Morphology of calcium silicate hydrate (C-S-H) gel: a molecular ...

Advances in Cement Research

http://dx.doi.org/10.1680/adcr.13.00079

Paper 1300079

Received 21/10/2013; revised 11/12/2013; accepted 12/12/2013

ICE Publishing: All rights reserved

Advances in Cement Research

Morphology of calcium silicate hydrate(C-S-H) gel: a molecular dynamic studyHou, Ma and Li

Morphology of calcium silicatehydrate (C-S-H) gel: amolecular dynamic studyDongshuai HouPhD Candidate, Department of Civil and Environmental Engineering,Hong Kong University of Science and Technology, Hong Kong

Hongyan MaPost-doctoral Fellow, Department of Civil and Environmental Engineering,Hong Kong University of Science and Technology, Hong Kong

Zongjin LiProfessor, Department of Civil and Environmental Engineering, HongKong University of Science and Technology, Hong Kong

Due to its complexity at nanoscale, calcium silicate hydrate (C-S-H), the dominant binding phase in cement hydrates,

is not yet completely understood. In this study, molecular dynamics was employed to simulate the hydration

products at low and high calcium/silicon ratios. It was found that two morphologies of calcium silicate hydrate gels

can be distinguished – a branched structure at low calcium/silicon ratios and an ellipsoid particle structure at high

calcium/silicon ratios. Using virtual X-ray diffraction (XRD), nuclear magnetic resonance (NMR) and small-angle

neutron scattering (SANS) techniques, the simulated structures were characterised, confirming that they show

features of calcium silicate hydrate as revealed by experimental approaches. The short-range structures of calcium

and silicon atoms and the distorted calcium tetrahedrons resemble the features of silicate glasses obtained from

experiments, implying the amorphous nature of the local structure in calcium silicate hydrate gel. Furthermore,

formation mechanisms for the two morphologies are proposed. In the hydration process, calcium ions play roles in

depolymerising the silicate structure and preventing the amorphous network formation. Therefore, at low calcium/

silicon ratios, the reaction is governed by silicate skeleton growth, but at high calcium/silicon ratio, aggregations of

calcium ions and short silicate chains dominate.

IntroductionAs cement material is so widely utilised all over the world, much

research has been carried out on the intrinsic properties of cement

paste (Li, 2011; Ma and Li, 2013; Ma et al., 2013). The

shrinkage, creep and strength of cement pastes are significantly

affected by the structure of calcium silicate hydrate (C-S-H) gel,

which has thus been investigated and is regarded as an essential

topic for cement-based materials. The multi-scale characteristics

and complicated chemical composition of calcium silicate hydrate

gel bring challenges in interpreting the morphology of the

hydration product. At microscale, calcium silicate hydrate gels

observed by scanning electron microscopy (SEM) do not distri-

bute uniformly and exist in various morphologies, such as fibres,

flakes, honeycombs and tightly assembled grains (Li, 2011). The

diversity of the morphologies is caused by many factors, such as

the water content, calcium/silicon ratio and hydration degree.

High-resolution transmission electron microscopy (TEM) can

distinguish the calcium silicate hydrate gel into inner products of

foil morphology and outer products of a fibrillar structure

(Groves, 1987). However, because calcium silicate hydrate gels

cannot be directly observed at nanoscale, their molecular struc-

ture was only inferred by combining experimental and computa-

tional techniques.

Small-angle neutron scattering (SANS) and small-angle X-ray

scattering (SAXS) techniques have shown that the average calcium/

silicon ratio of calcium silicate hydrate gel is 1.7 and its density is

2.6 g/cm3 (Allen et al., 2007). Based on these results and silicate

chain information from nuclear magnetic resonance (NMR) testing

(Cong and Kirkpatrick, 1996) and calcium silicate layer informa-

tion from X-ray diffraction (XRD) tests, a realistic model of

calcium silicate hydrate gel (Pellenq et al., 2009) was proposed

with the assistance of molecular simulation. However, even though

this realistic model can give a good interpretation on the atomic

structures and mechanical properties of calcium silicate hydrate gel

to within 1 nm, a gap still exists between the molecular scale and

the microscale. Large-scale molecular simulation, constructing a

colloid model from molecular level, could provide linkage between

the two levels. Dolado et al. (2007, 2011) employed the silica

formation reaction in the presence of calcium hydroxide (Ca(OH)2)

to simulate the cement hydration reaction. The simulation box

included more than 60 000 atoms. A hydration product – a

branched structure of size around 5 nm – was achieved after a

2.3 ns simulation and also identified with features of calcium

silicate hydrate gels. It is noted, however, that because of simulation

time limitations, the branch structure of calcium silicate hydrate

may not be the final hydration product. In addition, when the

1

calcium/silicon ratio is 1.65, the single morphology cannot repre-

sent the colloid phases of calcium silicate hydrate gel at different

calcium/silicon ratios.

In this work, to bridge the gap between calcium silicate hydrate

gel at molecular level and microlevel, molecular dynamics was

utilised to simulate the hydration process of more than

80 000 atoms within a 10 nm system. Calcium silicate hydrate

phases with high and low calcium/silicon ratios of 1.75 and 1.00

were simulated to investigate the calcium/silicon influence on the

morphology of the gel at mesoscale. Such simulation results can

provide information on the dynamics and structural role of

calcium and silicon in the hydration process. Hydration simula-

tion can directly describe the colloid cluster formation mechan-

ism, validating the theoretical and experimental models. The

local structure of calcium and silicon, Q species evolution and

cluster size development were analysed for a comprehensive

prediction of the colloid structure of calcium silicate hydrate gel.

Simulation methods

Reaction mechanism

Cement hydration, the production process of calcium silicate

hydrate gel, is based on the sol–gel reaction in the presence of

calcium ions. The oligomerisation of silicate monomers follows

the water-producing condensation reaction (Feuston and Garofali-

ni, 1990)

Si(OH)4þ Si(OH)4 ! (HO)3SiOSi(OH)3þH2O

As illustrated in Figure 1(a), the reaction leads to the linkage of

silicic monomers to form a dimer structure and the dissociation

of one water molecule. The polymerisation reaction produces a

different silicate cluster. The different silicate products are

defined as Qn, where n is the number of bridging oxygen atoms

for an individual silicate tetrahedral and varies from 0 to 4. It is

shown in Figure 1(b) that Qn species represent different silicate

structures – Q0 is the isolated monomer, Q1 and Q2 are chains,

and Q3 and Q4 reflect the interconnected structures.

Potential model and simulation procedures

The Feuston–Garofalini (FG) model, developed for the silica

system (Feuston and Garofalini, 1990), was used for the current

sol–gel reaction. The model contains three interaction potentials:

the Born–Mayer–Huggins (BMH) potential, the Stillinger–We-

ber (SW) potential and the Rahman–Stillinger–Lemberg (RSL2)

potential. The interaction among calcium (Ca) and silicon (Si),

oxygen (O) and hydrogen (H) can be achieved from a previous

study about calcium silicate hydrate (I) phases (Faucon et al.,

1996).

The experimental sol–gel reaction and the calcium silicate

hydrate gel hydration take more than 24 h. In order to simulate

the time-consuming reaction at a nanosecond scale, a high-

temperature (1800–2500 K) simulation environment is normally

preferred to accelerate the reaction rate (Dolado et al., 2007;

Garofalini and Martin, 1994). In the current simulation, 1800 K

was adopted as the simulation condition.

Molecular dynamics simulations were performed using the

Lammps package (Lammps, 2008) to study the calcium silicate

cluster evolution at calcium/silicon ratios of 1.75 and 1.00. The

initial configuration was obtained by adding silicic acid (Si(OH)4),

calcium hydroxide (Ca(OH)2) and water (H2O) molecules ran-

domly in a box of size 10 3 10 3 10 nm to achieve an overall

density of about 1.2 g/cm3: Detailed information of the simulation

system is given in Table 1. It should be noted that around

100 000 atoms can provide stable and accurate statistical results.

These initial configurations were relaxed at 300 K for 100 ps to let

the atoms diffuse to random positions. In order to increase the

reaction rate, the system was heated for 100 ps, linearly increasing

the temperature to 1800 K. The reactions were implemented at

1800 K for 10 000 ps, followed by a linearly decreasing tempera-

ture to 300 K for 1000 ps. Finally, the systems were maintained at

Hydrogen

Oxygen

Silicon+

(a)

(b)

Q0

Q1

Q2

Q4

Q3

Figure 1. (a) Polymerisation reaction for silicic monomers.

(b) Structures for sol–gel reaction products, Qn (n ¼ 0, 1, 2, 3, 4)

2

Advances in Cement Research Morphology of calcium silicate hydrate(C-S-H) gel: a molecular dynamic studyHou, Ma and Li

300 K and further equilibrated for 50 ps before a long production

run for 500 ps. Atom coordination information was recorded every

1 ps for later analysis. A time step of 1 3 10�3 ps was utilised.

Results and discussion

Calcium silicate hydrate morphology

During around 10 ns sol–gel reaction time in solution, calcium

silicate clusters with different morphologies are segregated from

the water solution. As shown in Figure 2(a), at a low calcium/

silicon ratio of 1.00, the calcium silicate cluster grows across the

simulation box with a branch size larger than 3.5 nm. More

interestingly, the network phase, growing in radiated style,

demonstrates amorphous features similar to the silica glass

achieved from a sol–gel reaction (Feuston and Garofalini, 1990).

On the other hand, the cluster with a calcium/silicon ratio of 1.75

nucleates and develops into an ellipsoid particle with a diameter

of approximately 5 nm, as shown in Figure 2(b). Due to the long

simulation time, the hydration degree is supposed to be higher

than that reported by Dolado et al. (2011), which can transform

the structure to an egg-like particle. Compared with the amor-

phous network structure, the hydration product at the high

calcium/silicon ratio demonstrates a more ordered state. In the

current simulation, the egg-like phase and network phase indicate

that the composition of the calcium silicate hydrate gel, especially

the calcium/silicon ratio, is an important factor influencing the

morphologies of calcium silicate hydrate gel at the nanoscale.

The simulated morphologies can give some theoretical under-

standing of microscale observations. The gel structure of cement

hydrate has been investigated by TEM at microscale. Two types

of calcium silicate hydrate phases can also be observed on TEM

photographs at the scale of around 100 nm (Groves, 1987). It is

widely accepted and experimentally confirmed that the inner

product (IPs) and outer product (OPs) are distinguished in

accordance with the density and geometry morphology (Richard-

son and Groves, 1993). As shown in Figure 3, the IP is

aggregated by small globule particles packed in a relatively dense

state with a maximum pore size of less than 10 nm. On the

contrary, the OP has a fibril structure, with a large length/width

ratio and large pores between the fibril structures. Microanalysis

of cement hydrate composition has demonstrated that the

calcium/silicon ratio of the IPs is larger than that of the OPs

(Richardson, 2004). According to the simulation results presented

in this study, ellipsoid particles with a high calcium/silicon ratio

are more likely to assemble together to construct the high-density

phase, while the network structure with a low calcium/silicon

ratio may transform into the loose fibril morphology. Further-

more, Figures 3(b), 3(c) and 3(d) show the morphologies of

(a)

(b)

Figure 2. Hydration products at different calcium/silicon ratios.

(a) Branched cluster with calcium/silicon ratio of 1.00. (b) Ellipsoid

cluster with calcium/silicon ratio of 1.75. Balls represent calcium;

thick sticks represent Si–O bonds; short sticks represent O–H

bonds

Calcium/

silicon

ratio

Number of atoms Size:

nm

Density:

g/cm3

Calcium Silicon Oxygen Hydrogen

1.00 3375 3375 33 750 47 250 10.5 1.17

1.75 2401 1372 29 498 48 706 9.7 1.21

Table 1. Simulation box information

3


synthesised gel at calcium/silicon ratios of 0.66, 1.50 and 2.00

respectively (Sasaki et al., 1996; Viallis-Terrisse et al., 2001). It

can thus be concluded that, with increasing calcium/silicon ratios,

the calcium silicate hydrate gel changes from a loose fibril-like

morphology to dense granular-like particles.

At a calcium/silicon ratio of 1.75, quite close to the average value

of the calcium silicate hydrate gel, the simulated colloid particle

with a diameter of around 5 nm matches well with the globule

particle proposed by Jennings (2000, 2008) and predicted from

SANS (Allen et al., 2007). To further study the properties of the

simulated gels, the clusters were characterised with results from

various virtual experimental techniques.

Comparison with experimental data

Simulated SANS and XRD curves were produced to further

identify the geometry and the crystalline arrangement of the

hydrated products. As shown in Figure 4, the simulated scattering

curve has a good agreement with that achieved from the SANS

test. The slope of the curve changes from the region �2.51 to

�2.85 to �3.83 to �3.85, when the scattering vector is between

0.012 nm�1 and 0.016 nm�1, corresponding to an average size of

5–8 nm. As shown in Figure 5, in the simulated diffraction curve,

two XRD intensity peaks can be observed at ,78 and 308. The

former corresponds to 1.0–1.4 nm spacing in the calcium silicate

cluster. It should be noted that this peak is also present in the

tobermorite and synthesised calcium silicate hydrate gels. The

(a)

IP

OP

500 nm

(b)

(c)

200 nm

200 nm

(d)

200 nm

Figure 3. High-resolution TEM photograph for different morphologies of calcium silicate hydrate at microlevel: (a) inner product (IP) and

outer product (OP) (Groves, 1987); (b) calcium/silicon ratio ¼ 0.66 (Viallis-Terrisse et al., 2001); (c) calcium/silicon ratio ¼ 1.50 (Viallis-

Terrisse et al., 2001); (d) calcium/silicon ratio ¼ 2.00 (Sasaki et al., 1996)

4


distance ranging from 0.9–1.4 nm is widely accepted as the

interlayer space between calcium silicate sheets in calcium

silicate hydrate analogues such as tobermorite and jennite

(Hamid, 1981; Janika et al., 2001; Merlino et al., 2001). By

identification with experimental results, it is confirmed that the

simulated gels in this study follow the features of calcium silicate

hydrate gels.

Glass nature of the calcium silicate hydrate gel

For bonded atoms, the relevant bond distances can be readily

determined by the positions of the peaks in the corresponding

radial distribution function (RDF). In calcium silicate hydrate gel

systems, the Si–O bond distance shows a clear dependence on

the calcium concentrations. Figure 6(a) shows the Si–O RDFs of

calcium silicate hydrate gel with calcium/silicon ratios of 1.00

and 1.75. The position of the first peak shifts slightly from

0.162 nm to 0.160 nm as the calcium/silicon ratio increases from

0.1 nm to 0.175 nm. Note that the Si–O bond lengths resemble

those in amorphous silica (i.e. 0.162 nm) (Cormack and Du,

2001; Du and Cormack, 2004; Mead and Mountjoy, 2006).

Two types of oxygen atoms are present in the calcium silicate

hydrate gels, as shown in Figure 6(b). An oxygen atom that

bridges two neighbouring silicon atoms is defined as a bridging

oxygen atom (BO) and an oxygen atom that is bonded to only

one silicon atom is defined as a non-bridging oxygen atom

(NBO). Thus, the Si–O RDF can be further decomposed into Si–

BO and Si–NBO RDFs, and the two components are plotted in

Figure 6(b) for the case of calcium/silicon ¼ 1.00. From Figure

6(b), a difference of 0.006–0.007 nm was observed between

0·001 0·01 0·1

SAN

S in

tens

ity: a

rbitr

ary

unit

Q: nm�1

SANS resultsCalcium/silicon 1·00�

Calcium/silicon 1·75�

Slope 3·85to 3·85

� ��

Slope 2·51to 2·85

� ��

Figure 4. Comparison of experimental SANS curves (Allen et al.,

2007) and simulated cluster for calcium/silicon ratios of 1.00 and

1.75

1

2

3

4

10 20 30 40

XRD

inte

nsity

: arb

itrar

y un

its

2 : degreesθ

Tobermorite (experiment)Calcium silicate hydrate(experiment)



Figure 5. XRD curves: comparison between experimental results

(tobermorite (Janika et al., 2001) and synthesised calcium silicate

hydrate (Martin, 1994)) and simulated clusters for

calcium/silicon ratios of 1.00 and 1.75

0

4

8

12

16

0 0·2 0·4 0·6 0·8 1·0

g r()

Distance: nm(a)

Calcium/silicon 1·00Calcium/silicon 1·75

��

0

4

8

12

16

0·12 0·14 0·16 0·18 0·20

0

4

8

12

16

0 0·2 0·4 0·6 0·8

g r()

Distance: nm(b)

Si-BOSi-NBOSi-O

0

4

8

12

16

0·12 0·14 0·16 0·18 0·20

Figure 6. (a) Si–O RDFs of calcium silicate hydrate systems.

(b) Si–O RDF and its Si–BO and Si–NBO components for calcium

silicate hydrate gel at a calcium/silicon ratio of 1.75

5


Si–BO and Si–NBO bond distances. Furthermore, due to the

smaller shielding effect from the neighbouring silicon atoms, the

NBOs may show more electronegativity than BOs. The presence

of calcium atoms can depolymerise the silica networks by break-

ing Si–BO bonds and hence increase the number of Si–NBO

bonds. Such an effect results in a decrease in the Si–O bond

distance at a high calcium/silicon ratio.

Figure 7(a) shows the Ca–O RDFs obtained from simulations of

the calcium silicate hydrate gel systems at calcium/silicon ratios

of 1.00 and 1.75. For the calcium silicate hydrate gel with a

calcium/silicon ratio of 1.00, the result shows that the Ca–O

bond distance is between 0.236 and 0.242 nm. Previous experi-

mental studies of calcium metasilicate (CaSiO3) glass using

XRD, extended X-ray absorption fine structure (EXAFS) and

neutron diffraction (Mastelaro et al., 2000; Taniguchi et al.,

1997) showed a value between 0.236 and 0.249 nm. This implies

that the local environment of calcium atoms in calcium silicate

hydrate gel is quite close to that in the amorphous silicate glass.

Two peaks located at approximately 0.24 nm and 0.3 nm were

found in both of the Ca–O RDFs. It is instructive to further

decompose each Ca–O RDF into Ca–NBO and Ca–BO parts. As

shown in Figure 7(b) for the case calcium/silicon ¼ 1.75, the

NBOs contribute to the first peak and BOs to the second (much

smaller) peak. Thus, the Ca–NBO and Ca–BO distances are

0.24 nm and 0.3 nm respectively.

Cluster formation mechanism

The local structure of the simulated calcium silicate hydrate gel

has been discussed and shows the nature of calcium silicate glass

in short-range order. In this section, the long-range order will be

further investigated by discussing cluster development and Q

species evolution at low and high calcium/silicon ratios respec-

tively.

Cluster development at low calcium/silicon ratio

At a low calcium/silicon ratio, cluster evolution demonstrates the

polymerisation process of silicic acid in the presence of calcium

atoms. The silicon cluster is defined as those silicon atoms

connected together by BOs and the minimum cluster is a dimer.

Cluster evolution is characterised by a maximum cluster size and

cluster numbers, which can be calculated by a recursive algorithm

proposed in previous simulation (Rao and Gelb, 2004).

Three obvious stages – initial small cluster formation, cluster–

cluster aggregation and large cluster densification – construct the

whole process of hydration. In the initial stage, a dimer structure

first forms and small chains randomly distribute in the water box,

isolated by calcium atoms and water molecules. As shown in

Figure 8(b), the number of small structures, dimers or long

chains, widely increases in this period and reaches the peak value

in the cluster number evolution (Figure 8(a)). Maximum cluster

development has a low growth rate in this period because dimer

formation is the dominant reaction and connections between

different isolated silicate polymers have not widely occurred. The

second period starts at around 1.3 ns and the cluster growth rate

suddenly changes in the maximum cluster curve. During the

cluster–cluster aggregate period, small clusters begin to aggregate

with neighbouring ones and combine to form large clusters.

Small clusters further aggregate and the density of the silicon

clusters continues to increase. As shown in Figure 8(c), small

clusters collide with nearby ones and the connection reaction

between different clusters occurs widely. This leads to a rapid

decrease in cluster number and an increase in maximum cluster

size. At the beginning of the third stage, a newly formed large

cluster size remains unchanged for more than 3 ns. This equili-

brium period is also when the cluster transforms into energy

preferable structures. Afterwards, the size of the large cluster

continues to increase. Additionally, water molecules dissociating

from the sol–gel reaction dissolve in the solution. Phase separa-

tion between calcium silicate clusters and water clusters can be

observed in the final stage. Finally, a calcium silicate branch with

a diameter of 3.5 nm is formed in the water environment, as

illustrated in Figure 8(d).

0

0·5

1·0

1·5

2·0

0 0·2 0·4 0·6 0·8

g r()

Distance: nm(a)

0

0·5

1·0

1·5

2·0

0·20 0·24 0·28 0·32 0·36 0·40

Calcium/silicon 1·00�Calcium/silicon 1·75�

0

0·4

0·8

1·2

1·6

2·0

2·4

0·1 0·2 0·3 0·4 0·5 0·6

g r()

Distance: nm(b)

Ca–BOCa–NBOCa–O

Figure 7. (a) Ca–O RDFs of calcium silicate hydrate gel. (b) Ca–O

RDF and its Ca–BO and Ca–NBO components for a calcium/silicon

ratio of 1.75

6


Cluster evolution at high calcium/silicon ratio

As shown earlier, cluster development at a low calcium/silicon

ratio is based on silicate chain skeleton growth resulting from the

sol–gel reaction. With increasing calcium concentrations, the

silicate chains are broken into short segments. At a high calcium/

silicon ratio, cluster evolution is based on a different mechanism.

Calcium atoms, associating with the NBOs in the silicon tetrahe-

dral, play a role in bridging the isolated silicate polymer structure

and construct the Si–Ca cluster. Due to the high diffusion rate,

calcium atoms can bridge with NBO atoms quickly and, at the

very beginning, small Si–O–Ca clusters form (Figure 9(a)).

Meanwhile, in the local silicon-rich region, silicate chains can

grow by the polymerisation reaction in a relatively slow mode.

Therefore, the structure of the calcium silicate hydrate gel

depends on the competition between the sol–gel reaction and the

NBO–Ca association.

More than 95% of the Si–Ca aggregations finish within 3 ns,

which is far quicker than the sol–gel cluster formation at low

calcium concentrations. Silicate chains and calcium atoms ini-

tially nucleate into small clusters that are attributed to both sol–

gel reaction and Ca–NBO association. Due to the large surface

energy of the nanoscale colloid in water solution, unstable small

0

25

50

75

100

0

20

40

60

80

0 3000 6000 9000

Clu

ster

siz

e: %

Time: ps(a)

Clu

ster

num

ber

(b)

(c) (d)

MaximumCluster number

Figure 8. Cluster evolution at low calcium/silicon ratio: (a) cluster

size and cluster number change with time; (b) snapshots of initial

small cluster formation; (c) cluster–cluster aggregation and

(d) large cluster densification. The sticks represent silicate chains

7


clusters, with a high kinetic potential, collide with their neigh-

bours and merge together. In the merging period, branch struc-

tures are able to be produced, as shown in Figure 9(b). Such

structures have also been observed by Dolado et al. (2011).

However, a branch involving a large percentage of calcium and

silicon in the system is just a temporary structure in the middle

period of hydration. Because the network constructed by the

Ca–NBO–Si bonds is much weaker than a skeleton consisting of

(a) (b)

(c) (d)

Figure 9. Snapshots of cluster growth at high calcium/silicon

ratio: (a) small cluster formation; (b) cluster–cluster aggregation

and branch structure formation; (c) densification of cluster; (d)

ellipsoid cluster formation

8


Si–BO–Si bonds (Shahsavari et al., 2009) at low calcium/silicon

ratios, the branch-like structure is likely to transform to a more

stable morphology with a larger colloid size (Figure 9(c)). The

continuous interaction between water and Ca–Si clusters can

finally result in densification of the calcium silicate hydrate gel

structures that are energetically stable ellipsoid particles (as

shown in Figure 9(d)).

After 10 ns equilibrium time, the morphology of the calcium

silicate hydrate gel at both high and low calcium/silicon ratio

remained unchanged in the current simulation. However, it is

worth noting that extending the simulation time can further

enhance the hydration degree. Considering the periodic condi-

tions of the simulation box, in a longer simulation run, for the

low calcium/silicon ratio case, the network of the calcium silicate

hydrate gel can grow into the fibril-like structure in a radiative

manner. On the contrary, at high Ca/Si ratio, for a longer

simulation run, isolated calcium silicate hydrate ellipsoids can

aggregate together to form a relatively dense packing morph-

ology.

The cluster formation mechanism for the calcium silicate hydrate

gel also provides a molecular insight into curing conditions such

as the temperature and the introduction of the mineral admix-

tures. The sol–gel reaction is based on the dimerisation reaction

between silicate monomers and elevating the temperature can

help accelerate the hydration degree. On the one hand, high

temperatures help overcome the reaction energy barrier so that

the monomers can de-proton and associate with each other more

easily and, on the other, the mobility of the atoms is increased at

high temperature so that they collide with each other more often

and accelerate the reaction rate. NMR studies have also shown

that a high curing temperature can enhance the polymerisation

degree in calcium silicate hydrate gel (Brough et al., 1994).

Mineral admixtures also have a great influence on the morph-

ology of calcium silicate hydrate gel. For example, adding silica

fume to cement can introduce many silicate phases to the

hydration process, reducing the calcium/silicon ratio. At a lower

calcium/silicon ratio, the morphology of calcium silicate hydrate

gel is based on silicate skeleton growth and a network structure is

more likely to form. Other mineral phases can result in alumina

and sodium atoms in the hydration products, which might have

an influence on network formation. In future work, other ions in

cement, such as Al3þ and Naþ, will be added to the simulation

system to explore their influence on the structure of hydration

products at nanoscale.

Connectivity factor Q evolutions

Previous cluster analysis gives insights about the calcium silicate

hydrate gel growth in respect of colloid development. In order to

better interpret the change of intrinsic molecular structures, Q

species evolution should be taken into consideration (Feuston and

Garofalini, 1988). As shown in Figure 10, the Q0 species

decreases monotonically and the growth rates of Q1 and Q2 are

larger than those of Q3 and Q4: The evolution process indicates

that the sol–gel reaction is the structural transformation from

monomers to silicate chains, cross-linkage structures or networks.

The main difference in Q species between the high calcium/

silicon phase and the low calcium/silicon phase is the percentage

of Q2 + Q3 species. At a low calcium/silicon ratio, Q2 and Q3

species take up more than 30% of the whole gel, indicating that

the long chain structures can interconnect with each other, which

is similar to connected silicate chains in the tobermorite model of

Merlino et al. (2001). Nevertheless, at a high calcium concentra-

tion, the network structure (Q4) and the cross-branched structure

(Q3) almost disappear. The discrepancy in Q species composition

mainly results from the de-polymerisation role of calcium atoms.

On the one hand, to connect with a neighbouring silicate

structure, silicate monomers have to compete with calcium atoms.

However, as proposed in a previous calcium silicate glass

simulation (Mead and Mountjoy, 2006), NBOs energetically

prefer to connect the calcium atoms. Consequently, adding more

calcium contributes to more Si–O–Ca bonds and decreases the

amount of Si–O–Si bonds. On the other hand, with respect to

dynamics, calcium atoms slow down the diffusion of silicon

0

20

40

60

80

100

0 2 4 6 8 10

Qn:

%

Time: ns(a)

Q0Q1Q2Q3Q4

0

20

40

60

80

100

0 2 4 6 8 10

Qn:

%

Time: ns(b)

Q0Q1Q2Q3Q4

Q0

Q1

Q2

Q3Q4

Q0Q1Q2

Q3

Q4

Figure 10. Q species evolution for calcium/silicon ratio of (a) 1.75

and (b) 1.00

9


atoms. Thus, compared with calcium atoms, the restricted silicon

atoms have less probability of colliding with nearby NBOs.

It should be noted that Q3 and Q4 species are the main structural

discrepancy between the calcium silicate hydrate morphology at

low and high calcium/silicon ratios. Q3 and Q4 species also have

a great influence on the mechanical performance of the calcium

silicate hydrate gel. A previous ab initio calculation (Shahsavari

et al., 2009) showed that the presence of Q3 species can improve

the interlayer stiffness of tobermorite, a mineral analogue of

calcium silicate hydrate gel (Merlino et al., 2001), by forming a

hinge mechanism. The skeleton role of the silicate structure is

based on the percentage of Q3, and the binding energy of Si–O

bonds is larger than that of Ca–O bonds. Therefore, the absence

of Q3 and Q4 species in a composite with a high calcium

concentration weakens the mechanical properties of the structure.

The mean chain length, 2(Q1 + Q2 + Q3)/Q1, is an important

structural factor used to estimate the connectivity of the silicate

chains (Chen et al., 2004). Figure 11 shows that the mean silicate

chain length decreases with increasing calcium/silicon ratio. The

chain length in the simulated system is consistent with the trend

as revealed in experiments, implying that the simulated calcium

silicate hydrate gel can describe the inner silicate structure well.

The silicate chain length evolution as a function of calcium/

silicon ratio is an important characteristic for the calcium silicate

hydrate gel. Generally, nanostructure development depends

mainly on the silicate skeleton evolution. At a low calcium/

silicon ratio, cluster evolution is based on the growth of the

silicate chains. Calcium atoms fill the cavities in the silicate

deficient parts and do not influence the skeleton extension to

network or branch structures. On the other hand, at a high

calcium/silicon ratio, the cluster aggregation process results from

the associations of calcium ions and isolated silicate composites.

In this respect, a high calcium concentration prefers to form the

crystalline calcium silicate hydrate phase, while a silicon-rich

cluster favours amorphous gels. Therefore, from a low to a high

calcium/silicon ratio, the morphology of the calcium silicate

hydrate gel gradually transforms from an amorphous network

structure to an ordered ellipsoid-like structure.

According to the previous discussion, the morphology of calcium

silicate hydrate gel at nanoscale can be generally described. The

molecular mechanism for cluster formation produces some ideas

of how to optimise the structure of calcium silicate hydrate gel at

nanoscale. At a low calcium/silicon ratio, even though the silicate

skeleton, including Q3 and Q4 species, improves the stiffness of

the structure to some extent, the fibril morphology, characterised

by large nanoporosity, weakens its mechanical behaviour. On the

contrary, despite no Q3 species, the relatively dense and ordered

morphology of calcium silicate hydrate gel at a high calcium/

silicon ratio, with a low porosity, can maintain cohesive forces at

a high level. Therefore, an ‘optimum’ nanostructure should

attempt to reduce the nanoporosity and increase the percentage of

Q3 and Q4 species. To obtain an ‘optimum morphology’, an

appropriate calcium/silicon ratio, hydration temperature and the

introduction of mineral admixtures should be comprehensively

considered.

ConclusionTwo types of calcium silicate hydrate phases at the mesoscale – a

branch structure and an ellipsoid particle structure – were

produced by a molecular dynamics simulation. The discrepancy

in morphology is mainly a result of the calcium/silicon ratio. The

simulated high calcium/silicon ratio and low calcium/silicon ratio

phases are quite similar to the inner and outer products respec-

tively. In particular, it is valuable to note that the ellipsoid particle

of around 5 nm size is consistent with Jennings’ model (Jennings,

2000, 2008) and SANS results (Allen et al., 2007). The simula-

tion constructs a bridge between molecular and colloid morphol-

ogies and gives physical insights into the formation of colloid

particles in the cement hydration process. Further density,

composition, XRD and SANS tests confirm that the structural

features of the simulated calcium silicate hydrate gels match the

experimental results well. For the local structure, the bond length

distributions of Si–O and Ca–O imply short-range disorder of

the covalent and ionic connections inside the calcium silicate

hydrate gel. The defective silicate network and distorted Ca–O

octahedrons in the simulated calcium silicate hydrate gels

resemble the amorphous nature of silicate glass.

In order to understand the hydration mechanisms of the two

calcium silicate hydrate phases, the calcium silicate hydrate gel

evolution was analysed in terms of cluster growth and Q species

evolution. At a calcium/silicon ratio of 1.00, calcium silicate

hydrate gel growth is based on the silicate skeleton growth,

which is more likely to form an amorphous network and extends

in a radiative mode. On the contrary, at a calcium/silicon ratio

2

4

6

8

10

1·0 1·5 2·0 2·5 3·0

Mea

n ch

ain

leng

th

Calcium/silicon ratio

Experiment 1Experiment 2Simulation

Figure 11. Mean chain length: comparison of NMR tests (Chen et

al., 2004) and simulation result

10


of 1.75, calcium atoms de-polymerise the network formation of

the silicate structures and the branch structure (Q3) is not likely

to form. The cluster nucleates to an ellipsoid structure. Calcium

and silicon influence the cluster morphologies in different ways:

the former contributes to an ordered structure while the latter

increases the amorphous state of the structure. Furthermore, a

wide range of calcium/silicon ratios should be taken into

consideration to observe morphology evolution; this study is

currently ongoing. Other ions rich in cement (e.g. Al3þ and

Naþ) will also be added to the simulation system to explore

their influence on the structure of hydration products at meso-

scale.

AcknowledgementThe China Ministry of Science and Technology (grant

2009CB623200) is gratefully acknowledged.

REFERENCES

Allen AJ, Thomas JJ and Jennings HM (2007) Composition and

density of nano-scale calcium silicate hydrate in cement.

Nature Materials 6(4): 311–316.

Brough A, Dobson C, Richardson IG and Groves G (1994) In situ

solid-state NMR studies of Ca3SiO5: hydration at room

temperature and at elevated temperatures using 29Si

enrichment. Journal of Materials Science 29(15): 3926–3940.

Chen JJ, Thomas JJ, Taylor HFW and Jennings HM (2004)

Solubility and structure of calcium silicate hydrate. Cement

and Concrete Research 34(9): 1499–1519.

Cong X and Kirkpatrick R (1996) 29Si MAS NMR study of the

structure of calcium silicate hydrate. Advanced Cement Based

Materials 3(3–4): 144–156.

Cormack A and Du J (2001) Molecular dynamics simulations of

soda-lime-silicate glasses. Journal of Non-Crystalline Solids

293–295(1): 283–289.

Dolado JS, Griebel M and Hamaekers J (2007) A molecular

dynamic study of cementitious calcium silicate hydrate (C-S-

H) gels. Journal of the American Ceramic Society 90(12):

3938–3942.

Dolado JS, Griebel M, Hamaekers J and Heberc F (2011) The

nano-branched structure of cementitious calcium-silicate-

hydrate gel. Journal of Materials Chemistry 21(12): 4445–

4449.

Du J and Cormack AJ (2004) The medium range structure of

sodium silicate glasses: a molecular dynamics simulation.

Journal of Non-Crystalline Solids 349(1): 66–79.

Faucon P, Delaye JM and Virlet J (1996) Molecular dynamics

simulation of the structure of calcium silicate hydrates: I.

Ca4þxSi6O14þ2x(OH)4�2x(H2O)2(0 <x< 1). Journal of Solid

State Chemistry 127(1): 92–97.

Feuston BP and Garofalini SH (1988) Empirical three-body

potential for vitreous silica. Journal of Chemical Physics

89(9): 5818–5824.

Feuston BP and Garofalini SH (1990) Oligomerization in

silica sols. Journal of Physics and Chemistry 94(13):

5351–5356.

Garofalini SH and Martin G (1994) Molecular simulations of the

polymerization of silicic acid molecules and network

formation. Journal of Physical Chemistry 98(4): 1311–1316.

Groves G (1987) TEM studies of cement hydration. Materials

Research Society Proceedings 85: 3–12.

Hamid S (1981) The crystal structure of the 11 A natural

tobermorite Ca2:25Si3O7:5(OH)1.5H2O. Zeitschrifit fur

Kristallographie 154(3–4): 189–198.

Janika JA, Kurdowsk W, Podsiadey R and Samset J (2001) Fractal

structure of CSH and tobermorite phases. Acta Physica

Polonica A 100(4): 529–537.

Jennings HM (2000) A model for the microstructure of calcium

silicate hydrate in cement paste. Cement and Concrete

Research 30(1): 101–116.

Jennings HM (2008) Refinements to colloid model of C-S-H in

cement: CM II. Cement and Concrete Research 38(3): 275–

289.

Lammps (2008) LAMMPS Molecular Dynamics Simulator. See

http://lammps.sandia.gov/ (accessed 11/01/2012).

Li Z (2011) Advanced Concrete Technology. Wiley, Hoboken, NJ,

USA.

Ma H and Li Z (2013) Realistic pore structure of Portland cement

paste: experimental study and numerical simulation.

Computers and Concrete 11(4): 317–336.

Ma H, Tang S and Li Z (2013) New pore structure assessment

methods for cement paste. Journal of Materials in Civil

Engineering, http://dx.doi.org/10.1061/(ASCE)MT.1943–

5533.0000982, in press.

Martin SI (1994) Synthesis of Tobermorite: A Cement Phase

Expected Under Repository Conditions. Lawrence Livermore

National Laboratory, Livermore, CA, USA.

Mastelaro VR, Zanotto ED, Lequeux N and Cortes RJ (2000)

Relationship between short-range order and ease of

nucleation in Na2Ca2Si3O9, CaSiO3 and PbSiO3 glasses.

Journal of Non-Crystalline Solids 262(1–3): 191–199.

Mead RN and Mountjoy G (2006) A molecular dynamics study

of the atomic structure of (CaO)x(SiO2)1�x glasses. Journal

of Physical Chemistry B 110(29): 273–278.

Merlino S, Bonnacorsi E and Armbruster T (2001) The real

structure of tobermorite 11 A: normal and anomalous forms,

OD character and polyptic modifications. European Journal

of Mineralogy 13(3): 577–590.

Pellenq RJM, Kushima A, Shahsavari R et al. (2009) A realistic

molecular model of cement hydrates. Proceedings of the

National Academy of Sciences of the United States of

America 106(38): 16 102–16 107.

Rao NZ and Gelb LD (2004) Molecular dynamics simulation of

the polymerization of aqueous silicic acid and analysis of the

effects of concentration on silica polymorph distributions,

growth mechanisms and reaction kinetics. Journal of Physical

Chemistry B 108(33): 12 418–12 428.

Richardson IG (2004) Tobermorite/jennite-and tobermorite/

calcium hydroxide-based models for the structure of C-S-H:

applicability to hardened pastes of tricalcium silicate,

h-dicalcium silicate, Portland cement, and blends of Portland

11


http://lammps.sandia.gov/

http://dx.doi.org/10.1061/(ASCE)MT.1943--5533.0000982,




cement with blast-furnace slag, metakaol. Cement and

Concrete Research 34(9): 1733–1777.

Richardson IG and Groves GW (1993) Microstructure and

microanalysis of hardened ordinary Portland cement pastes.

Journal of Materials Science 28(1): 265–277.

Sasaki K, Masuda T, Ishida H and Mitsuda T (1996) Synthesis of

calcium silicate hydrate with Ca/Si ¼ 2 by mechanochemical

treatment. Journal of the American Ceramic Society 80(2):

472–476.

Shahsavari R, Buechler MJ, Pellenq RJM and Ulm FJ (2009) First-

principles study of elastic constants and interlayer

interactions of complex hydrated oxides: case study of

tobermorite and jennite. Journal of the American Ceramic

Society 92(10): 2323–2330.

Taniguchi T, Okuno M and Matsumoto TJ (1997) X-ray

diffraction and EXAFS studies of silicate glasses containing

Mg, Ca and Ba atoms. Journal of Non-Crystalline Solids

211(1–2): 56–63.

Viallis-Terrisse H, Nonat A and Petit JC (2001) Zeta-potential

study of calcium silicate hydrates interacting with alkaline

cations. Journal of Colloid and Interface Science 244(1):

58–65.

WHAT DO YOU THINK?

To discuss this paper, please submit up to 500 words to

the editor at www.editorialmanager.com/acr. Your con-

tribution will be forwarded to the author(s) for a reply

and, if considered appropriate by the editorial panel, will

be published as a discussion in a future issue of the

journal.

12


Morphology of calcium silicate hydrate (C-S-H) gel: a molecular ...

Documents

Transcript of Morphology of calcium silicate hydrate (C-S-H) gel: a molecular ...