRP-bonded concrete system should

Y.L. Yaphary et al. / Composites Part B 131 (2017) 165e172 171

account for deterioration of adhesion at epoxy-silica interface aspreviously mentioned. In this case the decrement in load car-rying capacity of FRP-bonded concrete system due to the dete-rioration at epoxy-silica interface should be considered. It is alsorecommended to apply the pretreatment on the surface ofconcrete substrate prior to the application of epoxy.

4. Conclusion

In this study, the effect of NaCl solution on the adhesion ofepoxy-silica interface has been evaluated using molecular dy-namics simulations. Our finding indicates that NaCl solutionweakens the adhesion energy of epoxy-silica interface. This resultcomplies with experimental observation on epoxy-concrete inter-face at macroscale level that also shows the weakening perfor-mance of concrete-FRP bonded system in the presence of NaClsolution. This multiscale observation suggests that the durability atmacroscale is the manifestation from the degeneration at atomisticlevel. This finding provides the piece of information that can beused in synthesis and design practice towards the more durableinterface from severe influence in salt environment.

Acknowledgement

The authors are grateful to the support from Croucher Founda-tion through the Start-up Allowance for Croucher Scholars with theGrant No. 9500012, as well as the support from the Research GrantsCouncil (RGC) in Hong Kong through General Research Fund (GRF)with the Grant No. 11255616.

References

[1] Yu Z, Zhou A, Lau D. Mesoscopic packing of disk-like building blocks in cal-cium silicate hydrate. Sci Rep 2016;6. 36967.

[2] Yu Z, Lau D. Nano-and mesoscale modeling of cement matrix. Nanoscale ResLett 2015;10(1):173.

[3] Yaphary YL, Yu Z, Lam RHW, Lau D. Effect of triethanolamine on cement hy-dration toward initial setting time. Constr Build Mater 2017;141:94e103.

[4] Kang W-S, Rhee KY, Park S-J. Thermal, impact and toughness behaviors ofexpanded graphite/graphite oxide-filled epoxy composites. Compos Part BEng 2016;94:238e44.

[5] Jia Z, Hui D, Yuan G, Lair J, Lau K-t, Xu F. Mechanical properties of an epoxy-based adhesive under high strain rate loadings at low temperature environ-ment. Compos Part B Eng 2016;105:132e7.

[6] Ren X, Peng S, Zhang W, Yi C, Fang Y, Hui D. Preparation of adipic acid-polyoxypropylene diamine copolymer and its application for tougheningepoxy resins. Compos Part B Eng 2017;119:32e40.

[7] Qiao H, Chao M, Hui D, Liu J, Zheng J, Lei W, et al. Enhanced interfacialinteraction and excellent performance of silica/epoxy group-functionalizedstyrene-butadiene rubber (SBR) nanocomposites without any couplingagent. Compos Part B Eng 2017;114:356e64.

[8] Lau D. Moisture-induced debonding in concrete-epoxy interface. HKIE Trans2012;19(3):33e8.

[9] Maljaee H, Ghiassi B, Lourenço PB, Oliveira DV. Moisture-induced degradationof interfacial bond in FRP-strengthened masonry. Compos Part B Eng 2016;87:47e58.

[10] Au C, Büyük€oztürk O. Peel and shear fracture characterization of debonding inFRP plated concrete affected by moisture. J Compos Constr 2006;10(1):35e47.

[11] Silva MA, Biscaia H. Degradation of bond between FRP and RC beams. ComposStruct 2008;85(2):164e74.

[12] Subhani M, Al-Ameri R. Strength reduction in square columns confined withCFRP under marine environment. Compos Part B Eng 2016;97:183e92.

[13] Fang H, Zou F, Liu W, Wu C, Bai Y, Hui D. Mechanical performance of concretepavement reinforced by CFRP grids for bridge deck applications. Compos PartB Eng 2017;110:315e35.

[14] Lau D, Büyük€oztürk O. Fracture characterization of concrete/epoxy interfaceaffected by moisture. Mech Mater 2010;42(12):1031e42.

[15] Lau K, Dutta P, Zhou L, Hui D. Mechanics of bonds in an FRP bonded concretebeam. Compos Part B Eng 2001;32(6):491e502.

[16] Zhou A, Qin R, Feo L, Penna R, Lau D. Investigation on interfacial defect crit-icality of FRP-bonded concrete beams. Compos Part B Eng 2017;113:80e90.

[17] Lau D, Broderick K, Buehler MJ, Büyük€oztürk O. A robust nanoscale

experimental quantification of fracture energy in a bilayer material system.Proc Natl Acad Sci 2014;111(33):11990e5.

[18] Fang H, Xu X, Liu W, Qi Y, Bai Y, Zhang B, et al. Flexural behavior of compositeconcrete slabs reinforced by FRP grid facesheets. Compos Part B Eng 2016;92:46e62.

[19] Kalfat R, Al-Mahaidi R. Mitigation of premature failure of FRP bonded toconcrete using mechanical substrate strengthening and FRP spike anchors.Compos Part B Eng 2016;94:209e17.

[20] Zhou A, Büyük€oztürk O, Lau D. Debonding of concrete-epoxy interface underthe coupled effect of moisture and sustained load. Cem Concr Compos2017;80:287e97.

[21] Karbhari V, Engineer M. Effect of environmental exposure on the externalstrengthening of concrete with composites-short term bond durability. J ReinfPlastics Compos 1996;15(12):1194e216.

[22] Boyajian DM, Ray I, Davalos JF. Freeze-thaw cycling under a calcium chlorideenvironment: effects on CFRP strengthened concrete structures. Int J MaterProd Technol 2006;28(1e2):89e102.

[23] Zhou A, Tam L-h, Yu Z, Lau D. Effect of moisture on the mechanical propertiesof CFRPewood composite: an experimental and atomistic investigation.Compos Part B Eng 2015;71:63e73.

[24] Tam L-h, Zhou A, Yu Z, Qiu Q, Lau D. Understanding the effect of temperatureon the interfacial behavior of CFRP-wood composite via molecular dynamicssimulations. Compos Part B Eng 2017;109:227e37.

[25] Büyük€oztürk O, Buehler MJ, Lau D, Tuakta C. Structural solution using mo-lecular dynamics: fundamentals and a case study of epoxy-silica interface. IntJ Solids Struct 2011;48(14):2131e40.

[26] Lau D, Büyük€oztürk O, Buehler MJ. Characterization of the intrinsic strengthbetween epoxy and silica using a multiscale approach. J Mater Res2012;27(14):1787e96.

[27] Fang Y, Wang K, Hui D, Xu F, Liu W, Yang S, et al. Monitoring of seawaterimmersion degradation in glass fibre reinforced polymer composites usingquantum dots. Compos Part B Eng 2016;112:93e102.

[28] Xiang Y, Liu Y, Mi B, Leng Y. Molecular dynamics simulations of polyamidemembrane, calcium alginate gel, and their interactions in aqueous solution.Langmuir 2014;30(30):9098e106.

[29] Tam L-h, Lau D. Moisture effect on the mechanical and interfacial properties ofepoxy-bonded material system: an atomistic and experimental investigation.Polymer 2015;57:132e42.

[30] Tam L-h, Lau D. A molecular dynamics investigation on the cross-linking andphysical properties of epoxy-based materials. RSC Adv 2014;4(62):33074e81.

[31] Sun H. Ab initio calculations and force field development for computersimulation of polysilanes. Macromolecules 1995;28(3):701e12.

[32] Accelrys Software Inc.: Materials Studio.[33] Thurmond V, Potter II R, Clynne M. The densities of saturated solutions of NaCl

and KCL from 10 degrees to 105 degrees C. U. S Geological Survey (Open-FileReport 84-253), 1984.

[34] Martínez L, Andrade R, Birgin EG, Martínez JM. PACKMOL: a package forbuilding initial configurations for molecular dynamics simulations. J ComputChem 2009;30(13):2157e64.

[35] Plimpton S. Fast parallel algorithms for short-range molecular dynamics.J Comput Phys 1995;117(1):1e19.

[36] Rosa A, El-Barbary A, Heggie M, Briddon P. Structural and thermodynamicproperties of water related defects in a-quartz. Phys Chem minerals2005;32(5e6):323e31.

[37] Yu Z, Xu Z, Lau D. Effect of acidity on chitineprotein interface: a moleculardynamics study. BioNanoScience 2014;4(3):207e15.

[38] Cui J, Yu Z, Lau D. Effect of acetyl group on mechanical properties of chitin/chitosan nanocrystal: a molecular dynamics study. Int J Mol Sci 2016;17(1):61.

[39] Floriano WB, Nascimento MAC. Dielectric constant and density of water as afunction of pressure at constant temperature. Braz J Phys 2004;34(1):38e41.

[40] Yu Z, Lau D. Development of a coarse-grained a-chitin model on the basis ofMARTINI forcefield. J Mol Model 2015;21(5):128.

[41] Chen Q, Liu J, Thundat T, Gray MR, Liu Q. Spatially resolved organic coating onclay minerals in bitumen froth revealed by atomic force microscopy adhesionmapping. Fuel 2017;191:283e9.

[42] Florin E-L, Moy VT, Gaub HE. Adhesion forces between individual ligand-receptor pairs. Science-AAAS-Weekly Pap Edition-including Guide Sci Inf1994;264(5157):415e7.

[43] Ackbarow T, Chen X, Keten S, Buehler MJ. Hierarchies, multiple energy bar-riers, and robustness govern the fracture mechanics of a-helical and b-sheetprotein domains. Proc Natl Acad Sci 2007;104(42):16410e5.

[44] Bell GI. Models for the specific adhesion of cells to cells. Science1978;200(4342):618e27.

[45] Helfand BT, Chang L, Goldman RD. Intermediate filaments are dynamic andmotile elements of cellular architecture. J cell Sci 2004;117(2):133e41.

[46] Walton EB, Lee S, Van Vliet KJ. Extending Bell's model: how force transducerstiffness alters measured unbinding forces and kinetics of molecular com-plexes. Biophysical J 2008;94(7):2621e30.

[47] Cranford SW, Ortiz C, Buehler MJ. Mechanomutable properties of a PAA/PAHpolyelectrolyte complex: rate dependence and ionization effects on tunableadhesion strength. Soft Matter 2010;6(17):4175e88.

[48] Dolan C, Tanner J, Mukai D, Hamilton H, Douglas E. Research report for

Y.L. Yaphary et al. / Composites Part B 131 (2017) 165e172172

evaluating the durability of bonded CFRP repair/strengthening of concretebeams. Washington DC: NCHRP; 2009.

[49] Karbhari VM, Ghosh K. Comparative durability evaluation of ambient tem-perature cured externally bonded CFRP and GFRP composite systems for

repair of bridges. Compos Part A Appl Sci Manuf 2009;40(9):1353e63.[50] Lau D, Qiu Q, Zhou A, Chow CL. Long term performance and fire safety aspect

of FRP composites used in building structures. Constr Build Mater 2016;126:573e85.