You are on page 1of 11

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/229350363

Elastic Properties of the Main Species Present in Portland Cement Pastes

Article  in  Acta Materialia · March 2009


DOI: 10.1016/j.actamat.2008.12.007

CITATIONS READS

110 213

3 authors:

Hegoi Manzano J.s. Dolado


Universidad del País Vasco / Euskal Herriko Unibertsitatea Tecnalia
72 PUBLICATIONS   1,221 CITATIONS    80 PUBLICATIONS   2,106 CITATIONS   

SEE PROFILE SEE PROFILE

A. Ayuela
Universidad del País Vasco / Euskal Herriko Unibertsitatea
166 PUBLICATIONS   3,822 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

InnovaConcrete: Innovative materials and techniques for the conservation of 20th century concrete-based cultural heritage View project

CEMFF: A force field database for cementitious materials including validations, applications and opportunities View project

All content following this page was uploaded by Hegoi Manzano on 19 March 2018.

The user has requested enhancement of the downloaded file.


This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy

Available online at www.sciencedirect.com

Acta Materialia 57 (2009) 1666–1674


www.elsevier.com/locate/actamat

Elastic properties of the main species present in Portland cement pastes


H. Manzano a,*, J.S. Dolado a,b, A. Ayuela c
a
LABEIN-Tecnalia, Parque Tecnológico de Bizkaia, Edificio 700, 49160 Derio, Spain
b
Nanostructured and Eco-efficient Materials for Construction Unit, Associated Unit LABEIN-Tecnalia/CSIC, Spain
c
Unidad de Fı́sica de Materiales, Centro Mixto CSIC-UPV/EHU and Donostia Internacional Physics Center, PO Box 1072, San Sebastián/Donostia, Spain

Received 29 August 2008; received in revised form 4 December 2008; accepted 4 December 2008
Available online 8 January 2009

Abstract

The elastic properties of the most important phases present in cement pastes have been studied by force field atomistic methods. Cal-
culations reproduce the mechanical properties of crystalline species alite, belite and portlandite fairly well. The elastic properties of amor-
phous calcium–silicate–hydrate (C–S–H) gel are also explained in terms of the commonly used structural models 1.4 nm tobermorite and
jennite, once the intrinsic porosity and the silicate chain length are concurrently considered through the appropriate micromechanical
models. The findings conclude that bringing the atomic structure into contact with the mechanical properties of C–S–H gel may modify
the actual view of the gel.
Ó 2008 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Calcium–silicate–hydrate (C–S–H); Crystal structure; Elastic behavior; Atomic potentials; Nanostructure

1. Introduction Thus, the topic of this work is understanding their struc-


tural and micromechanical characteristics and how they
The cement matrix is a complex material with many are linked to the mechanical behavior of cements.
crystalline phases. Both unreacted species and hydration The mechanical properties of alite and belite have
products are present and agglomerated by an amorphous been the object of different studies in synthetic and
product, the calcium–silicate–hydrate (C–S–H) gel. In in situ samples, as well as in the perfect stoichiometric
ordinary Portland cement, the raw materials are limestone structures C3S and C2S. When measured with techniques
and clay. This implies a typical composition of clinkers of nanoindentation and resonance frequency, the elastic
with 67% CaO and 22% SiO2 with mainly alite (C3S) moduli of alite and belite are similar, with values
and belite (C2S) species, where in cement chemistry nota- between 130 and 150 GPa [1,2]. Portlandite is also a
tion C stands for CaO, S for SiO2 and H for H2O. After well-known mineral, and its mechanical properties have
hydration of the clinker, the main components formed been determined by many methods such as Brillouin
are portlandite (CH) and C–S–H gel. The C–S–H gel is spectroscopy, X-ray spectroscopy and first-principles
the most important hydration product, up to 70% of the atomistic calculations based on density functional theory
final volume, and it is responsible for the cohesion and (DFT). Several nanoindentation measurements obtained
mechanical properties of cement pastes. However, when similar Young’s moduli (E) for portlandite: 36 GPa [2]
one considers the intrinsic porosity in cement pastes, these and 40 GPa [3]. These results are in agreement with (i)
properties are not only dependent on the C–S–H gel, but the extrapolations of different samples to zero porosity,
also on the contribution of previous unhydrated species. E = 35.2 GPa [4] and E = 48 GPa [5], and also with (ii)
the values calculated by Monteiro and Chang [6], where
*
Corresponding author. Tel.: +34 946073300; fax: +34 946073349.
E ranges between 39.8 and 42.2 GPa after numerical
E-mail addresses: hmanzano@labein.es, hegoi.manzano@gmail.com analysis of data measured by Brillouin spectroscopy [7]
(H. Manzano). and X-ray diffraction [8].

1359-6454/$34.00 Ó 2008 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.actamat.2008.12.007
Author's personal copy

H. Manzano et al. / Acta Materialia 57 (2009) 1666–1674 1667

Regarding C–S–H gel, different nanoindenting tests expected to interact with each other in different ways.
show the existence of two types of gels in terms of their Therefore, before any simulation, the employed sets of
elastic properties: a low stiffness (LS) C–S–H gel with the potentials must be carefully checked with the experimental
indentation modulus M  19 GPa and a high stiffness data to assess their transferability.
(HS) variety with M  28 GPa [2,9]. This bimodal distribu- The starting subset of interatomic potentials is the one
tion in the indentation modulus fits perfectly with Jennings’ employed in Ref. [15] to simulate a tobermorite-like struc-
model [10,11]. In this model, globular 5 nm particles pack ture. Preliminarily, this scheme was used to describe the
themselves into a loose low density (LD) C–S–H gel with mechanical properties of many crystalline-related C–S–H
a packing efficiency gLD = 64%, or into a high density phases [16]. The interactions between silicate ions are based
(HD) C–S–H gel with gHD = 74%. Several works [3,9] have on the well-known potentials for zeolites [17–19]. The Si–O
further advanced the correlation between LS–HS and LD– interaction uses Buckingham potentials with a three-body
HD varieties. They obtained the C–S–H indentation mod- harmonic potential to simulate the bond hybridization.
ulus using the back analysis procedure. These works iden- The Ca–O ionic interactions are modeled by Buckingham
tify the Jennings globule with a single crystal, which has an repulsive potentials [20]. The intermolecular interactions
indentation modulus Ms  60 GPa and a Young’s modulus between water molecules and other species are described
E  55 GPa [3,12]. These works provide strong evidence using Buckingham potentials, whereas the interaction
for the existence of a unique C–S–H nanoparticle. The pic- between different water molecules are described using a
ture that emerges from coupling Jennings’ model with nan- Lennard–Jones potential. The intramolecular interactions
oindenting results is also supported by recent atomistic within water molecules are simulated with the Buckingham
simulations which found Young’s modulus for Hamid’s potential and a three-body harmonic potential for the H–
structure (C/S = 1) to be E = 57.1 GPa [13]. O–H angle [21]. The coulomb interaction within water mol-
The present work calculates the elastic properties of the ecules has been subtracted. Thus, the charges of hydrogen
calcium–silicate-based species present in cement pastes and oxygen atoms serve only to describe their interaction
using an atomistic method. Specifically, bulk, shear and with other atoms, and not to determine the structure in
Young’s moduli are calculated together with their Poisson water molecules. By eliminating this coulombic term first,
ratio. Although this method is appropriated for crystalline one drastically reduces the number of interactions that
phases, a direct calculation of the elastic properties of C–S– must be computed and adds more flexibility to the config-
H gel is impossible, owing to its unsettled atomic structure. uration of water. The polarizability of oxygen atoms is
To overcome the problem, a starting point is taken using described by the core–shell approach [22]. In this model,
the related crystalline structures of 1.4 nm tobermorite the atoms are treated as internal cores linked to external
and jennite. Then, well-known (nano)characteristics of massless shells. Both the core and the shell are linked by
the C–S–H gels were applied to these crystals. Specifically, a spring potential and have specific charges. It must be
intrinsic nanoporosity and the finite length of silicate noted that not all oxygen or calcium atoms are equal.
chains are accounted for. This scheme can reasonably bring The environment surrounding a specific atom has a great
into contact the bulk crystalline properties with those of C– influence on its chemical behavior, bond distances, bond
S–H gels. This process adds new insights for C–S–H gels strengths and other properties. Hence, the differences are
into the links between structure and mechanical properties. distinguished between atoms with different values for the
corresponding potential parameters (see Table 1). (Oxygen
2. Force field methods atoms can be in water, hydroxyl groups, oxides or silicate
chains. Calcium atoms can be bonded to oxygen, water
The calculations use the force field method [14]. The or OH groups.)
atoms are represented as spheres, where their electronic The simulations were done with the GULP code [23].
nature is taken into account implicitly with some parame- The experimental data for the crystalline structures were
ters, such as their charge and their radius. The interaction optimized by relaxing the unit cell parameters and the
between atoms is described by parameterized interatomic atomic forces. The symmetries were eliminated once the
potentials so the system energy can be traced back from experimental full unit cells were generated to allow aniso-
their position. The force field method gives accurate infor- tropic variations, even if the atoms were in specific symme-
mation on structures, thermodynamic properties and try positions. The search of local minima followed the
mechanical properties with a low computational effort. Newton–Raphson procedure with the Broyden–Fletcher–
The election of suitable potentials is the cornerstone of Goldfarb–Shannon (BFGS) [24] scheme to update the Hes-
any force field simulation. There are a large number of sian. Then, the elastic constant tensor was calculated from
published potentials for a wide range of systems. However, the second derivative matrix in the optimized structures.
there is no guarantee that certain potentials could properly The bulk (K) and shear (G) moduli were determined from
simulate structures different from the originals used to fit the elastic constants according to the averaged Hill defini-
them. The potentials are usually built up to describe the tion [25]. The average Young’s modulus (E) and Poisson’s
interactions of atoms under a given neighborhood. Conse- ratio (t) were calculated through the standard relations for
quently, the same atoms in different environments are isotropic media:
Author's personal copy

1668 H. Manzano et al. / Acta Materialia 57 (2009) 1666–1674

Table 1
Potentials and parameters employed in this work*.
Species Charge core Charge shell Spring constant Ref.
Si 4.00 [17,18]
Cainter–intra–port 2.00 [20]
Ow 1.25 2.05 209.45 [21]
Osi 0.87 2.87 74.92 [17–19]
Ooh–port 0.87 2.27 74.92 [19]
Hw–oh–port 0.40 [21]
Water intramolecular potentials
Morse D (eV) a (Å2) H (Å) Cut off (Å)
Ow–Hw 6.2 2.22 0.924 0.0–1.4 [21]
Three-body harmonic k (eV rad2) h (°) Cut off 1–2–3 (Å)
Ow–Hw–Hw 4.2 108.69 0.0–1.2/0.0–1.2/0.0–2.4 [21]
1.4 nm tobermorite, jennite
Buckingham A (eV) q (A) C (eV Å6) Cut off (Å)
Si–O 1283.907 0.32052 10.66158 0.0–12.0 [17,18]
Si–Ooh 983.557 0.32052 10.66158 0.0–12.0 [17,18]
Si–Ow 983.557 0.32052 10.66158 0.0–12.0 [19]
Cainter–Osi 1090.4 0.3437 0.0–12.0 [20]
Cainter–Ooh 777.27 0.3437 0.0–12.0 [20]
Cainter–Ow 777.27 0.344 0.0–12.0 [20]
Caintra–Osi 1090.4 0.3437 0.0–12.0 [20]
Caintra–Ooh 1090.4 0.3437 0.0–12.0 [20]
Caintra–Ow 777.27 0.3437 0.0–12.0 [20]
Osi–Osi 22,764.0 0.32052 27.879 0.0–12.0 [17,18]
Osi–Ooh 22,764.0 0.149 13.94 0.0–12.0 [17,18]
Osi–Ow 22,764.0 0.149 0.149 0.0–12.0 [19]
Ooh–Ooh 22,764.0 0.149 6.97 0.0–12.0 [19]
Ooh–Ow 22,764.0 0.149 0.149 0.0–12.0 [19]
Ooh–Hw 311.97 0.25 0.0–12.0 [19]
Ow–Hw 396.0 0.25 0.0–12.0 [21]
Ow–Hoh 311.97 0.25 0.0–12.0 [19]
Morse D (eV) a (Å2) q (Å) Cut off (Å)
Ooh–Hoh 7.0525 3.1749 0.9428 0.0–1.4 [19]
Lennard–Jones A (eV Å6) B (eV Å12) Cut off (Å)
Ow–Ow 39.344.98 42.15 0.0–12.0 [21]
Three-body harmonic k (eV rad2) h (°) Cut off 1–2–3 (Å)
Si–O–Si 2.0972 109.47 0.0–1.8/0.0–1.8/0.0–3.2 [17,18]
Portlandite
Morse D (eV) a (Å2) q (Å) Cut off (Å)
Oport–Hport 6.2 2.22 0.924 0.0–1.4 [21]
Lennard–Jones A (eV Å6) B (eV Å12) Cut off (Å)
Oport–Oport 39.344.98 42.15 0.0–12.0 [21]
Buckingham A (eV) q (A) C (eV Å6) Cut off (Å)
Caport–Oport 777.27 0.3437 0.0 0.0–12.0 [20]
*
The subscripts indicate the atomic environments: oh, hydroxyl group; si, SiO4 group; w, water molecule; port, portlandite; inter, interlaminar atom;
intra, intralaminar atom.

9G sizes large enough to be seen with nanoindentation measure-


E¼ ð1Þ
3 þ G=K ments, and the measured elastic properties are compared
with the computed results. Secondly, the complex C–S–H
and gel is examined by applying a homogenizing sub-micro-
3  2G=K scopic model with defects in the involved crystalline compo-
m¼ ð2Þ nents, which are 1.4 nm tobermorite and jennite (Fig. 2).
6 þ 2G=K
3.1. Highly ordered phases present in cement pastes
3. Results and discussion
3.1.1. C3S
The results are presented in two different sections. First, The computed values for the cell parameters are shown
C3S, C2S and portlandite are studied (see Fig. 1). They are in Table 2. The unit cell parameters are in good agreement
present in the cement matrix as crystalline structures with with the experiments [26] with errors <2.5%. The atomic
Author's personal copy

H. Manzano et al. / Acta Materialia 57 (2009) 1666–1674 1669

Fig. 1. Geometries of crystalline phases. Optimized structures of: (a) C3S in the xz plane; (b) C2S in the xz (right panel) and the yz plane (left panel); (c)
portlandite in the xz plane. Silicon tetrahedra are in dark purple, calcium atoms in large orange spheres, oxygen atoms in small red spheres/sticks, and
hydrogen atoms in light white sticks. (For interpretation of color mentioned in this figure the reader is referred to the web version of the article.)

Fig. 2. Crystals related to C–S–H gel: optimized structures of: (a) 1.4 nm tobermorite; (b) jennite. The left and right panels are the projections in the xz
and yz planes, respectively. The atom labels are given as in the previous figure. Additionally, water molecules are indicated by joined sticks.

arrangement and the coordination tetrahedra do not dentation measurements with a result of E = 135 GPa
change greatly from the input experimental coordinates [1,2]. Using resonance frequency techniques and extrapo-
(see Fig. 1a). Also the computed values for the mechanical lating the results of samples with different porosities to zero
properties are in agreement with the experimental data porosity, a slightly larger value of E = 147 GPa was
given in Table 3. There are no direct experimental measure- obtained [1]. The theoretical result of E = 138.9 GPa is
ments of the bulk (K) and shear moduli (G) for this clinker between these two experimental values and in very good
phase. Young’s modulus (E) was determined by nanoin- agreement. The computed Poisson ratio, t = 0.28, is also
Author's personal copy

1670 H. Manzano et al. / Acta Materialia 57 (2009) 1666–1674

Table 2
Cell parameters of highly ordered phases present in cement pastes; the differences between experimental data and the computed values are also shown.
Alite (C3S) Belite (b-C2S) Portlandite (CH)
Exp. [26] This work % Error Exp. [28] This work % Error Exp. [29] This work % Error
a (Å) 33.08 33.45 +1.13 5.50 5.51 +0.16 3.59 3.55 1.17
b (Å) 7.02 7.12 +1.39 6.76 6.92 +2.44 3.59 3.55 1.17
c (Å) 18.50 18.91 +2.24 9.32 9.94 +6.58 4.91 4.94 +0.72
a (°) 90 90 0.0 90 90 0.0 90 90 0.0
b (°) 94.1 94.9 +0.81 94.1 94.0 0.17 90 90 0.0
c (°) 90 90 0.0 90 90 0.0 120 120 0.0

Table 3
Bulk (K), shear (G) and Young’s moduli (E) in GPa together with their Poisson ratio (t) for the crystalline phases present in cement pastes.
K G E m
Exp. This work Exp. This work Exp. This work Exp. This work
Alite (C3S) – 103.0 – 54.5 147–135 [1,2] 138.9 0.30 [27] 0.28
Belite (b-C2S) – 111.0 – 53.1 140–130 [1,2] 137.9 0.30 [27] 0.30
Portlandite (CH) 39.65 [6] 31.1 16.36 [6] 13.4 40.3 [3], 36 [2] 35.2 0.30–0.325 [6] 0.31

close to the estimated averaged value of t = 0.30 for all this anisotropic effect with values of Ex = Ey = 93.6 and
phase constituents in the clinker [27]. Ez = 32.8 GPa.

3.1.2. C2S 3.2. Cementitious C–S–H gel


The computed cell parameters for belite (Table 2) are
close to the experimental values [28], except for the c axis, 3.2.1. Mechanical properties of tobermorite and jennite
which increases 0.6 Å. Nevertheless, this expansion is Much of the actual knowledge about C–S–H gel is due
negligible regarding the atomic arrangement (shown in to its comparison with some crystals, specifically 1.4 nm
Fig. 1b) and the crystal properties. The silicate tetrahedron tobermorite and jennite (Fig. 2). Both 1.4 nm tobermorite
orientation and the coordination of Ca+2 ions reproduce and jennite have a sheet-like structure of Ca–O layers
the experimental data. Both the computed Young’s modu- ribbed by silicate chains. These chains where the Si–O tet-
lus (E = 137.9 GPa) and Poisson ratio (t = 0.30 GPa) are rahedra repeat themselves every three units follow the dre-
also in good agreement with the experimental data. Note ierketten arrangement. In 1.4 nm tobermorite, two of these
that the experimental Young’s moduli are 130 and tetrahedra are intimately connected to the Ca–O layer
140 GPa, obtained with nanoindentation and frequency (called pairing tetrahedra), and the remaining one is ori-
resonance, respectively [1,2]; and their t ratio values are ented to the interlaminar space (called the bridging tetrahe-
estimated as 0.30 [27], as shown in Table 3. dron). The final stoichiometry (Ca5Si6O16(OH)27H2O with
Ca/Si ratio 0.83) was resolved in Ref. [31] along with its
3.1.3. Portlandite exact crystalline structure. Jennite has a similar structure,
After relaxing the experimental input parameters, the with two differences. First, only half the oxygen atoms
unit cell of portlandite keeps the trigonal symmetry (see cell belonging to the Ca–O layer are linked to the silicate
parameters a and b between layers in Fig. 1c). The output chains, whereas the rest are bonded to OH groups and to
cell dimensions are in agreement with experimental data an extra Ca–O sheet. Secondly, the spatial arrangement
[29], and their differences are slightly >1%, as listed in of silicate chains is slightly different owing to this extra
Table 2. The present calculated value for Young’s modulus sheet that bridges the tetrahedra and partially connects
is E = 35.4 GPa, close to the value obtained experimen- them to the Ca–O layers (see Fig. 2). The stoichiometry
tally. Furthermore, Monteiro and Chang [6] obtained the (Ca9Si6O18(OH)65H2O with a Ca/Si = 1.5) and its crystal-
values of the bulk and shear moduli together with their line structure were determined in Ref. [32].
Poisson ratio. The present calculated values for these mag- The crystalline structures of 1.4 nm tobermorite and jen-
nitudes are smaller, but they are also in agreement with the nite were optimized. The cell parameters and structures
experimental data, as shown in Table 3. The elastic proper- after optimization are shown in Table 4 and Fig. 2. The
ties of portlandite crystals have also been calculated by first small variation in the volume of both species is 3.5%,
principles using DFT [30]. In such work, Young’s moduli mainly due to the cell expansion in the z axis and the per-
in the axis directions of the unit cell are anisotropic because pendicular direction to the layers. This expansion can be
of the layered structure of portlandite, with Ex = Ey = 89.0 attributed to the shielding effect that the interlayer water
and Ez = 38.6 GPa. Although not discussed here, it should molecules made in the dispersive and coulombic attraction
be mentioned that the present calculations also reproduce between different sheets. Concerning mechanical proper-
Author's personal copy

H. Manzano et al. / Acta Materialia 57 (2009) 1666–1674 1671

Table 4
Cell parameters for highly ordered phases, 1.4 nm tobermorite and jennite, related to C–S–H gel; the differences between experimental data and computed
values are also shown.
1.4 nm tobermorite Jennite
Exp. [31] This work % Error Exp. [32] This work % Error
a (Å) 6.74 6.76 +0.29 10.57 10.63 +0.56
b (Å) 7.42 7.34 1.09 7.26 7.34 1.10
c (Å) 27.99 29.15 +4.14 10.93 11.13 +1.82
a (°) 90 90 0.0 101.3 102.7 +1.38
b (°) 90 90.3 +0.33 97 93.9 3.19
c (°) 123.3 122.9 0.32 109.7 110.1 +0.36

ties, Poisson’s ratio (t) for both 1.4 nm tobermorite and built a bulk material of tobermorite-like species (T), and
jennite are in reasonably agreement with the experimental jennite-like species (J), formed by dimeric (T2, J2), penta-
data for C–S–H gel [33]. This agreement can be explained meric (T5, J5) and octameric (T8, J8) chains. The omission
because t is related to the ratio between elastic moduli. of bridging tetrahedra in jennite does not change the total
However, the calculated Young’s moduli (E) of tobermor- charge of the unit cell when one eliminates a silicon atom
ite and jennite are 49.9 and 56.0 GPa, respectively, which (+4) and two oxygen atoms (2). In the case of tobermor-
are much larger than those of experimental findings for ite, one has to remove a silicon atom (+4), an oxygen (2)
C–S–H gels, with E  16–25 GPa [3,9]. The calculations and an OH group (1). With this procedure, the system
of both bulk (K) and shear (G) modulus with values of would gain one electron. To return the unit cell to its neu-
44.8 and 40.2 GPa, respectively, are also overestimating tral state, two possibilities have been considered. In the first
the experimental data of C–S–H gels. It was found experi- case, the extra charge has been compensated by saturating
mentally that K = 18 GPa [34]. To the best of the authors’ half of the terminal oxygen atoms of silicate chains with H,
knowledge, the shear moduli of C–S–H gel have not been so that they form terminal OH groups (species denoted by
measured, although an approximated value obtained after TOHn). In the second case, the necessary number of calcium
a homogenization analysis gives G  9.7 GPa [9]. The cal- atoms has been introduced in the interlayer space to coun-
culated G values are 19.0 and 22.1 GPa for 1.4 nm toberm- terbalance the charge (TCan).
orite and jennite, respectively, which are again significantly The relaxation of these new tobermorite-like and jen-
larger than those in the experiments. nite-like structures with defects leads to small changes in
The disagreement between experimental and theoretical their structure. For tobermorite-like species, the main
values has been traditionally attributed to (nano)porosity- change is the movement of interlayer Ca ions towards the
related effects. The intrinsic values for the Young’s modu- vacancy left after removing the bridging silicon tetrahe-
lus of C–S–H gel have been estimated by reverse analysis dron. For jennite-like structures, the absence of bridging
based on micromechanical models. The Mori–Tanaka tetrahedra implies a slight atomic rearrangement in the
(MT) model gives E  47 GPa [35], and the self-consistent Ca–O layer. The new terminal oxygens, previously bonded
(SC) model gives E  65 and 60 GPa [3,12]. These values to the silicon bridging atom, now enter in coordination
compare well with the prediction reported by Pellenq and with this Ca–O layer, and change the orientation of the sil-
Van Damme (E = 57.1 GPa) [13]. The experimental values icate chains. This possibility of charge compensation makes
are also in agreement with the computed data for 1.4 nm the Ca interlayer atoms remain in their positions. The large
tobermorite (E = 48 GPa) and jennite (E = 56 GPa). How- atomic rearrangement in jennite-like structures does not
ever, several experimental studies by 29Si nuclear magnetic happen in those based in tobermorite.
resonance show that the chain length has discrete values The mechanical properties for these shortened species
following the 3n  1 rule: those formed with n = 2, 5 and are listed in Table 5. As expected, the mechanical proper-
sometimes 8 silicon tetrahedra are especially stable [36– ties of both species increase as the chain length increases,
38]. Consequently, the picture drawn is still incomplete, owing to the stability given by the longer silicate chains
and the situation is far more complex and intriguing. As attached to the Ca–O layers. The main improvement in
advanced in a preliminary work [16], it seems dubious (or the elastic properties is due to the incorporation of the first
at least not completely justified) that the intrinsic mechan- bridging tetrahedron, which merges two dimer structures
ical properties of solid C–S–H particle could be directly into a pentamer. This suggestion is supported by the
correlated with those of perfect crystalline structures. non-lineal relationship between the mechanical properties
and the chain length. The mechanical properties have the
3.2.2. Tobermorite-like and jennite-like structures with largest change between dimers and pentamers, whereas
shortened silicate chains the rise from pentamers to octamers is smaller. For
There is overall agreement that the silicate chain struc- tobermorite-like species, the structures with pentamer
ture of C–S–H gel is formed by the omission of bridging chains have almost the same properties as their infinite
tetrahedra [39,40]. Based on this statement, the authors crystals. For jennite-like structures, the mechanical proper-
Author's personal copy

1672 H. Manzano et al. / Acta Materialia 57 (2009) 1666–1674

Table 5
Calculated bulk (K), shear (G) and Young’s moduli (E) in GPa and their Poisson ratio for the tobermorite-like and jennite-like species.*
Ca/Si ratio K G E m
TOH2 1.25 26 14 35.61 0.27
TOH5 1.00 33.7 18.5 46.92 0.27
TOH8 0.94 36.8 19.1 48.85 0.28
TCa2 1.50 25 15.7 38.95 0.24
TCa5 1.10 33.9 18.7 47.39 0.27
TCa8 1.00 47.8 18.5 49.16 0.33
T1 0.83 44.8 19 49.94 0.31
J2 2.25 23.9 13.8 34.72 0.26
J5 1.80 31.8 18.4 46.27 0.26
J8 1.69 37.8 21 53.16 0.27
J1 1.50 40.2 22.1 56.03 0.27
Exp. data C–S–H gel 0.7–2.3 [40] 18 [34] 9.7 [9] 19–28 [9] 0.25 [33]
*
T denotes tobermorite; J, jennite. The subscripts OH and Ca refer to the charge used compensation method (see text). The end number is the chain length
n, where n is the number of silicon atoms.

ties are not close to the crystal ones when using the penta- with air void inclusions. Thus, the mechanical properties of
mer chains, and one must look to the values of octamer C–S–H gel are written as a function of those values for
chains. This finding could be explained by the bridging tet- solid phases (i.e., those of the Tn and Jn structures) and
rahedra playing a more active role here in the stabilization of certain packing factors (g) which are related to the gel
of Ca–O layers than in tobermorite. The differences porosity (p) through the equation g = 1  p. Within the
between the tobermorite-like species balanced by OH SC approach [43], the bulk and shear moduli of C–S–H
(TOH) groups and by Ca ions (TCa) are also examined. gel (KC–S–H gel and GC–S–H gel) are given in the following
The added calcium ions take the place of removed silicon expressions:

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
GC–S–H gel 1 5 3 1 2
¼  ð1  gÞ  r0 ð2 þ gÞ þ 144ð1  r0  480g þ 400g2 þ 408r0 g  120r0 g2 þ 9r20 ð2 þ gÞ ð3Þ
G0 2 4 16 16

tetrahedra and hold together, giving continuity to the sili-


cate chains. Nevertheless, there is no significant improve- 4gG
ment in the mechanical properties, which change from K C–S–H gel G0
¼ 4G
ð4Þ
E = 35.6 GPa to E = 38.9 GPa in the case of dimers. This K0 Goþ3ð1gÞr0
effect is even less important in the case of longer chains.
where g is the packing factor, K0 and G0 are the bulk and
In addition, it was observed that the Ca/Si ratios increase
the shear modulus of crystalline species, and r0 is a param-
from 0.83 in tobermorite and 1.5 in jennite to values in
eter defined as the ratio K0/G0. In the MT approach [44],
the range between 0.94 and 2.25, respectively. This rise is
the bulk and shear modulus are written as
in agreement with the ideas in Refs. [41,42]: the higher
Ca/Si ratio of C–S–H gel with respect to tobermorite and K C–S–H gel 1a
¼g ð5Þ
jennite crystals is due to the omission of bridging tetrahe- K0 1  ag
dra and the incorporation of extra Ca ions to balance the GC–S–H gel 1b
charge, which was the authors’ procedure to build the ¼g ð6Þ
G0 1  bg
highly ordered phases related to C–S–H gel.
where g is again the packing factor, and a and b are two
To sum up this section, in the present simulations there
parameters dependent on the bulk and the shear modulus
is a noticeable disparity between the mechanical properties
of crystalline species (K0 and G0):
of T2 and J2 and those with longer tobermorite- and jen-
nite-like structures (T5, J5, T8 and J8). 3K 0
a¼ ð7Þ
ð3K 0 þ 4G0 Þ
3.2.3. Mechanical properties of C–S–H gel
ð6K 0 þ 12G0 Þ
Following Refs. [3,12], the properties of the Tn and Jn b¼ ð8Þ
crystals are scaled up to those of C–S–H gel using the ð15K 0 þ 20G0 Þ
homogenization procedure of the SC and the MT micro- Thus, the values obtained for both models are applied to
mechanical models. Here the C–S–H gel is pictured as a Eqs. (1) and (2), and Young’s modulus and Poisson’s ratio
composite made of a solid phase (the Tn and Jn structures) are recalculate. For a better comparison with the
Author's personal copy

H. Manzano et al. / Acta Materialia 57 (2009) 1666–1674 1673

experimental data, the indentation modulus (M) of C–S–H ture the range measured for LS- and HS-C–S–H gels.
gels is calculated using The main difference between both schemes depends on
E the change in indentation modulus with the packing, espe-
M¼ ð9Þ cially for the SC model. Within the SC scheme, the packing
ð1  m2 Þ
factor is the main reason for the change from the LS region
Both SC and MT models are applied with two packing fac- to the HS one. In comparison, the MT gap between the LS
tors. The g values are 0.64 for the LD-C–S–H gel and 0.74 and HS regions can be crossed by changing the silicate
for the HD-C–S–H gel. The indentation modulus of C–S– chain length. Several conclusions can be drawn, in spite
H gels (MC–S–H gel) as a function of the length of Tn and Jn of the micromechanical model employed. The HS-C–S–H
crystals is displayed in Fig. 3a and b for the SC and the MT gel is composed only of highly polymerized silicate chains.
scheme, respectively. The triangles indicate low packing From the present results, T2 and/or J2 units can never par-
g = 0.64, while the squares indicate high packing ticipate in the HS arrangements. Additionally, the short sil-
g = 0.74. The experimental ranges found in Ref. [9] for icate chains are always involved in the description of LS-C–
LS-C–S–H and HS-C–S–H gels are also shown as S–H gel. Whether present or not, longer chains in the LS-
shadowed areas. For both micromechanical models, the C–S–H gel cannot be fully elucidated from the present sim-
stiffness depends clearly on the packing factor and the ulations, as they provide a different conclusion, depending
silicate chain length. As expected, the indentation modulus on the micromechanical model employed. Further detailed
increases with the packing factor and with the silicate chain correlations between experiment and theory will require
length. Moreover, it is remarkable that both models cap- additional C–S–H measurements and calculations.

4. Conclusions
a 35
J This work investigated, through force field atomistic
TCa simulations, the elastic properties of the most important
Indentation Modulus (M) GPa

TCa
30 TOH
phases present in Portland cement pastes. The following
TOH
J
HS were proved.
J
25 TCa
TCa TCa
TOH (i) The lattice parameters of studied crystalline species
TOH TOH
J J (C3S, C2S and CH) are in agreement with the exper-
20 TCa imental results. This agreement corroborates the suit-
TOH LS
J ability of the method and the transferability of the
15 potential parameters. The predictions for their
η = 0.64) mechanical properties are also in agreement with
MT (η
(η = 0.74) the experimental data obtained by nanoindentation
10 MT
and other techniques, such as Brillouin spectroscopy
2 3 4 5 6 7 8 and frequency resonance.
Silicate Chain Length (n) (ii) Simulations have shown that the mechanical proper-
ties of 1.4 nm tobermorite and jennite (usually
b 35 employing structural models of CSH gel) are about
twice as great as those of C–S–H gel. Nevertheless,
J when applying characteristics such as intrinsic poros-
Indentation Modulus (M) GPa

30
TCa TCa HS ity and the finite chain length of the C–S–H gel to the
TOH
TOH crystals, the mechanical properties decrease and reach
25 J values in excellent agreement with nanoindentation
TCa
measurements for C–S–H gels.
20 TOH J (iii) Though the elastic properties of C–S–H gels have
J TCa TCa been traditionally explained in terms of solely poros-
TOH LS
TOH ity-related effects, simulations show that the silicate
15 TCa J
chain length also matters. In fact, the results indicate
TOH that the longer the silicate chains, the better the
J
SC (η = 0.64)
10 SC (η = 0.74) mechanical properties.
(iv) Interesting linkages are suggested between the atomic
2 3 4 5 6 7 8 structure and the mechanical properties of LS- and
Silicate chain length (n) HS-C–S–H gels. According to the simulations, the
Fig. 3. Indentation modulus vs silicate chain length for: (a) SC microme- mechanical properties of HS-C–S–H gel can only be
chanical model; (b) MT model. Squares indicate HD values (g = 0.74); explained if it consists of pentamers and/or larger sil-
triangles indicate low densities (g = 0.64). icate units. Unfortunately, the simulations cannot
Author's personal copy

1674 H. Manzano et al. / Acta Materialia 57 (2009) 1666–1674

unequivocally determine the units which make up the [13] Pellenq RJM, Van Damme H. MRS Bull 2004;29:319.
LS-C–S–H gel. Further work combining experiment [14] Leach AR. Molecular modeling, principles and applications. 2nd
ed. Harlow: Pearson Education; 2001.
and theory must be done in order to clarify this [15] Gmira A. Thesis, University of Orléans, France; 2003.
situation. [16] Manzano H, Dolado JS, Ayuela A, Guerrero A. Phys Sta Sol (a)
2007;204:1775.
[17] Schroder KP, Sauer J, Leslie M, Catlow CRA, Thomas JM. Chem
Acknowledgements Phys Lett 1992;188:320.
[18] Gale J, Henson NJ. J Chem Soc Faraday Trans 1994;90:3175.
[19] Du Z, De Leeuw NH. Surf Sci 2004;554:193.
This work was supported by the Basque Government [20] De Leeuw NH, Watson GW, Parker SC. J Phys Chem 1995;99:17219.
through the NANOMATERIALS Project (IE05-151) un- [21] De Leeuw NH, Parker SC. Phys Rev B 1998;58:13901.
der the ETORTEK Program (NANOMAT), Spanish Min- [22] Dick BG, Overhauser AW. Crystals Phys Rev 1958;112:90.
isterio de Ciencia y Tecnologı́a (MCyT) of Spain (Grant [23] Gale JD. J Chem Soc Faraday Trans 1997;93:629.
[24] Shannon DF. Math Comput 1970;24:647.
No. Fis 2004-06490-C03-00 and MONACEM Project) [25] Nye JF. Physical properties of crystals. New York: Oxford University
and the European Network of Excellence NANO- Press; 1957.
QUANTA (NM4-CT-2004-500198). The computing re- [26] De la Torre AG, Bruque S, Campo J, Aranda MGA. Cem Concr Res
sources of the Donostia International Physics Center 2002;32:1347.
(DIPC) and those of the Supercomputation Center of Gali- [27] Olivier WC, Pharr GM. J Mater Res 1992;7:613.
[28] Midgley CM. Acta Crystallogr 1952;5:307.
cia (CESGA) are gratefully acknowledged. [29] Henderson DM, Gutovsky HS. Bull Geol Soc Am 1956;67:1705.
[30] Laugesen JL. In: Bartos PJM, editor. Nanotechnology in construc-
References tion symposium proceedings. RSC Pub; 2004.
[31] Bonaccorsi E, Merlino S. J Am Ceram Soc 2005;88:505.
[1] Velez K, Maximilien S, Damidot D, Fantozzi G, Sorrentino F. Cem [32] Bonaccorsi E, Merlino S, Taylor HFW. Cem Concr Res
Concr Res 2001;31:555. 2004;34:1481.
[2] Acker P. In: Ulm F-J, Bazant ZP, Wittmann FH, editors. Creep, [33] Le Bellego C. Ph.D. dissertation, ENS de Cachan, France; 2001.
shrinkage and durability of concrete and other quasi-brittle materi- [34] Cho SW, Yang CC, Huang R. J Mar Sci Technol 2002;10:8.
als. Amsterdam: Elsevier; 2001. [35] Ulm F-J, Constantinides G, Heukamp FH. Conc Sci Eng 2004;37:43.
[3] Constantinides G, Ulm F-J. J Mech Phys Solids 2007;55:64. [36] Cong X, Kirkpatrick RJ. Adv Cem Based Mater 1996;3:144.
[4] Beaudoin JJ. Cem Concr Res 1983;13:319. [37] Brough AR, Dobson CM, Richardson IG, Groves GW. J Mater Sci
[5] Wittmann FH. Cem Concr Res 1986;16:971. 1994;29:3926.
[6] Monteiro PJJ, Chang CT. Cem Concr Res 1995;25:1605. [38] Ayuela A, Dolado JS, Campillo I, de Miguel YR, Erkizia E, Sánchez-
[7] Holuj F, Drozdowski M, Czajkowski M. Solid State Commun Portal E, et al. J Chem Phys 2007;127:164710.
1985;56:1019. [39] Taylor HFW. J Am Ceram Soc 1986;73:464.
[8] Made C, Jeanloz R. Geophys Res Lett 1990;17:1157. [40] Richardson IG. Cem Concr Res 2004;34:1733.
[9] Constantinides G, Ulm F-J. Cem Concr Res 2004;34:67. [41] Taylor HFW. Cement chemistry. 2nd ed. London: Thomas; 1997.
[10] Jennings HM. Cem Concr Res 2000;30:101. [42] Richardson IG. Cem Concr Res 1999;29:1131.
[11] Tennis PD, Jennings HM. Cem Concr Res 2000;30:855. [43] Kröner E. Statistical continuum mechanics. CISM Courses and
[12] Jennings HM, Thomas J, Gevrenov JS, Constantinides G, Ulm F-J. Lectures, 92, Udine; 1971.
Cem Concr Res 2007;37:329. [44] Mori T, Tanaka K. Acta Metall 1973;21:1605.

View publication stats

You might also like