Reactive molecular dynamics and experimental study of graphene-cement composites: Structure, dynamics and reinforcement mechanisms

Accepted Manuscript Reactive molecular dynamics and experimental study of graphene-cement composites: Structure, dynamics and reinforcement mechanisms Dongshuai Hou, Zeyu Lu, Xiangyu Li, Hongyan Ma, Zongjin Li PII:

S0008-6223(17)30023-4

DOI:

10.1016/j.carbon.2017.01.013

Reference:

CARBON 11622

To appear in:

Carbon

Received Date: 15 October 2016 Revised Date:

4 January 2017

Accepted Date: 6 January 2017

Please cite this article as: D. Hou, Z. Lu, X. Li, H. Ma, Z. Li, Reactive molecular dynamics and experimental study of graphene-cement composites: Structure, dynamics and reinforcement mechanisms, Carbon (2017), doi: 10.1016/j.carbon.2017.01.013. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT

Reactive

molecular

dynamics

and

experimental

study

of

graphene-cement composites: structure, dynamics and reinforcement

RI PT

mechanisms

Dongshuai HOUa*, Zeyu LUb*, Xiangyu LIc, Hongyan MAd, Zongjin LIe

a. Professor, Corresponding author, Department of Civil Engineering, The Qingdao

SC

University of Technology, Qingdao, China; email address: [email protected]; Tel: +86 532 85071330;

M AN U

b. Post Doctor Fellow, Department of Civil and Environmental Engineering, The Hong Kong University of Science and Technology, Clear Water Bay, Hong Kong; [email protected];

c. Lecture, College of Arckitecture and Civil Engineering, Taiyuan University of Technology, Taiyuan, China; email address: [email protected];

TE D

d. Assistant Professor, Department of Civil, Architectural and Environmental Engineering, Missouri University of Science and Technology, Rolla, MO, USA e. Professor, Department of Civil and Environmental Engineering, The Hong Kong

EP

University of Science and Technology, Clear Water Bay, Hong Kong; email address: [email protected].

AC C

* Corresponding author.

Abstract

The remarkable properties of graphene and graphene oxide (GO) make it as an ideal candidate for high performance cement-based composites. This paper firstly investigated the effects of graphene and GO on the hydration, microstructures and mechanical properties of cement paste. The incorporation of 0.16 wt. % GO into the cement matrix can enhance the flexural strength of the material by 11.62 % due to the higher hydration degree, nano-filler effect and cracking-bridging effect. On the other

ACCEPTED MANUSCRIPT hand, graphene reduces the hydration development and mechanical behavior of cement paste due to its poor dispersibility in alkaline environments. Furthermore, the different interaction mechanisms between graphene/GO and cement hydrates have been deeply studied by reactive force field molecular dynamics (MD), revealing that

RI PT

the functional hydroxyl groups in GO provide non-bridging oxygen (NBO) sites that accept hydrogen-bonds of interlayer water molecules in the calcium silicate hydrate (C-S-H). In the presence of interface counter-ions, protons transfer from the -OH in GO to the NBO sites in C-S-H, which further contributes to the polarity of the GO

SC

surface and enhances the bonding with neighboring species. Besides the H-bond connections, the Ca2+ and Al3+ ions near the surface of C-S-H play a mediating role in

M AN U

bridging oxygen atoms in silicate chains and hydroxyl groups in GO, which increases the silicate chain length and heals the defective GO structure. Dynamically, the aluminate-silicate chains, calcium ions and functional hydroxyl groups establish the “cages”, and strongly prevent the freely diffusion of the interface water molecules, stabilizing the connections between C-S-H and GO structures. Finally, the uniaxial

TE D

tensile simulation indicated that while the high cohesive force and enhance plasticity in the GO modified cement composite is mainly contributed by the strong structural H-bonds and calcium aluminate skeleton, the weakest mechanical behavior of the

EP

graphene/C-S-H composite is attributed to poor bonding and the instability of atoms in the interface region. Reactive force field couples the chemical and mechanical responses that the water dissociation, protons exchange between C-S-H and GO

AC C

structures and silicate-aluminate-carbon network de-polymerization occurred to resist the tensile loading.

1.

Introduction

Graphene, the single-atom-thick sheet constructed by sp2-bonded carbon atoms, have drawn much attention due to its numerous unique properties, including excellent carrier mobility, outstanding tensile strength, extremely large specific surface area and high thermal conductivity [1]. During the past few years, graphene-based materials, such as graphene sheet and graphene oxide (GO), have been incorporated into the

ACCEPTED MANUSCRIPT matrix of various materials for a wide range of functional applications: extraordinary high-strength polymer nano-composites with graphene platelets [2], shape-memory and enzyme-sensing bio-materials incorporated by the graphene and amyloid fibrils [3] and high-performance thermally insulating materials developed by nanocellulose and

RI PT

GO [4]. In particular, the superior mechanical performance of graphene sheet (Young’s modulus: 1TPa, tensile strength 130GPa) makes it as an ideal candidate for advanced filler materials in cement-based composites [5]. Reinforcement of traditional cementitious materials with graphene can open up new pathways for

SC

improving the mechanical properties and durability of building materials, and can significantly reduce the consumption of traditional concrete materials. In respect to

M AN U

the sustainable development, it is of great importance and urgency to develop novel building materials, especially considering that the manufacture and application of concrete results in the release of 5~10% of global Green House Gas each year [6]. However, difficulties in dispersing graphene and the high cost of production limits its widespread application in construction and building materials. As a graphene

TE D

derivative, GO is mono-layer of sp2-hybridized carbon atoms modified by a mixture of carboxyl, hydroxyl and epoxy functionalities [7]. The oxygen functional groups, attached on the basal planes and edges of GO sheets, significantly alter the van der

EP

Waals interactions between the GO sheets and therefore improve its dispersion in water. The extraordinary mechanical properties, highly dispersible property in water

AC C

and lower cost, make GO a promising material for enhancing the mechanical properties of cement composites. Pan et al. [8] has shown that the addition of 1.0 wt. % GO could simultaneously improve the strength and toughness of GO-chitosan composites. This improvement has been attributed to the enhanced nanofiller-matrix adhesion/interlocking arising from the wrinkled surface and the two-dimensional geometry of the graphene platelets. Lv et al. [9] showed that cement composites exhibited a remarkable increase in tensile strength (78.6 %), flexural strength (60.7 %) and compressive strength (38.9 %) by incorporating 0.03 wt. % GO. Pan et al. [10] found that the introduction of 0.05 wt. % GO increased the compressive strength of

ACCEPTED MANUSCRIPT GO-cement composite by 15-33 % and the flexural strength by 41-59 %, respectively. Lu et al. [11] demonstrated that 0.05 wt. % GO led to 11.1% and 16.2% increase in compressive strength and flexural strength of cement paste, and also showed the effect of the hybrid GO/CNTs composites on the mechanical behavior of cement paste.

RI PT

Abrishami and Zahabi [12] have studied reinforcing graphene oxide/cement composite with NH2 functionalizing group and found that the flexural strength can be increased by 38.4% by compositing 0.1 wt.% GO in the cement paste. They attributed the strength enhancement to the porosity decreasing and improvement of interfacial

SC

strength between C-S-H and GO. Furthermore, the empirical force field molecular dynamics simulation has been performed to study the molecular-scale energetic,

M AN U

structural, and dynamic properties of the interface between surface functionalized graphitic structures and calcium-silicate-hydrate (C-S-H) [13] [14]. The simulation work provided molecular insights on the H-bonds interaction between C-S-H and different functional groups in GO structures. However, the empirical force field fixed the bonds in the functional groups on the GO and cannot allow the chemical reactions

TE D

between the C-S-H gel and GO. Hence, it is necessary to further investigate the interaction between GO and more realistic C-S-H model by reactive force field molecular dynamics.

EP

Many studies have been conducted to combine graphene/GO with cement-based materials [15] [16] [17] [18] [19], but they mainly focused on improving the

AC C

mechanical strength and transport behavior. The interaction mechanism between graphene, GO and C-S-H gel has not been understood comprehensively on the nanoscale level. In this study, on the one hand, hydration development, microstructure and mechanical behavior of graphene/GO modified cement-based materials will be characterized by using SEM, Raman and XPS techniques. On the other hand, reactive force field MD simulation will be utilized to investigate the C-S-H and graphene/GO interface properties, including surface energy, molecular structure and cohesive strength, which provide scientific guidelines for enhancing macroscopic properties of cement-based materials. The comprehensive study of the nanointeraction,

ACCEPTED MANUSCRIPT microstructure and macro-behavior of graphene/GO modified cement-based materials can help guide the sustainable design of materials using a fundamental approach. 2.

Methods and Characterization

2.1 Experimental methods

RI PT

The early-age hydration process of graphene/GO modified cement paste was measured by a Thermometric TAM Air C80 isothermal calorimeter and Renishaw RM 3000 Micro-Raman spectroscopy. For the thermometric test, cement was firstly

SC

weighted and then placed into the testing container. The pre-dispersed GO and graphene aqueous solutions were transferred into the testing needle tube, which was

M AN U

then placed on the top of the testing container and sealed. The container-needle tube filled with cement and aqueous solutions was put into the calorimeter setup for 24 h to make the system equilibrium. Then, the aqueous solutions in the needle-tube were injected to the container with internal-mixing for 8 min to ensure that cement particles were fully mixed with the aqueous solutions. The heat generated initially from the

TE D

beginning of mixing was captured and continuously monitored. In addition, Raman scattering of pre-milled samples was conducted using a 633 nm laser source. The microstructures of the composite that were pre-coated with golden films around 5 nm were observed using scanning electron microscope (SEM, JEOL 6390). X-ray

EP

photoelectron spectroscopy (XPS, Physical Electronics 5600 multi-technique system) was carried out to investigate the carbon status in the graphene and GO. Due to the

AC C

graphene and GO are dispersed in the aqueous solutions, it should be pretreated by freeze-drying method to obtain the powder-like or floc-like particle before XPS test. For the compressive strength test, the specimens were placed in a Materials Testing System (MTS 2000) and loaded at the speed of 1 kN/s until 10 % failure. For the three-point bending test, three specimens with dimensions of 150 mm × 30 mm × 10 mm were used with a span of 90 mm and a stroke control at a loading rate of 0.25 mm/min. Two linear variable differential transformers (LVDTs) were set up on each side of the specimen to measure the mid-point deflection.

ACCEPTED MANUSCRIPT 2.2 Simulation methods Force field: Molecular dynamics simulation, including the model construction and mechanical testing, was carried out by using the ReaxFF force field. The fundamental feature of ReaxFF is that the connectivity is not assigned for fixing for the covalent

RI PT

bonds such as Si-O, C-O and Al-O. Instead, bond orders, estimated by instantaneous atomistic distances, are updated continuously to simulate bond breakage and formation. The ReaxFF set of parameters employed in this study was created by merging the Si−O−H, C-O-H and Ca−O−H sets developed by van Duin et al [20] [21]

SC

and Manzano et al. [22]. The Reactive Force field has been widely utilized in water confined in carbon nanotubes [23], silica-water interfaces [24], calcium silicate

M AN U

hydrate gel [25] and nano-crystals [26]. The parameterization and performance of the force field for Ca, Si, O and H can be directly obtained from previous published reference data [21] [27].

Model construction: The simulation of the crystal structure is based on the 11 Å tobermorite structure determined by Hamid in 1981 [28] that is taken as the C-S-H

TE D

substrate. Cleavage of the tobermorite structure in the [001] direction leads to non-bridging oxygen (NBO, Si-O-). The silicate chains are assumed to be infinite with no defective part. The bridging oxygen (BO) and NBO from the silicate chains

EP

become the predominated elements in the channel. Some interlayer calcium atoms are also adsorbed near the silicate channels. In the simulation, the tobermorite super-cell contains 2 × 3 × 1 crystallographic unit cells. The dimensions of the super-cell are

AC C

a=22.32 Å, b=22.02 Å, c=22.77 Å and α=90°, β=90°, γ=90°. After completion of cleaving the super-cell, three models with different interlayer graphene and GO are constructed.

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Fig. 1 Model construction of Graphene/C-S-H, GO/C-S-H and GO/C-A-S-H composite. Left figure represent cleaved tobermorite; three figures in the middle are G and GO sheets; Right figures are the initial model of G/C-S-H, GO/C-S-H and

composite model.

TE D

GO/C-A-S-H before molecular dynamics. The ball-stick styles represent the The red, green, gray, yellow, white and purple balls represent

oxygen, calcium, carbon, silicon, hydrogen and aluminum atoms, respectively. The

EP

white-red stick represents the hydroxyl bond, the gray stick is the C-C bond, the yellow-red stick is the silicate bond.

AC C

The pristine graphite unit cell with lattice parameters a=2.46 Å, b=4.26 Å, c=3.4 Å and α=90°, β=90°, γ=90° was constructed, following previous research [14]. The cell was then replicated 9 times along the a direction and 5 times along the b direction, while the thickness remained as 3.4 Å. The graphene sheet, including 180 carbon atoms, was inserted into the interlayer region created by the cleaved C-S-H substrates. This first model is defined as graphene_CSH (G_CSH). Additionally, the GO structure was created by attaching 18 functional groups (–OH) on the graphene sheet, with a surface coverage of 10%. According to different degree of oxidation, the coverage ratio might vary and the 10 % coverage of –OH groups is in the range of the

ACCEPTED MANUSCRIPT typical surface treated carbon fiber materials published in reference [29] [30] [14]. The GO confined by the C-S-H substrates construct the second model defined as GO_CSH. To consider the interaction between the GO and Al rich C-S-H, the 16 Al atoms are introduced into the interlayer region. To ensure the charge balance, 24

RI PT

interlayer Ca atoms are removed. The third model is defined as GO_CASH. At the very beginning，all the atoms in the graphene/GO were located 5 Å away from either substrate surface. The model construction procedure is schematically shown in Fig. 1.

SC

The REAX package in LAMMPS [31] was employed to perform the reactive force field MD run. The trajectories of the atoms were calculated by the Verlet algorithm

M AN U

with a time step of 0.25 fs. Initially, the Nosé-Hoover thermostat algorithm was used to simulate the interaction between the C-S-H gel and graphene/GO in the NPT ensemble at 300 K and 1 atm for 250 ps. When the system reached equilibrium, it turned to the canonical ensemble at 300 K 100 ps. Finally, a further 1000 ps NPT production run was performed to obtain the atomistic trajectories for structural and

TE D

dynamics analysis.

Uniaxial Tension test: Three samples achieved by extending the model in last section by factors of 3, 2, 4 along the a, b and c axis, respectively, were subjected to uniaxial

EP

tensile strain through progressive elongation at strain rates of 0.08/ps. Hence, the three samples contain more than 22000 atoms, and the large number of atoms can ensure statistical stability during the tensile test. The tensile deformation of the

AC C

samples was obtained by a stepwise displacement of the atoms. During all the simulation period, an isobaric-isothermal ensemble was implemented using the Nose-Hoover thermostat, and the Verlet integration scheme. As the C-S-H tensioned along the y direction, the pressure in the x and z directions was coupled to zero, so as to consider the Poisson effect.

ACCEPTED MANUSCRIPT 3.

Results and discussion

3.1 Mechanical behavior The compressive and flexural strength of graphene and GO modified cement paste at 14 days are shown in Fig. 2. The experimental results indicate that

the compressive

RI PT

and flexural strength of GO modified cement paste reache 86.9 MPa and 8.66 MPa, which are 3.21 % and 11.62 % higher than that of control sample. Moreover, it is indicated that GO is more effective to enhance the flexural strength than compressive strength of cement paste which is attributed to not only the excellent mechanical

SC

properties of the GO itself, but also the crack-bridging effect of GO on controlling the micro-cracks in the cement matrix at the nanoscale level. As shown in Fig. 3, it is the

M AN U

first time to report that GO is trapped in both sides of the micro-crack ~500 nm in width, which can bridge the nano-crack and interlock the cement hydration products. More loads can thus be transferred and shared by the GO, which leads to the enhanced mechanical behavior of cement paste. However, 0.16 wt. % graphene reduces the compressive and flexural strength of the cement paste by 3.36 % and 10.59 %

TE D

compared with the control sample. A possible reason is that graphene with few functional groups creates little interaction with the hydration products of cement, such as C-S-H and Ca(OH)2. Graphene, as an isolated phase, has poor bonding with the

EP

cement matrix, leading to the decreased mechanical behavior. The molecular mechanism of the graphene/GO on the mechanical performance of cement hydrates is

AC C

further discussed in section 3.6.

Fig. 2 Compressive and flexural strength of graphene and GO modified cement paste.

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Fig. 3 SEM image of crack-bridging GO in cement matrix. 3.2 Hydration

The influence of graphene and GO on the cement hydration at the early stage (3 days),

TE D

monitored by thermometric TAM air, is shown in Table 1 and Fig. 4. It can be seen that the dissolution rate of cement in water increases with GO addition but decreases with graphene addition. GO modified cement paste (sample 2) shows the shortest time

EP

(3.8 min) to arrive the dissolution peak with higher accelerating rate and heat flow. However, graphene modified cement paste (sample 3) shows an obviously longer time

AC C

(9.0 min) to reach the dissolution peak, and the accelerating rate and heat flow are lower than that of sample 2. In addition, GO not only increases the dissolution rate of cement, but also accelerates the hydration process of cement. As shown in Fig. 4b, with the addition of GO, the peak of heat flow shifts to the left with increased magnitude, and more importantly, the magnitude increment of the 2nd peak (stands for C3A hydration) is higher than that of 1st peak (stands for C3S hydration), which means that C3A hydration is more affected by GO addition than C3S hydration. However, graphene modified cement paste shows a contrary tendency. The shoulder peaks shift to the right with decreased magnitude in heat flow, indicating that graphene is not beneficial for cement hydration. This is because graphene inevitably agglomerates in

ACCEPTED MANUSCRIPT the alkaline cementitious environment and therefore graphene might not be uniformly distributed in the cement matrix. Due to the hydrophobic property of graphene, the cement near the agglomerated graphene has less water for hydration and the hydration degree and rate will be decreased, and the more severe the graphene agglomeration,

RI PT

the more time delayed the cement hydration. In order to confirm the conclusion above, Raman spectroscopy was conducted to investigate the effects of graphene and GO on the hydration development of cement paste, as shown in Fig. 5. It can be seen that the peaks for Ca(OH)2, C3S/C2S and gypsum can be identified at Raman shifts of 500-800

SC

cm-1, 852 cm-1 and 1079 cm-1, respectively. Plain cement paste (Sample 1) shows an obvious characteristic peak of Ca(OH)2 due to the rapid hydration of C3S at the early

M AN U

age. However, the peak almost disappears after GO addition (Sample 2) because Ca2+ takes part in the chemical reaction with –COOH in the GO to form Ca(HCOO)2. Fig. 6a depicts the Ca 2p spectra of graphene modified cement paste. The fitted curves indicate that Ca2+ exists in terms of Ca(OH)2, CaCO3 and CaO. Graphene has no reaction with Ca(OH)2. However, Ca(HCOO)2, centered at 347.4eV, is detected in the

TE D

GO modified cement paste, as shown in Fig. 6b, which is the reaction product between Ca(OH)2 and –COOH in GO. Ca(OH)2 hydrated from C3S will be continuously consumed by the GO and more C-S-H can be generated, which is the

EP

major reason leading to the improved hydration rate of GO modified cement paste. In addition, it is clear that the graphene modified cement paste shows higher peak intensity of C3S/C2S but lower peak intensity of Ca(OH)2, which results from the

AC C

lower hydration rate of the cement. Fewer C3S hydrates to form C-S-H at the early age because graphene is a hydrophobic material with less functional groups, which is easier to agglomerate in the cement matrix, and thus reduces the hydration of cement paste. However, for the GO modified cement paste, firstly, the oxygen functional groups in GO make it more active in interacting with the cement hydration products, and a denser microstructure of cement paste can be achieved by GO addition. Secondly, the functional groups can absorb water molecules and make GO a nucleation site for cement hydration, namely, cement particles are more likely to react with water molecules adsorbed on functional groups in the GO, forming crystal

ACCEPTED MANUSCRIPT hydrate nuclei. Finally, C-S-H will be increasingly generated due to the consumption of Ca(OH)2 by GO addition. Table 1 Hydration exothermic rate and hydration heat of GO reinforced cement paste Time to dissolution peak (min)

1 2 3

4.9 3.8 9.0

Hydration heat (J/g)

RI PT

Sample

TE D

M AN U

SC

183 195 140

Fig. 4 Hydration heat evolution of the graphene and GO modified cement paste (a)

AC C

EP

dissolution peak, (b) shoulder peaks.

Fig. 5 Raman spectroscopy of (1) plain cement paste, (2) GO and (3) graphene modified cement paste.

RI PT

ACCEPTED MANUSCRIPT

SC

Fig. 6 XPS spectrum of (a) graphene and (b) GO modified cement paste. 3.3 Microstructures

M AN U

The microstructures of GO modified cement paste at 1 day are observed in Fig. 7. Fig. 7a shows the SEM image of two major cement hydration products, calcium hydroxide (Ca(OH)2, CH) and ettringite (Ca6Al2(SO4)3(OH)12·26H2O, Aft), in GO modified cement paste. EDX analysis shown in Fig. 7a and 7b reveals that C atoms are contained in both CH and Aft, indicating that CH and Aft are more likely to form on

TE D

the surface of GO, which confirms the nucleation effect of GO on the cement hydration. Fig. 8 indicates the early-age (3h) microstructures of graphene and GO modified cement paste. The crystals of the hydration products in graphene modified

EP

cement paste (Fig. 8b) are not as large and mature as in plain cement paste (Fig. 8a), and the microstructure of graphene modified cement paste is not as denser as that in

AC C

plain cement paste, which indicates that graphene reduces the hydration development of cement at the early age. However, the crystals become larger and more mature, and the microstructures are denser in GO modified cement paste, indicating that GO is beneficial for interlocking the cement hydration products and forming a compacted microstructure, which leads to the improved mechanical behavior of cement paste. The SEM results are consistent with the results obtained from calorimetric measurements.

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

Fig. 7 Hydration products in GO modified cement paste.

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Fig. 8 Microstructures of (a) plain cement paste, (b) graphene and (c) GO reinforced cement paste. To better understand the role that graphene/GO plays in reinforcing the cement hydrate, the interface model of C-S-H and graphene/GO has been constructed by

ACCEPTED MANUSCRIPT molecular dynamics. The molecular structure, reactivity, dynamics and cohesive strength of the model have been investigated systematically. 3.4 Molecular Structural Properties of graphene/GO and C-S-H model 3.4.1 Molecular Structure of graphene/GO and C-S-H

RI PT

Molecular structures of the composite of the C-S-H and graphene/GO are shown in Fig.9. Correspondingly, Fig. 10 demonstrates the intensity profile of the Ca, Si, O, H, Al and C atoms distribution along the z direction. For the C-S-H structure, the silicate

SC

chains rooted in the calcium sheet form the calcium silicate layers. The directly connected silicate tetrahedron is defined as paired (P) and the silicate tetrahedron

M AN U

connecting two paired tetrahedrons is defined as the bridging tetrahedron (B). The water molecules, hydroxyl groups and the interlayer calcium ions are located near the surface of the bridging tetrahedron or in the vicinity of the channel between neighboring bridging tetrahedrons. Differing from the initial structure of tobermorite 11 Å without hydroxyl groups, there are some silicate hydroxyl (Si-OH) and calcium hydroxyl (Ca-OH) groups produced after the reactive simulation. Some of the water

TE D

molecules near the layer are dissociated into H+ and OH-. While the H+ is associated with the non-bridging oxygen atoms in the bridging tetrahedron, the OH- is bonded with the interlayer calcium atoms [32]. Both the non-bridging sites and adsorbed

EP

calcium atoms contribute to the hydrophilic characteristic of the C-S-H surface [33]. Interestingly, it can be clearly observed from Fig. 9a and Fig. 9b that the GO/C-S-H

AC C

interface has more Si-OH groups than that of the graphene/C-S-H one. It indicates that proton transferring might occur between the C-S-H and GO, and is further explained in the following section. Despite of the presence of the G/GO in the interlayer region, the silicate chain with infinite length in the tobermorite crystal maintained the calcium silicate structure in the ordered layer state. In the real C-S-H gel, the calcium silicate layer might be disturbed in some extent as the silicate chains turn defective.

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Fig.9 Molecular structure of the (a) G/C-S-H; (b) GO/C-S-H (c) GO/C-A-S-H in the equilibrium state after molecular dynamics simulation. The meanings of the

AC C

EP

TE D

ball and sticks with different color are illustrated in Fig.1 caption.

Fig. 10 Density profile for the atoms along z direction (a) G_CSH; (b) Go_CSH; (c) Go_CASH; (d) the crystal parameter c (cell size along z direction)

ACCEPTED MANUSCRIPT

The graphene and GO confined in the interlayer region of the C-S-H gel demonstrates different morphologies. In the graphene, the carbon atoms, orderly arranged in the xy plane, are distributed away from the C-S-H surface. No extension along the z

RI PT

direction can be observed in the graphene/C-S-H interface in Fig. 9a. As shown in Fig.10b, the sharp peak of the C atoms with a few overlapping with hydrogen atoms in the C-S-H gel, also confirms the hydrophobic nature of the graphene sheet. On the other hand, as shown in Fig.9b, the protruding hydroxyls on both sides of the GO

SC

point towards the C-S-H interface. The C-C bonds in the GO are stretched in the z direction and the graphene sheet is disturbed to some extent. In Fig. 10b, the

M AN U

distribution of hydrogen atoms growing through the interlayer region suggests improved interaction between the C-S-H and GO. The oxygen-containing groups can weaken the repulsive effect between the single-atomic carbon layer and the C-S-H gel and reduce the interlayer distance, as shown in Fig. 10d. Furthermore, the aluminum atoms in the interface region can greatly change the GO structure greatly. As shown in

TE D

Fig. 9c, the ordered two dimensional GO sheets are transformed to the amorphous three dimensional carbon network structures, losing the repetitive pattern of hexagons for the carbon atoms. The broad distribution of the carbon atoms in Fig.10c implies an

EP

expansion of the interlayer region. It can be observed that the aluminum atoms, forming the Si-O-Al-O-C connection, play the role in bridging the silicate tetrahedron and the carbon hexagons. This enhances the interaction between the C-S-H and GO. A

AC C

healing effect for the Al atoms in the defective silicate chains has also been found in the C-A-S-H gel [34]. Because GO sheet is strongly stretched by the neighboring atoms in the z direction, the interlayer distance increases to around 18.8 Å, as seen in Fig. 10d.

3.4.2 Local structure of the graphene and GO in the interlayer The spatial correlation for the carbon atoms in the graphene/GO can be further characterized by the radial distribution function of C-C and C-O. As shown in Fig. 11a, in the short range, there are three peaks with decreasing intensity located at 1.44

ACCEPTED MANUSCRIPT Å, 2.52 Å and 2.91 Å in the RDF of the C-C for graphene, representing the C-C bond length and typical diagonal distances in the same regular carbon hexagon. The second and third peaks emerge to a broader one in the C-C spatial distribution for the GO, mainly resulting from distortion of the carbon hexagons and defections in the

RI PT

structure. In the medium range, in the graphene, the local maxima of RDF are positioned at 3.81Å, 4.33 Å and 5.07Å, indicating the strong spatial correlation between neighboring carbon hexagons. By geometric calculation, the relative atomistic positions in two different hexagons also confirm that the carbon atoms are

SC

in the same plane in a range around 5 Å. However, the weaker peak intensities of the GO suggests that the confined GO cannot remain in a flat state even in the medium

M AN U

range. The neighboring hexagons rotate around their shared edge and the torsion results in the wrinkles of the GO structures, which can be clearly observed in Fig. 9b and 9c. Furthermore, the intensity peaks of the graphene can extend to around 9 Å, implying that the structural order can persist for a long distance. This can be attributed to the two dimensional extension of the graphene structure. In the long range ( > 6 Å),

TE D

no obvious spatial correlation can be observed in the GO that shows amorphous

AC C

EP

branch structures being ultra-confined in the interlayer region.

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Fig. 11 (a) radial distribution function (RDF) for the C-C atoms in graphene/GO systems; (b) RDF for the C-O atoms in graphene/GO systems; (c) short range RDF for C-O atoms in GO systems; (d) molecular structure of the C-O and C-OH bonds in GO; (e) schematically description for the C-OH dissociation. The

TE D

meaning of different colored ball and sticks is in Fig.1 caption.

The RDF of the C-O atoms can further describe the local structures of the oxygen-containing groups grafted in the graphene. As expected in Fig.11a, the RDF

EP

intensity remains at a zero value until the distance reaches 2.5 Å in the graphene and C-S-H interface. In the range from 2.5 to 10 Å, no intensity maximum is detected in

AC C

the RDF, implying no spatial correlation with the neighboring oxygen atoms in the silicate hydroxyl and water molecules near the surface of C-S-H. On the other hand, the oxygen atoms bonding to the GO result in pronounced peaks positioned at 1.41 Å, 2.54 Å and 3.76 Å in the RDF, which reflect the direct C-O covalent bond, indirect C-C-O and the C-C-C-O connections in the single benzene-ring like structure. It should be noted that a C-O RDF from 1.3 to 1.8 Å demonstrates bimodal distribution in Fig. 11c, indicating two different C-O bond types. As shown in Fig. 11d, the first type is C-O with a shorter length around 1.37 Å and the second type is the hydroxyl groups connected with carbon atoms, with bond length around 1.43 Å. Additionally,

ACCEPTED MANUSCRIPT while the first peak height in RDF for the GO/C-S-H is quite lower than that of the second one, the heights of the double peaks for the GO/C-A-S-H show little difference. It means that the C-OH groups occupy a predominant percentage in the GO/CSH, yet in the GO/CASH, the ratios of C-O and C-OH are similar. Since no C-O

RI PT

groups were prepared at the beginning of the model construction, they were all produced after the reactive simulation process. The chemical reaction pathway can be illustrated in the following sequence in Fig. 11e: the interlayer calcium or aluminum ions diffuse and approach the C-OH groups due to electronic attraction; the

SC

electronegativity of the oxygen atoms are shared with neighboring Ca and Al atoms and the O-H bond is stretched; the C-OH dissociates into H atoms and the C-O group;

M AN U

the de-bonded hydrogen atom diffuses and associates with the non-bridging oxygen atom in the silicate tetrahedron and forms the Si-OH connection. Besides, the free hydrogen atoms can also associate with the calcium hydroxyl groups, producing water molecules. This reaction explains the observation that more silicate tetrahedrons are protonated in the C-S-H surface in the presence of GO. In the current simulation, -OH

TE D

groups on the GO surface can dissociate into H+ and –O- , resembling the ionization in phenol where there is an equilibrium between the –OH, H+, and –O- groups in an aqueous solution [35]. Under the conjugated structure of un-oxidized benzene rings, the –O- groups could exist in a stable manner by sharing the electrons. The ionization

EP

degree can be improved by the reaction between H+ and hydroxyl groups in the C-S-H gel. Compared with calcium atoms, the aluminum species are more likely to form

AC C

stable Al-O-C covalent bonds, which can accelerate the proton transfer between the GO and C-S-H surface. In previous research, Sanchez et al. [14] investigated the interaction between functionalized graphitic structures and calcium silicate hydrate and found that the affinity to C-S-H was closely related with the polarity of the functional group. In current study, the de-protonation reaction can contribute to the high polarity of the GO system, which to some extent enhances the interaction with the C-S-H structures. 3.4.3 Local structure of interlayer Ca and Al ions The molecular structure of the GO in the vicinity of the C-S-H surface is greatly

ACCEPTED MANUSCRIPT affected by the ions and hydroxyl groups in the interlayer region. Fig. 12a exhibits the RDF for the Caw (calcium atoms in the interlayer region) and Oc (oxygen atoms bonded to carbon) in the samples of GO/CSH and GO/CASH. The sharp peaks in both cases are located at 2.67 Å, implying a stable Ca-O connection in the interface.

RI PT

Meanwhile, the oxygen atoms in the samples are further categorized into oxygen atoms in the C-S-H gel (Os) and Oc. In Fig. 12b, strong correlations for both the Al-Oc and Al-Os have also been observed at distance 1.92 Å, which represents the Al-O

SC

covalent bond.

The local structures of the Al and Ca atoms are also analyzed by the coordination

M AN U

numbers (CN). As shown in Fig. 12c, on average, the Ca atoms have around 5.9 nearest oxygen neighbors, including 4.4 Os and 1.5 Oc. A CN of nearly 6 has been widely found in silicate glass [36] and cement-based materials [37] in previous investigations. The 6-coordinated Ca-O clusters are more likely to form disordered octahedrons. On the other hand, 2.2 Os atoms and 1.8 Oc atoms contribute to 4

TE D

coordinated atoms of Al species. Al and O atoms aggregate to build the aluminate tetrahedron. The aluminate tetrahedron structures occupy predominant percentage in the C-A-S-H gel, which matches well with recent NMR characterization on the

EP

silicate-aluminate structures [38]. Hou et al. [39] and Qomi et al. [34] studied the local structure of Al species present in the interlayer region of C-S-H gel by molecular dynamics. Both studies found that Al species can associate with non-bridging sites of

AC C

the defective silicate chains, constructing 5- and 6-fold polyhedrons. Since the infinite long chains in the tobermorite 11 Å have fewer non-bridging atoms, the CN of Al species reduces to around 4 in the current research. As illustrated in Fig. 12d, the counter-ions in the vicinity of the C-S-H surface play an essential mediating role in bridging the coordinated atoms in the C-S-H structure and GO.

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

EP

Fig. 12 (a) Radial distribution function for Ca-O (b) RDF of Al-O (c) coordination number of Ca-O and Al-O , and local molecular structure of Al-O and Ca-O

AC C

cluster (d) molecular structure of Os-Al-Oc and Os-Ca-Oc connecting the

neighboring graphene oxide and C-S-H gel. The meaning of different colored ball and sticks is in Fig.1 caption.

3.4.4 The local structure of water and hydroxyl groups In addition to Ca-O and Al-O bonds, the H-bonds are important components for linking the C-S-H and GO structures. The RDFs of Oc-H and Os-H are utilized to characterize the structural feature of the H-bonds network in the interfacial region. In the interface between graphene and C-S-H, as exhibited in Fig. 13a, the first peak in

ACCEPTED MANUSCRIPT the RDF of Os-H, distributed from 1.4 to 2.0 Å, reflects the H-bonds interaction between the silicate chains and neighboring water molecules. The H-bond is defined according to two requirements: bond length is smaller than 2.45 Å and the minimum donator-hydrogen-acceptor angle is 150o [40]. As shown in Fig.13b, the water

RI PT

molecules or the decomposed hydroxyls can either accept the H-bonds from the Si-OH or donate the bonds to the bridging oxygen atoms in the silicate skeleton. These H-bonds are defined as the structural H-bonds in the C-S-H gel that are dramatically different from the H-bonds between neighboring water molecules in the

SC

bulk water solution [22]. The silicate chains provide a large number of H-bonds accepting sites for the water molecules, which can further restrict the mobility for

M AN U

hydroxyl groups. On the other hand, as compared with the shallow distribution for the H-bonds in the silicate chains, the first peak for the RDF of Oc-H shifts towards a longer distance with a broader distribution from 1.48 to 2.62 Å. Three types of H-bonds can be observed in Fig. 13c: the de-protonated C-O accepts bonds from the hydrogen in Si-OH (C-O---HO-Si); the neighboring hydroxyl rooted in the GO

TE D

interact with each other (C-OH---C-OH); the carbon hydroxyl donates H-bonds to calcium hydroxyl (C-OH---OH-Ca). Three types of H-bonds show different bond lengths: C-O---HO-Si (1.77 Å) < C-OH---OH-Ca (1.99 Å) < C-OH---C-OH (2.36 Å).

EP

It indicates that the H-bonds properties are greatly influenced by the local environment. A similar H-bonds structure can also be observed in the Al rich region. As exhibited in Fig. 13d, the hydroxyl in the silicate tetrahedron accepts H-bonds to

AC C

neighboring aluminate tetrahedron, and meanwhile acts in the role of H-bonds donor attracting the C-OH groups. In this respect, the first and third types are important chemical bonds for bridging the C-S-H gel and GO.

Dipole moment distribution of a single water molecule has been employed to study the interaction between water molecules and their C-S-H or graphene/GO substrates quantitatively. The dipole moment is sensitive to the local environment of the water molecules: while the hydrophilic confinement contributes to a higher value of dipole moment, the hydrophobic confinement tends to a lower value [33]. The dipole

ACCEPTED MANUSCRIPT moment has been calculated according to the methodology mentioned by Manzano et al. [22]. In previous ReaxFF simulation on the bulk water solution [41], the dipole moment for bulk water was found to be around 2.15 D and

can be taken as reference.

It can be observed from Fig. 13e that the average values for the dipole moment in all

RI PT

cases are significantly larger than 2.15 D. It indicates that the environment in which the water molecules located is hydrophilic in both graphene/C-S-H and GO/C-S-H.

As compared with graphene/C-S-H, the dipole moment distribution of water in

SC

GO/C-S-H shifts towards a larger value, with the mean value rising from 2.48 D to 2.54 D. It confirms the hydrophilic nature of the functionalized C-OH groups in GO.

M AN U

When the water molecules are confined between the GO and C-S-H, the upshift for the dipole moment distribution is mainly attributed to the electronic field induced by the oxygen-containing groups and silicate framework that polarize the confined water molecules to some extent by stretching the O-H bonds, bending the H-O-H angles and changing the charges [42]. It should be noted that the carbon basal sheet still consists un-oxidized

benzene

rings,

which

are

hydrophobic.

The

hydrophilic

TE D

of

oxygen-containing groups and hydrophobic benzene rings contribute to the am-hydrophilic property of GO and the am-hydrophilicity is influenced by the local

EP

PH and chemical compositions [43]. It can be observed that there is a further positive shift toward large values of dipole moment as the Al atoms are present in the interfacial region. This enhancement can be explained by the following two reasons.

AC C

First, the effect of the interlayer cations on the dipole moment has been investigated in previous research, and it was found that Ca2+ and Al3+ ions contribute to the hydrophilic nature of the C-S-H gel [33] and Al species provide more positive charges for the polarizing water to balance them. Secondly, as discussed in the previous section, the Al species result in proton dissociation from C-OH, increasing the polarity of the functional group in GO. A GO framework with more polarity also leads to a hydrophilic feature.

These H-bonds, together with the ionic-covalent Al-O and Ca-O connections

ACCEPTED MANUSCRIPT construct a dimensional interfacial network of chemical bonds. The stability, strength and the mechanical contribution for different bonds is discussed in the following

AC C

EP

TE D

M AN U

SC

RI PT

sections.

Fig. 13 (a) RDF of Os-H in the G/CSH and Oc-H in Go/CSH, Go/CASH (b) molecular structure of the structural H-bonds network (c) molecular structure of three types of H-bonds between C-S-H and graphene oxide (d) H-bond network in the Al-rich region; (e) dipole moment distribution for water molecules in graphene/C-S-H, GO/C-S-H and GO/CASH. The meaning of different colored ball and sticks is in Fig.1 caption.

ACCEPTED MANUSCRIPT

3.5 Dynamic Properties of the graphene/GO and C-S-H model 3.5.1 Dynamic properties of carbon atoms

RI PT

The dynamic properties of the atoms in the graphene/GO and C-S-H gel are first quantitatively evaluated by mean square displacement (MSD) that is defined by the following Eq. 1:

(1)

SC

MSD(t) =< |  (t) −  (0)| >

The variation of the MSD value with time t is the statistical average for the square of

M AN U

the spatial difference between the position at current time t and the initial position at time zero for the specific atoms (i ranges from 1 to total number of investigated atoms). Considering the 2-dimensional structure of graphene and the interlayer space, the MSD (t) is further decomposed to MSDxy and MSDz, characterizing the movement of atoms parallel and normal to the C-S-H surface. Fig.14 records the MSD of carbon

TE D

atoms in the graphene/GO structure during 100 ps. As shown in Fig. 14a, in the graphene/C-S-H system, MSD continues increasing to 9 Å2 during the initial 3 ps; subsequently drops to 2.6 Å2 and fluctuates at this value until 30 ps; finally jumps to

EP

around 12 Å2 and reaches a plateau in the last 60 ps. Fig. 14b shows the MSDxy and MSDz for carbon atoms in graphene/C-S-H. The variation of MSDxy is almost simultaneous with that of MSD, implying that the movement of C atoms in the xy

AC C

plane plays a predominant role. On the contrary, the ultra-confinement restricts the mobility of C atoms in the z direction, resulting in lower MSDz values (< 1 Å2). The dramatically large fluctuation of the MSD between 2.6 Å2 and 12 Å2 reflects that the graphene oscillates in the domain of 3 Å, which is the typical size of a single carbon-hexagon cell.

The contour map of the atomistic positions in the xy plane are shown in Fig. 14c in order to better understand the dynamic behavior of the C atoms in the nanometer channel. The trajectories of neighboring C atoms completely overlap and only the

ACCEPTED MANUSCRIPT hexagon-like organization with a cavity in the center can be distinguished. It means that during 100 ps, the C atoms frequently exchange positions with their bonded atoms in the short range but cannot escape the cage in the middle range. The swinging behavior of the graphene structure in the xy plane is mainly attributed to the

RI PT

heterogeneous repulsive effect from the bottom and top surfaces for the C-S-H gel. Since no bonds are built in the interfacial region, the unstable graphene cannot remain

AC C

EP

TE D

M AN U

SC

in a fixed position for a long time.

SC

RI PT

ACCEPTED MANUSCRIPT

Fig. 14 (a) Mean square displacement for C atoms in graphene/CSH, GO/CSH

M AN U

and GO/CASH; (b) MSDxy and MSDz for C atoms in graphene/CSH; (c) trajectories of the atoms in xy plane in graphene/CSH; (d) in GO/CSH; (e) in GO/CASH.

On the other hand, the C atoms in the graphene oxide show different MSD evolution as compared with those in graphene. After the jump in the first 0.3 ps, the MSD

TE D

slowly climbs to 1~2 Å2 with slight fluctuations. The extremely low MSD values, even at 100 ps, indicate that C atoms cannot escape from the cage built by their coordinate atoms. The trajectories for the C atoms in GO are shown in Fig.14d and

EP

14e. Almost all the carbon atoms are distributed separately, without any overlapping with other neighboring atoms. Hence, they can only vibrate or rotate at the fixed

AC C

positions. The discrepancy between the dynamic properties of C atoms in graphene and GO indicates that functional groups in the GO can significantly improve the chemical stability of the C atoms near the C-S-H surface. Additionally, during 100 ps, the MSD values of C in GO/CASH are slightly slower than those of C in GO/CSH. It suggests that the presence of Al species in the interlayer region can further stabilize the bonding with GO. 3.5.2 Dynamic properties for hydrogen atoms The MSDs of different atoms in GO/CSH have been calculated and are shown in Fig. 15a. According to the MSD values at 100 ps, the mobility of atoms is ranked in the

ACCEPTED MANUSCRIPT following order: H > C > Ca > Si. MSD values below 1 Å2 for Si and Ca atoms indicate that the calcium silicate sheet plays skeletal role in stabilizing the whole structure. In the graphene oxide, the C atoms can be categorized into “immobile” atoms that are directly bonded with C-S-H and “mobile” ones, away from the

RI PT

interfacial chemical bonds. Compared with the stable C-S-H skeleton, the slightly higher MSD values of the C atoms are mainly attributed to the movement of the “mobile” atoms. In the interlayer region, the highest MSD values of the H atoms suggest that H atoms are the most unstable species in the C-S-H gel. Not only can

SC

some interlayer water molecules or hydroxyl groups diffuse in silicate channel, but

AC C

EP

TE D

M AN U

the protons can transfer from C-OH to the Si-OH, as discussed in the previous section.

Fig. 15 (a) MSD for the Ca, Si, C and H atoms in GO/CSH; (b) MSD for the H atoms in graphene/CSH, GO/CSH and GO/CASH; (c) schematically description

ACCEPTED MANUSCRIPT of water transport between graphene and C-S-H; (d) the trajectories of hydrogen atoms in xy plane for graphene-CSH (left) and GO-CSH (right). The meaning of

RI PT

different colored ball and sticks is in Fig.1 caption.

Additionally, as shown in Fig. 15b, the MSD values of the hydrogen atoms show large discrepancy for the water molecules or hydroxyl groups confined by graphene and GO. In the graphene, during 100 ps, the MSD continuously increases to around

SC

20 Å2, that is more than five times the value in GO. As seen in Fig. 15c, the dynamic behavior of the water molecules can be considered as the single-file water transport

M AN U

between hydrophobic graphene and the hydrophilic C-S-H substrate. Meanwhile, the trajectories of the hydrogen atoms, shown in Fig.15d, indicate that some water molecules can diffuse in the xy plane of the graphene as far as 8 Å, but most of hydroxyl groups near the GO surface can only vibrate at the original positions.

TE D

The dynamic properties of the hydrogen atoms can be further investigated by the self-part of the van Hove function that is expressed in Eq. 2 

 (, ) = ∑ 〈( − | () −  (0)|)〉 

(2)

EP

The physical meaning of 2Gs(r,t) is the probability that particle i has moved a distance r in time t. Fig.16 a and b demonstrate the van Hove functions of water

AC C

confined in C-S-H gel with graphene and GO substrates at different times. As shown in Fig. 16a, at short time scale (t < 1ps), the sharp peak located at 0.5 Å means there is a restriction effect from the cage. The average displacement shifts toward 0.9 Å at time 2.5 ps. At 25 ps, an clear intensity increase can be observed at 2 Å and 4 Å in Gs(r,t). While the former length is a typical value for OH bonds vibrating at fixed Si-O- sites or the rotation around a fixed axis, the latter one is slightly larger than the length of one water molecule (~3 Å). Another intensity peak at 45 ps indicates that the hydrogen atoms can travel as far as 8 Å, which is consistent with the observation in the contour maps in Fig. 6d. It means that some water molecules have diffusive

ACCEPTED MANUSCRIPT behavior between two neighboring silicate tetrahedrons at very short simulation times. The long distance diffusion is not possible for the water molecules confined in C-S-H gel is so short a time, due to strong attraction from the non-bridging oxygen sites and the over-solvated counter-ions [32], and the obstacles from the protruding bridging

RI PT

silicate tetrahedron [44]. Nevertheless, the graphene sheet does contribute to the transport of hydrogen atoms. Previous experimental study [45] and molecular simulation [46] proposed that water transport in the carbon nanotubes (CNTs) or in the vicinity of graphene is a fast phenomenon. The velocity accelerating effect of

SC

CNTs and graphene is due to the less friction interior surface that contributes to the super-lubricity for the water transport [47]. In this respect, the transport process for

M AN U

the water molecules is described as the following circle: molecules escape from the energy barrier of the C-S-H substrate, diffuse rapidly near the graphene surface and return to the C-S-H substrate. On the other hand, as time evolves, the shoulder located around 2Å gradually increases in the Gs(r, t) distribution for the movement of H atoms near the GO. The H atoms accumulation in the short distance domain implies that the

TE D

aluminate-silicate chains, calcium ions and C-OH groups construct the cages, strongly

AC C

EP

preventing the freely diffusion of the H atoms.

Fig. 16 Gs(r,t) distribution of H atoms in (a) graphene/CSH; (b) GO/CSH at different simulation time from 0.25 ps to 45 ps.

3.5.3 Time correlation function for chemical bonds In the molecular structural analysis mentioned above, the H-bonds, Ca-Oc and Al-Oc are important connections that link the GO and C-S-H gel, and it is necessary to

ACCEPTED MANUSCRIPT explore the strength and stability of the chemical bonds. The time correlated function (TCF) is utilized to describe dynamical properties of various chemical bonds. The TCF of a bond is described in Eq. 3:

 !(") !(#)$  !(#) !(#)$

(3)

RI PT

() =

where δb(t) = b(t)−< ' >, b(t) is a binary operator that takes a value of one if the pair (e.g. Ca-O) is bonded and zero if not; and is the average value of b over all

SC

simulation times and pairs. The chemical bonds break and form during the time evolution, which results in connectivity variation in the C-S-H gel. If the connectivity

M AN U

persists unchanged, the TCF of the bonds will maintain a constant value one. Otherwise, the breakage of the bonds leads to lower the TCF, and the more frequently the bond breakages occur, the lower the value of TCF. By comparing the deviations

AC C

EP

TE D

from one in the TCF curves, the stability of the various bonds can be estimated.

Fig. 17 (a) Time correlated function for H-bonds in graphene-CSH, GO-CSH and

ACCEPTED MANUSCRIPT GO-CASH. (b) TCF for different bonds in GO-CSH; (c) TCF for different bonds in GO-CASH.

As shown in Fig. 17a, with progressively increasing time, the TCF of H-bond in

RI PT

G and Go gradually reduces to 0.6 and 0.7, respectively. As compared with the G, the H-bond stability is enhanced to some extent in the Go system. This matches well with small MSD values of the hydrogen atoms near the Go surface that remain in fixed positions for a long time. Furthermore, the H-bonds can be

SC

divided into those connected with oxygen atoms in the C-S-H gel (Os-H) and those with oxygen groups in the GO sheet (Oc-H), as mentioned in the above

M AN U

section. As shown in Fig. 17 b and 17c, the bond strength of Oc-H is stronger than that of Os-H in both GO-CSH and GO-CASH, in respect of bond stability. It is worth noting that the H-bond lifetime of the C-OH or C-O- groups is longer than that of the H-bonds of the structural water molecules. In addition, it was also found that the H-bond lifetime in the GO system was favored by the increasing of

TE D

the GO polarity in previous molecular simulation [14]. On the contrary, according to the reduced value of TCF in Fig. 17b, the bond strength of Ca-Oc is weaker than that of Ca-Os. Combined with the coordination analysis in the structural

EP

section, Oc only occupies one sixth of the nearest neighbors to the Ca atoms. The bond stability of Ca-Oc is more likely influenced by other coordinates. Differing from the continuously reducing trend in TCF for the Ca-O and H-bonds, the TCF

AC C

for the Al-O bonds remains constant during the simulation time, suggesting the strong strength of the covalent bonds. It resembles the aluminate-silicate skeleton in the C-A-S-H gel or in some zeolite crystals [34]. In summary, based on bond strength, the chemical bonds connecting the C-S-H and Go structures rank in the following order: Al-O > Ca-Os >Ca-Oc >Oc-H> Os-H. These chemical bonds play

an essential role in withstanding the loading. The mechanical contribution of the different bonds is discussed in the following section concerning mechanical testing.

ACCEPTED MANUSCRIPT 3.6 Reinforcement mechanism of G/Go on C-S-H 3.6.1 Stress-strain relation and molecular structural evolution The stress-strain relation can distinguish the different mechanical behaviors of the graphene/GO reinforced C-S-H gel under tension loading. As shown in Fig. 18, the

RI PT

stress fluctuates around 0 GPa with the increasing tensile strain for the graphene/C-S-H sample. In order to validate the accuracy of the reactive force field in mechanical properties investigation, pure tobermorite 11 Å and pure graphene have been uniaxial tensioned, separately. The stress-strain relation has been analyzed in the

SC

supplementary material. As expected, the graphene without chemical bonding with the surrounding C-S-H substrates provides little mechanical contribution. This is also due

M AN U

to the instability of the graphene confined in the interlayer region. Without bonding with C-S-H, the reinforced role of graphene on cementitious materials is quite limited in respect of some physical interactions such as the enhancing friction and filling the large voids.

TE D

On the other hand, as shown in Fig.18 for the GO/C-S-H sample, the stress continues increasing to 1.5 GPa as the strain reaches 0.1 Å/Å, and subsequently reduces slowly in the post-failure regime. The tensile strength is within the range of previous findings

EP

on the cohesive force in C-S-H gel at the nano-scale (~3GPa) [48]. The functional groups C–OH and –C-O bridge the C-S-H substrate and Go together to resist the mechanical loading rather than the two independent species in the G/C-S-H sample. It

AC C

is worth noting that the post-failure region of the stress-strain relation (from 0.1 to 0.6Å/Å) is elongated to some extent due to the intrusion of GO as compared with that of the tensioned C-S-H gel in previous molecular dynamics study (from 0.1 to 0.4Å/Å) [32]. In particular, it can be observed the ladder-like stress drop at strain 0.4 Å/Å, which is a typical plasticity feature of polymers. It implies that the brittleness of the C-S-H gel is greatly improved along the interlayer direction. The uniaxial tension simulation at nano-scale level provides molecular insights on the interfacial strength between C-S-H and graphene or GO. It explains the mechanical difference between GO/cement and G/cement composites observed by the compressive and flexural

ACCEPTED MANUSCRIPT experiments in section 3.1. The functionalizing greatly improved the mechanical properties of GO/cement composite because of the interfacial strength between functionalized GO nanosheets and the C-S-H gel, which further confirms the crack-bridging behavior of GO in the composite as shown in Fig. 3.

Go_CASH Go_CSH G_CSH

-2.0 -1.5

SC

Stress (GPa)

-2.5

RI PT

-3.0

-1.0

0.0 0.0

M AN U

-0.5

0.2

0.4

0.6

0.8

Strain (Å/Å)

Fig. 18 Stress-strain relation for the G-CSH, GO-CSH and GO-CASH tensioned along the z direction.

TE D

Furthermore, the stress-strain relation in the tensioned GO/CASH sample can be divided into two stages in the stress increase process: as the strain rises to around 0.1 Å/Å, the stress-strain for the GO/CASH coincides with that of GO/CSH; from 0.1

EP

Å/Å to 0.5 Å/Å, the stress climbs to 2.5 GPa more slowly, with a smaller slope. Two-stage stress-strain means reduction of the modulus after the turning point 0.1 Å/Å, but both the cohesive strength and plasticity of the C-S-H gel are significantly

AC C

improved. It indicates that the Al species in the interlayer space further helps in bridging the C-S-H and GO structures in order to bear the loading more effectively.

To better understand the stress-strain evolution, it is necessary to monitor the morphology changes for GO/C-S-H samples during the tension process. As shown in Fig. 19a, as the strain increases from 0 to 0.1 Å/Å, the structure is elongated slightly in the interlayer region due to the stretching of the Ca-OH-C bonds and the H bonds. In the early post-failure region, bonds breakage is widely observed at the GO/C-S-H interfacial region, resulting in the birth of small cracks. At strain 0.4 Å/Å, while part

ACCEPTED MANUSCRIPT of the C-OH and C-O- bonds are pulled out from the C-S-H surface, some bonds are still deeply rooted between the silicate chains and surface calcium atoms. The GO structure is extended along z direction, bridging the gap between the neighboring C-S-H substrates. The GO structural extension can slow down the crack growth,

RI PT

preventing the sample from fracturing suddenly, and enhance the plasticity of the

AC C

EP

TE D

M AN U

SC

structure, which explains the ladder-like stress reduction in the stress-strain relation.

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Fig.19 Molecular structure evolution at z-direction tensile strain 0, 0.1, 0.2, 0.4 and 0.6 Å/Å (from top to bottom) for (a) GO/C-S-H; (b) GO/C-A-S-H composites. The meanings of the ball and sticks are illustrated in Fig. 1 caption. Compared with the morphological evolution of the GO structure, most of the interlayer calcium ions remain in the silicate surface and the calcium silicate sheet

ACCEPTED MANUSCRIPT maintains an ordered layer state during the tension process. Additionally, the structural evolution of the GO/CASH is shown in Fig. 19b. There are two obvious structural discrepancies for C-S-H and C-A-S-H during tension process. First, at the later stage of the stress rising from 0.1 to 0.4 Å/Å, the straight arrangement of the

RI PT

silicate chains in the layered structure is disturbed to some extent, and the deformation of the calcium silicate sheet and GO happens simultaneously. It means that damage occurs not just at the interface but is also transferred to the C-S-H skeleton. The deformation transference also indicates re-arrangement of the GO structure and is

SC

further discussed in the next section. Another difference is that the Al species is distributed through the interlayer region, not just at the interfacial region during the

M AN U

tensile process. Since Al atoms can continue healing the damaged structures, the cracks in the GO/CASH sample develop quite slowly.

3.6.2 Chemical reactions during tensile process

The enhancement of cohesive strength and plasticity can be investigated

TE D

quantitatively by analyzing the chemical reactions in the GO/C-S-H systems. The ReaxFF force field, describing the bond breakage and re-formation, provides a good opportunity to study the pathway of the atomistic structural transformation of the

EP

samples under large deformation [25]. As shown in Fig. 20a, the numbers of different hydroxyl groups are recorded with the evolution of tensile strain. The number of water molecules and other hydroxyl groups remain unchanged at the very beginning

AC C

of the tensile process, which corresponds to the short stress increasing stage from 0 to 0.1 Å/Å. As the strain increase from 0.1 to 0.3 Å/Å, while the numbers of Si-OH groups and water molecules decrease, those of C-OH and Ca-OH decrease. When the strain exceeds 0.3 Å/Å, the number of water molecules turns to decrease and that of C-OH groups continues decreasing. On the contrary, more Si-OH and Ca-OH groups are produced during this stage. The change of the hydroxyl number indicates that water dissociation and association frequently happen during the tensile process. The mechanism of “Hydrolytic reaction” has been discussed in a previous study on C-S-H gel under tensile loading: water molecules attack the Si-O-Ca bonds, weaken the bond

ACCEPTED MANUSCRIPT strength, and dissociate into H+ and OH-, finally producing the same number of Si-OH and Ca-OH bonds [32]. This reaction explains that after 0.3 Å/Å, the water number decreases and both Si-OH and Ca-OH numbers increase simultaneously. Compared with the C-S-H system, another reaction has been captured for the GO/C-S-H during

RI PT

the tensile process. As shown in Fig.20c, the ionization of C-OH occurs due to external loading: the protons in C-OH form strong H-bonds with non-bridging oxygen in the silicate chains; the tensile loading disturbs the local structure and the –OH is stretched broken; the dissociated protons bond with the Si-O-. This results in the

SC

transformation from C-OH to Si-OH. Meanwhile, as shown in Fig.20d, the protons can also re-bond with the neighboring calcium hydroxyl, producing water molecules.

M AN U

In this respect, anomalous reduction of the C-OH groups is explained by the ionization process.

Similarly, as shown in Fig.20b, the “hydrolytic reaction”, water dissociating to Si-OH and Ca-OH, has also been found in the sample of GO/CASH under tensile loading. As

TE D

compared with the reaction in the GO/CSH sample, more water molecules have been decomposed in the GO/CASH. The increasing amount of water dissociation is due to the frequently bond breakages during tensile testing. When the Al-O-Si bond is stretch

EP

broken, a non-bridging oxygen site is produced at the fracture interface. Compared with the bridging oxygen in Al-O-Si, the non-bridging atoms have higher reactive

AC C

energy. This can help the water molecules overcome the energy barrier of the “hydrolytic reaction”, contributing to the water dissociation [49].

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Fig. 20 Hydroxyl number evolution with tensile strain in (a) GO/CSH; (b)

AC C

GO/CASH; ionization pathway for (c) Si-OH production; (d) water production. The meaning of different colored ball and sticks is in Fig.1 caption.

The Al species play important role in bridging the C-S-H gel and the GO, and help construct a network structure in the GO/CASH sample. To better interpret the morphological evolution, the connectivity is estimated by the factor Qn (n=0, 1, 2, 3, 4) that is utilized to characterize the structure of silicate glass [50]. Here, n represents the

ACCEPTED MANUSCRIPT number of bridging oxygen atoms in the Si/C/Al species. In the silicate species, the Q0 is the monomer without connection with other silicate tetrahedron; Q1 represent the short chains; Q2 is long chain; Q3 and Q4 are branch and network structure. In the Go/CASH, as shown in Fig. 21a and 21b, Q0 species and Q1 species correspond to

RI PT

C-OH/C-O groups without bonding and binding to the Al atom, respectively. The Al species, linking GO and C-S-H, has a more polymerized degree and is in the form of Q2 and Q3. Meanwhile, the bridging silicate tetrahedron connected with aluminate tetrahedron also forms the Q3 branch structure and the tetrahedron pair, not directly

SC

bonded with Al species and C-OH groups, remains as the Q2 species. With increasing n, the Q species transforms from a 1-D chain-like structure to a 3-D branch with more

M AN U

connections to their neighbors. A previous study proposed that the Q3 species, present

EP

TE D

in the interlayer region, can greatly enhance the stiffness of the C-S-H structure [51].

AC C

Fig. 21 Molecular structure of the (a) Q0, Q1, Q2; and (b) Q3 species

ACCEPTED MANUSCRIPT

0.0

0.2

0.4

0.6

0.8

RI PT

Q3

SC

Percentage of Q species (%)

Q2

TE D

M AN U

Q1

0.0

0.2

0.4

0.6

Q0

0.8

Strain (Å/Å)

EP

Fig. 22 Evolution of Q species percentage with the strain for GO/CASH sample.

AC C

When the structure under goes tension loading, the Al-O-Si and C-O-Al bonds are stretched broken and re-constructed, resulting in the Q species evolution. As shown in Fig. 22, when the strain increases from 0 to 0.4 Å/Å (corresponding to the stress rising stage in the stress-strain relation), the Q3 and Q1 species gradually decrease and Q2 continuously decreases. In the post-failure stage, the Q species vary slightly, with some fluctuations. The Q3 reduction and Q2 increase are mainly attributed to the breakage between the Al species and the bridging silicate tetrahedron. The Q1 reduction is partly due to the fracture between the Al species and the C-OH or C-O groups. It confirms that the Al species plays a bridging role in withstanding the

ACCEPTED MANUSCRIPT loading. Interestingly, the Q0 species remains unchanged during the first 0.2 Å/Å and continues decreasing from 0.2 to 0.4 Å/Å. The reduction of Q0 species means that the C-OH or C-O can re-connect with the neighboring species, producing a C-O-Al connection. It should be noted that the turning point at 0.2 Å/Å agrees well with that

RI PT

observed in the stress-strain relation. When the structure is locally stretched broken, Al atoms can diffuse into the broken region and heal the structure to some extent. Therefore, a healed network can withstand more tensile strain and can enhance the plasticity of the GO/C-A-S-H sample. Conclusions

SC

4.

The intrinsic interaction between the graphene/GO and cement hydrate in

M AN U

high-performance cement-based composites has been investigated by a combinatorial approach of experiment and reactive molecular dynamics simulation at the atomic, micro and macro levels. The following conclusions were drawn from the investigations.

TE D

1. In the GO reinforced cement composite, the compressive and flexural strength is enhanced by 3.21 % and 11.62 %, respectively. It is attributed to the higher hydration degree, the nano-filler effect and the crack-bridging effect by the GO

EP

addition. GO can be regarded as a nucleation site to accelerate the hydration process and is likely to unite hydration products to form a denser microstructure. On the other hand, graphene reduces the hydration development and mechanical

AC C

behavior of cement paste due to the poor dispersibility of graphene in an alkaline environment.

2. The hydrophilic nature of the interface region between the GO sheet and C-S-H is mainly attributed to the fact that the silicate chains and functional hydroxyl groups in GO provide non-bridging oxygen sites, accepting hydrogen-bonds of the interlayer water molecules. Besides the H-bond connections, the Ca2+ and Al3+ ions near the surface of C-S-H play a mediating role in bridging the oxygen atoms in the silicate chains and hydroxyl groups in the GO. In particular, the Al3+ ions both increase the silicate chain length and heal a defective GO network,

ACCEPTED MANUSCRIPT constructing the silicate-aluminate-carbon skeleton in the interface region. 3. The functional hydroxyl groups can stabilize the atoms in both C-S-H and GO structures. The aluminate-silicate chains, calcium ions and C-OH groups construct the cages, strongly preventing the free diffusion of the interface water molecules.

RI PT

Additionally, the lifetime of the H-bonds accepted by oxygen-containing GO is longer than that of H-bonds formed by the structural water molecules in C-S-H. On the contrary, the less friction graphene interior surface, with the hydrophobic benzene-ring structures, accelerates the transport process of the water molecules,

SC

reducing the binding stability between the C-S-H and graphene system.

4. The mechanical strength of the C-S-H and graphene/GO composites is greatly

M AN U

influenced by the functional hydroxyl groups and the mediating counter-ions. While the high cohesive force and enhance plasticity of GO reinforced cement composites are mainly contributed by the H-bonds and covalent-ionic bonds (O-Ca-O or O-Al-O), the weakest mechanical behavior is attributed to poor bonding and the instability of the atoms in the interface region. This matches well

TE D

with the experimental findings and clearly explains the discrepancy in the reinforcement mechanism between the GO and graphene. More importantly, during the uniaxial tensile simulation, the Al atoms can reconnect with the

EP

neighboring C-OH or C-O groups, slowing down the crack propagation and enhancing the plasticity of the structures.

AC C

Acknowledgements

Financial support from National Natural science foundation of China under Grant 51508292, 51678317, and the China Ministry of Science and Technology under Grant 2015CB655100 are gratefully acknowledged.

References

[1] J. S. Bunch, A. M. Van Der Zande, S. S. Verbridge, I. W. Frank, D. M.

ACCEPTED MANUSCRIPT Tanenbaum, J. M. Parpia and P. L. McEuen, "Electromechanical resonators from graphene sheets," Science, vol. 315, no. 5811, pp. 490-493, 2007. [2] M. A. Rafiee, J. Rafiee, Z. Wang, H. Song, Z. Z. Yu and N. Koratkar, " Enhanced mechanical properties of nanocomposites at low graphene content," ACS Nano,

RI PT

vol. 3, no. 12, pp. 3884-3890, 2009. [3] C. Li, J. Adamcik and R. Mezzenga, "Biodegradable nanocomposites of amyloid fibrils and graphene with shape-memory and enzyme-sensing properties," Nature Nanotechnology, vol. 7, no. 7, pp. 421-427, 2012.

SC

[4] B. Wicklein, A. Kocjan, G. Salazar-Alvarez, F. Carosio, G. Camino, M. Antonietti and L. Bergstrom, "Thermally insulating and fire-retardant lightweight

M AN U

anisotropic foams based on nanocellulose and graphene oxide," Nature Nanotechnology, vol. 10, no. 3, pp. 277-283, 2015.

[5] C. Lee, X. D. Wei, J. W. Kysar and J. Hone, "Measurement of the elastic properties and intrinsic strength of monolayer graphene," Science, vol. 321, no.

TE D

5887, pp. 385-388, 2008.

[6] Z. Li, Advanced Concrete Technology, John Wiley & Sons, Inc, 2011. [7] V. Georgakilas, J. Tiwari, K. Kemp, J. Perman, A. Bourlinos, K. Kim and R. Zboril, "Noncovalent Functionalization of Graphene and Graphene Oxide for

EP

Energy Materials, Biosensing, Catalytic, and Biomedical Applications," Chemical reviews, vol. 116, no. 9, pp. 5464-5519, 2016.

AC C

[8] Y. Pan, T. Wu, H. Bao and L. Li, "Green fabrication of chitosan films reinforced with parallel aligned graphene oxide," Carbohydrate polymers, vol. 84, no. 4, pp. 1908-1915, 2011.

[9] S. Lv, Y. Ma, C. Qiu, T. Sun, J. Liu and Q. Zhou, "Effect of graphene oxide nanosheets of microstructure and mechanical properties of cement composites," Construction and building materials , vol. 49, pp. 121-127, 2013. [10] Z. Pan, L. He, L. Qiu, A. Korayem, G. Li, J. Zhu, F. Collins, D. Li, W. Duan and M. Wang, "Mechanical properties and microstructure of a graphene oxide–

ACCEPTED MANUSCRIPT cement composite," Cement and Concrete Composites, vol. 58, pp. 140-147, 2015. [11] Z. Lu, D. Hou, L. Meng, G. Sun, C. Lu and Z. Li, " Mechanism of cement paste reinforced by graphene oxide/carbon nanotubes composites with enhanced

RI PT

mechanical properties," RSC Advances, vol. 5, no. 122, pp. 100598-100605, 2015.

[12] M. E. Abrishami and V. Zahabi, "Reinforcing graphene oxide/cement composite with NH2 functionalizing group," Bulletin of Materials Science, vol. 39, no. 4,

SC

pp. 1-6, 2016.

[13] H. Alkhateb, A. Al-Ostaz, H. D. Cheng and X. Li, "Materials genome for nanocomposites,"

Journal

of

M AN U

graphene-cement

Nanomechanics

&

Micromechanics, vol. 3, no. 3, pp. 67-77, 2013.

[14] F. Sanchez and L. Zhang, "Molecular dynamics modeling of the interface between surface functionalized graphitic structures and calcium-silicate-hydrate: Interaction energies, structure, and dynamics," Journalo of Colloid and Interface

[15] A.

Peyvandi,

TE D

Science, vol. 323, no. 2, pp. 349-358, 2008. P.

Soroushian,

N.

Abdol

and

A.

M.

BalaChandra,

"Surface-Modified Graphite Nanomaterials for Improved Reinforcement

EP

Efficiency in Cementitious Paste," Carbon , vol. 63, no. 2, pp. 175-186, 2013. [16] S. Musso, J. M. Tulliani, G. Ferro and A. Tagliaferro, " Influence of carbon

AC C

nanotubes structure on the mechanical behavior of cement composites," Composites Science and Technology, vol. 69, no. 11-12, pp. 1985-1990, 2009.

[17] A. Sedaghat, M. K. Ram, A. Zayed, R. Kamal and N. Shanahan, "Investigation of physical properties of graphene-cement composite for structural applications," Journal of Composite Materials, vol. 4, no. 1, pp. 12-21, 2014. [18] M. A. Rafiee, T. N. Narayanan, D. P. Hashim, N. Sakhavand, R. Shahsavari and R. Vajtai, "Hexagonal boron nitride and graphite oxide reinforced multifunctional porous cement composites," Advanced Functional Materials, vol. 23, no. 45, pp.

ACCEPTED MANUSCRIPT 5624-5630, 2013. [19] M. Wang, R. Wang, H. Yao, S. Farhan, S. Zheng and C. Du, "Study on the three dimensional mechanism of graphene oxide nanosheets modified cement," Construction and Building Materials, vol. 126, pp. 730-739, 2016.

RI PT

[20] K. Chenoweth, A. C. T. van Duin and W. A. Goddard, "Reaxff reactive force field for molecular dynamics simulations of hydrocarbon oxidation," Journal of Physical Chemistry A, vol. 112, no. 5, pp. 1040-1053, 2008.

[21] A. van Duin, A. Strachan, S. Stewman, Q. Zhang, X. Xu and W. Goddard,

SC

"ReaxFFsio Reactive Force Field for Silicon and Silicon Oxide Systems," The Journal of Physical Chemistry A, vol. 107, no. 19, pp. 3803-3811, 2003.

M AN U

[22] H. Manzano, S. Moeini, F. Marinelli, A. van Duin, F. Ulm and R. Pellenq, "Confined Water Dissociation in Microporous Defective Silicates:Mechanism, Dipole Distribution, and Impact on Substrate Properties," Journal of the American Chemistry Society, vol. 134, no. 4, p. 2208−2215, 2011.

TE D

[23] D. W. Brenner, O. A. Shenderova, J. A. Harrison, S. J. Stuart, B. Ni and S. B. Sinnott, "A second-generation reactive empirical bond order (REBO) potential energy expression for hydrocarbons," Journal of Physics: Condensed Matt, vol. 14, no. 4, pp. 783-802, 2002.

EP

[24] S. Leroch and M. Wendland, "Simulation of Forces between Humid Amorphous Silica Surfaces: A comparison of empirical atomic force fields," The Journal of

AC C

Physical Chemistry C, vol. 116, pp. 26247-26261, 2012.

[25] H. Manzano, E. Masoero, I. Arbeloa and H. Jennings, "Molecular modelling of shear deformations in ordered and disordered Calcium Silicate Hydrates," Soft matter, vol. 9, no. 30, pp. 7333-7341, 2013.

[26] T. Lau, A. Kushima and S. Yip, "Atomistic Simulation of Creep in a Nanocrystal," Physical Review Letters, vol. 104, no. 17, p. 175501, 2010. [27] H. Manzano, R. Pellenq, F. Ulm and M. Buehler, "Hydration of Calcium Oxide Surface Predicted by Reactive Force Field Molecular Dynamics," Langmuir, vol.

ACCEPTED MANUSCRIPT 28, pp. 4187-4197, 2012. [28] S. Hamid, "The crystal structure of the 11 A natural tobermorite Ca2.25 Si3 O7.5(OH)1.5H2O," Zeitschrifit fur Kristallographie 154, pp. 189-198, 1981. [29] M. Montes-Morán, F. van Hattum, J. Nunes, A. Martínez-Alonso, J. Tascón and

RI PT

C. Bernardo, " A study of the effect of plasma treatment on the interfacial properties of carbon fibre–thermoplastic composites," Carbon, vol. 43, no. 8, pp. 1795-1799, 2005.

[30] R. D. Allington, D. Attwood, I. Hamerton, J. N. Hay and B. Howlin, "A model of

SC

the surface of oxidatively treated carbon fibre based on calculations of adsorption interactions with small molecules.," Composites Part A Applied Science and

M AN U

Manufacturing, vol. 29, no. 9, pp. 1283-1290, 1998.

[31] S. Plimpton, "Fast Parallel Algorithms for Short-Range Molecular Dynamics," Journal of computational physics, vol. 117, no. 1, pp. 1-19, 1995. [32] D. Hou, T. Zhao, H. Ma and Z. Li, "Reactive Molecular Simulation on Water

TE D

Confined in the Nanopores of the Calcium Silicate Hydrate Gel: Structure, Reactivity, and Mechanical Properties," The Journal of Physical Chemistry C , vol. 119, no. 3, pp. 1346-1358, 2015.

[33] M. Youssef, R. J. M. Pellenq and B. Yildiz, "Glassy Nature of Water in an

EP

Ultraconfining Disordered Material:The Case of Calcium silicate hydrate," Journal of American Chemistry Society , vol. 133, no. 8, p. 2499–2510, 2011.

AC C

[34] M. Qomi, F. Ulm and R. Pellenq, "Evidence on the Dual Nature of Aluminum in the Calcium-Silicate-Hydrates Based on Atomistic Simulations," Journal of the American Ceramic Society, vol. 95, no. 3, pp. 1128-1137, 2012.

[35] D. Li, M. Muller, S. Gilje, R. Kaner and G. Wallace, "Processable aqueous dispersions of graphene nanosheets," Nature Nanotechnology , vol. 3, no. 2, pp. 101-105, 2008. [36] R. N. Mead and G. Mountjoy, "A Molecular Dynamics Study of the Atomic Structure of (CaO)x(SiO2)1-x Glasses," J.Phys.Chem, vol. 110, no. 29, pp.

ACCEPTED MANUSCRIPT 273-278, 2006. [37] R. J. Pellenq, A. Kushima, R. Shahsavari, K. J. Van Vliet, M. J. Buehler and S. Yip, "A realistic molecular model of cement hydrates," PNAS, vol. 106, no. 38, p. 16102–16107, 2009.

RI PT

[38] E. L’Hôpital, B. Lothenbach, D. A. Kulik and K. Scrivener, "Influence of calcium to silica ratio on aluminium uptake in calcium silicate hydrate," Cement and Concrete Research, vol. 85, pp. 111-121, 2016.

[39] D. Hou, Z. J. Li and T. J. Zhao, "Reactive force field simulation on

SC

polymerization," RSC Advances, vol. 5, pp. 448-461, 2015.

[40] D. Hou and Z. Li, "Molecular Dynamics Study of Water and Ions Transported

M AN U

during the Nanopore Calcium Silicate Phase:Case Study of Jennite," Journal of Materials in Civil Engineering, vol. 26, no. 5, 2014.

[41] D. Hou, Z. Li, T. Zhao and P. Zhang, "Water transport in the nano-pore of the calcium silicate phase: reactivity, structure and dynamics," Physical Chemistry

TE D

Chemical Physics , vol. 17, pp. 1411-1423, 2015.

[42] F.-X. Coudert, R. Vuilleumier and A. Boutin, "Dipole Moment, Hydrogen Bonding and IR Spectrum of Confined Water," Physical Chemistry Chemical Physics , vol. 7, no. 12, p. 2464–2467, 2006.

EP

[43] W. Cai, R. Piner, F. Stadermann, S. Park, M. Shaibat, Y. Ishii, D. Yang, A. Velamakanni, S. An, M. Stoller, J. An, D. Chen and R. Ruoff, "Synthesis and

AC C

solid-state NMR structural characterization of 13C-labeled graphite oxide," Science, vol. 321, pp. 1815-1817, 2008.

[44] M. J. A. Qomi, M. Bauchy, F. J. Ulm and R. J. M. Pellenq, " Anomalous composition-dependent dynamics of nanoconfined water in the interlayer of disordered calcium-silicates," The Journal of chemical physics, vol. 140, no. 5, p. 054515, 2014. [45] J. K. Holt, H. G. Park, Y. Wang, M. Stadermann, A. B. Artyukhin, C. P. Grigoropoulos, A. Noy and O. Bakajin, "Fast mass transport through

ACCEPTED MANUSCRIPT sub-2-nanometer carbon nanotubes," Science, vol. 312, no. 5776, pp. 1034-1037, 2006. [46] K. Falk, F. Sedlmeier, L. Joly, R. R. Netz and L. Bocquet, "Molecular Origin of Fast Water Transport in Carbon Nanotube Membranes: Superlubricity versus

RI PT

Curvature Dependent Friction," Nano Letters, vol. 10, no. 10, pp. 4067-4073, 2010.

[47] A. Alexiadis and S. Kassinos, "Molecular simulation of water in carbon nanotubes," Chemical Reviews, vol. 108, no. 12, pp. 5014-5034, 2008.

SC

[48] P. Bonnaud, Q. Ji, B. Coasne, R. Pellenq and K. Van Vliet, "Thermodynamics of

11422–11432, 2012.

M AN U

Water Confined porous Calcium silicate hydrate," Langmuir, vol. 28, no. 31, p.

[49] T. Zhu, J. Li, X. Lin and S. Yip, "Stress-dependent molecular pathways of silica-water reaction," Journal of Mechanics and Physics of solids, vol. 53, no. 7, pp. 1597-1623, 2005.

TE D

[50] V. R. Mastelaro, E. D. Zanotto, N. Lequeux and R. J. Cortes, "Relationship between short-range order and ease of nucleation in Na2Ca2Si3O9, CaSiO3 and PbSiO3 glasses," Journal of Non-Crystalline Solids, vol. 262, no. 1-3, pp. 191-199, 2000.

EP

[51] S. Merlino, E. Bonnacorsi and T. Armbruster, "The real structure of tobermorite 11 A: normal and anomalous forms, OD character and polyptic modifications,"

AC C

European Journal of Mineralogy13, pp. 577-590, 2001.