You are on page 1of 7

Computational Modelling of Concrete Structures – Meschke, Pichler & Rots (Eds)

© 2018 Taylor & Francis Group, London, ISBN 978-1-138-74117-1

Fracture properties of cement hydrates determined from microbending


tests and multiscale modeling

J. Němeček, V. Šmilauer, J. Němeček & F. Kolařík


Faculty of Civil Engineering, Czech Technical University, Prague, Czech Republic

J. Maňák
Institute of Physics Academy of Sciences, Prague, Czech Republic

ABSTRACT: The paper shows experimental results received from bending tests performed on micro-
beams fabricated with focused ion beam milling. The micro-beams were bent by nanoindenter and load-
displacement curves recorded. From that, tensile strength and fracture energy were deduced for individual
cement paste constituents at the level of a few micrometers. The tensile strength of the C-S-H rich inner
product reached 700 MPa while the supremum of fracture energy was in the range 4–20 J/m2. The experi-
mentally obtained values served for verification of the multi-scale numerical model that was built for
scales starting from C-S-H globules (1–100 nm) to cement paste level (up to 1–100 μm). Simulation of
random packing of C-S-H globules reproduced well experimental data, yielding cohesive stress of the
globule as 2500 MPa. A weak size effect was found on the scale of C-S-H up to 1000 nm, signalizing
high ductility and weak strain localization. Further extension to cement paste was proposed, introducing
defects responsible for further strength reduction on scaling.

1 INTRODUCTION it is able to determine tensile strength of the individ-


ual phases and, in some cases, fracture energy needed
Many material characteristics like elastic modulus, to break the cantilever. The experimental part is a
tensile strength or fracture energy belong to basic stepping stone for the development for up-scaling
mechanical properties in fracture mechanics mod- and down-scaling models for fracture properties.
els. Traditionally, the properties are assessed on Thus, at second, we propose a three level numer-
centimeter to meter sized samples whose dimen- ical model for prediction of cement paste tensile
sions are far above the size of microstructural strength. In the down-scaling direction, identifi-
heterogeneities. Thus, microstructure-based pre- cation of cohesion of a C-S-H globule was done.
dictions that take microlevel heterogeneity into Smeared crack model with strain-softening was
account need experimentation at a substantially used in the framework of damage mechanics. In the
smaller scale, i.e. at the level of micrometers. More- up-scaling direction, the tensile strength of C-S-H
over, cementitious materials exhibit quasi-brittle products is described well in accordance with
behavior on all considered scales. In this paper, we experimental evidence gained on micro-beams. The
focus our attention to the basic microscopic level tensile strength of cement paste is further reduced
of cement paste which spans the range from 1 μm by internal defects whose size is beyond our experi-
to approximately 100 μm and which forms the ments as shown at the end of the paper.
binder of all cementitious systems.
First, we introduce a unique experimental method
2 EXPERIMENTAL PART
that allows cutting out a micrometer sized can-
tilever beam using the Focused Ion Beam (FIB)
2.1 Samples and material microstructure
milling technique from a heterogeneous composi-
tion of cement paste. High precision geometry of Samples of well hydrated cement paste were pre-
the beams is reached with FIB. The beam is loaded pared from pure Portland cement CEM-I 42,5R
in bending with the help of a nanoindenter and with a standard chemical composition given by
mechanical response is determined. Careful selection Table 1. Water to cement ratio used for prepar-
of homogeneous-like regions allows assigning results tion of the paste was 0,4. Samples were casted into
to dominant chemical phases, particularly the inner small cylindrical moulds and stored in water for
product and the outer product that are rich of C-S-H 7 years, thus the degree of hydration approaches
and Portlandite (CH) rich regions. From these tests, 100%. Then, samples were cut into 5 mm thick

113
Table 1. Chemical composition of cement.

CaO SiO2 Al2O3 Fe2O3 MgO TiO2

wt.% 63.77 20.51 4.74 3.3 1.05 0


K 2O Na2O SO3 MnO free Ca

wt.% 0.95 0.15 3.07 0.09 0

Figure 2. Microbeam fabricated in an outer product.

a current of 1 nA. Details on the fabrication can


be found in (Němeček, Šmilauer, Polívka, & Jäger
2016). After FIB milling, the micro-beams were
observed by SEM to scan their actual dimensions
and position in a particular microscopic phase. An
example of the milled beam is shown in Fig. 2.

Figure 1. Microstructure of cement paste. 2.3 Nanoindentation


Nanoindenter Hysitron Tribolab-700 system was
slices and prepared with a metallographic proce- used for characterization of mechanical response
dure (Němeček, Šmilauer, Polívka, & Jäger 2016) of micro-beams in two ways. First, standard
to get a flat and smooth surface with the roughness imprints were performed in the vicinity of the
of several tens of nm. Such sufrace is suitable for beams to reveal elastic properties of the tested
nanoindentation and FIB analyses. After polishing phase. Secondly, the indenter was used as a load-
samples were stored in an inert gas (argon) to limit ing tool to bend cantilevers and record the force-
carbonation besides testing times when samples displacement diagram in the displacement control-
were exposed to ambient lab humidity 20–40%. led regime. From the diagrams, both tensile strength
At the level of hydrates (i.e. below 100 μm) the sam- ft and fracture energy G fsup
u
have been calculated as
ples are characterized with a heterogeneous micro-
structure composing mainly of the inner and outer Fmax L h
products that are rich in C-S-H gels and amorphous ft = (1)
nano-crystallites of Ca(OH)2, then larger Portland- Iy 3
ite crystals and a small portion of unhydrous grains 1 wmax
G fsup = ∫
u
of clinker minerals, see Fig. 1. Also, the capillary Fdw
d (2)
porosity with the main pore radii in the level of 10−2 Af 0

−10−1 μm as measured by MIP (Němeček, Šmilauer,


Polívka, & Jäger 2016) is present at this scale. where Fmax is the maximum force measured by the
nanoindenter, L is the cantilever length, Iy the cross-
sectional second moment of inertia, h the height of
2.2 SEM and FIB
the triangular beam cross-section, wmax is the peak
After polishing samples were scanned and small deflection and Af is the fracture area (i.e. Af 12 bh
micro-beams prepared with the FIB technology. for the triangle). Note that the fracture energy was
FEI Quanta 3D FEG dual beam instrument com- computed as the supremum estimate based on the
bining SEM and FIB was used for fabrication assumption that the micro-beam behavior shows
of all samples. The FIB technique uses a finely neither snap-back nor softening and that the maxi-
focused beam of gallium ions for precise milling of mum force corresponds to the maximum energy
microscopic samples with various geometries. For release rate with a limiting stablecrack propagation.
our purposes, the sample geometry was chosen to
be a cantilever beam of 20 μm length and triangu-
2.4 Results of microbending tests
lar cross-section with approximate dimensions of
3–4 μm in height and width. The final milling step Results of the micro-bending tests as recorded by
was done at an accelerating voltage of 30 kV and nanoindenter and recalculated into non-dimensional

114
Chanvillard 2016). The basic assumption of the
model was that the tensile failure of C-S-H glob-
ules leads to progressive failure on higher scales.
At that time, downscaling approach identified
apparent uniaxial tensile strength of the C-S-HHD
as 107 MPa, which is almost 7 × lower than tensile
strength measured from microbending tests.
Recently, a molecular dynamics model of C-S-H
gel, formulated as cohesive polydisperse particles,
yielded tensile strength as C-S-HLD = 550 MPa and
C-S-HHD = 720 MPa (Davie & Masoero 2015).
Those values are comparable with microbend-
ing tests in Table 2, considering that outer prod-
uct contains certain portion of capillary porosity.
AFM nanoscale investigation of C-S-H cohe-
sion (modified C-S-HLD from alite hydration)
yielded 930 MPa, which is again comparable
with microbending results (Plassard, Lesniewska,
Pochard, & Nonat 2005).
In order to match current C-S-H experimental
data with the previous multiscale fracture model,
strength scaling needs further improvement. In this
regard, we propose three levels, see Figure 4, where
each level represents

Table 2. Results from nanoindetation and micro-


bending tests.

Outer product Inner product CH

E 24.9 ± 1.3 33.6 ± 2.0 39.0 ± 7.1


ft 264.1 ± 73.4 700.2 ± 198.5 655.1 ± 258.3
G fsup
u
4.4 ± 1.9 19.7 ± 3.8 19.9 ± 14.4
Figure 3. Results of micro-bending tests for individual
phases. n.t. 8 8 11

stress and relative deflection quantities are shown in Legend: E = Young’s modulus (GPa); ft = tensile strength
sup
u
(MPa); G f = supremum of fracture energy (J/m2),
Fig. 3. It can be seen in the figure that the beams
n.t. = number of evaluated tests.
behave approximately linearly up to the break and
the steep brittle fracture like unloading branch.
Althought the machine is depth controlled the sta-
bility of the post-peak control is not well maintained
and must be treated as approximate only. The tensile
strength is, therefore, captured well while the fracture
energy estimate was calculated as the supremum esti-
mate, Eq. 2. The results agree well with previously
obtained results of elastic properties of individual
cement paste constituents (Němeček, Králík, &
Vondřejc 2013) and also with energy calculations
derived from bulk nanoindentation (Němeček,
Hrbek, Polívka, & Jäger 2016).

3 MODELING PART

A four-scale fracture model was previously for-


mulated and validated for compressive strength of Figure 4. Three-level hierarchical multiscale model.
cement paste with w/c ∈ 0.157, 0.68 after approx- Microstructure images from E. Lachowski, P. Stutzman,
imately 1 day of hydration (Hlobil, Šmilauer, & D.P. Bentz et al.

115
• Level 1: CSH. C-S-H globules are intermixed It is assumed that a material contains randomly
with gel pores oriented 2D elliptical flat voids with various aspect
• Level 2: Cement paste. Level 1 acts with capil- ratios m = b/a. The voids have a negligible area and
lary porosity, other hydration products, unre- only represent stress concentrators and internal
acted clinker, unreacted SCMs. defects in a material. Under macroscopic biaxial
• Level 3: Defects. Level 2 is enriched with defects stress, the maximum tensile stress among all voids,
in the form of cracks and air voids. m ⋅ ση, appears on a critically inclined elliptical
void under a critical angle ψ
3.1 Material model for compressive and σ 3 σ1 σ1 1
tensile failure cos 2ψ = , ≥− (4)
2( 3 1 ) σ 3 3
Material model describing compressive of tensile
failure at each level is based on fracture/damage −(σ 1 − σ 3 )2
m ⋅ση = (5)
mechanics. Damage mechanics uses the concept 4(σ 1 + σ 3 )
of an equivalent strain, ε, as a descriptor of dam-
age evolution. Damage becomes initiated when the Crack formation occurs when the tangential
equivalent strain, ε, exceeds strain at the onset of tensile stress, m ⋅ ση, equals to the tensile strength
cracking, ε0 = ft/E, where E is the elastic modulus. of the matrix. Since ση and the crack geometry,
The Rankine criterion for tensile failure defines ε as m, cannot be measured directly, it is reasonable to
relate their product to the uniaxial macroscopic
σ1 tensile stress, σ 1, as proposed by Griffith (Griffith
ε = , σ1 > 0 (3)
E 1924).

where σ1 is the maximum positive effective prin- m ⋅ση


cipal stress of undamaged-like material. Figure 5 σ1 = (6)
elucidates the mechanism. 2
Under uniaxial compressive stress, crack initia-
tion occurs under a different mechanism. A homo- The material starts to crack when σ 1 equals to
geneous material experiences only one negative the uniaxial macroscopic tensile strength ft. Note
principal stress and deviatoric stresses. Cracking that the tensile strength of the homogeneous
in diagonal, shear band zone, is often encountered matrix (intrinsic strength) and the crack geometry
on cementitious specimens, however, the physi- remain unknown separately and we can assess only
cal mechanism is again tensile microcracking in apparent macroscopic tensile strength ft.
voids and defects in the underlying microstructure Plugging Equation (5) into Equation (6) and
(Bažant & Planas 1998, pp. 297). Such a behavior further into Equation (3) leads to the definition
has already been described in the work of Grif- of the equivalent strain, ε , under compression-
fith (Griffith 1924), and McClintock and Walsh dominant loading
(McClintock & Walsh 1962), and we briefly review
this theory and extend it with an equivalent strain 1 −(σ 1 σ 3 )2
to be used in the framework of damage mechanics. ε = ⋅ (7)
E 8(σ 1 + σ 3 )

An interesting feature of the Griffith model is


that the ratio of the uniaxial compressive-to-tensile
strength equals to 8

| fc | = ft (8)

This can be verified when plugging σ1 = ft into


Equation (3) which leads to ε = ft /E . Plugging
σ1 = 0, σ3 = −8ft into Equation (7) leads to the same
equivalent strain.
Since the equivalent strain may arise from the
Rankine of Griffith criterion, it is necessary to
compare both Equations (3) and (7) and to select
Figure 5. Crack evolution during (a) uniaxial tensile the higher equivalent strain. Since the damage
stress and (b) compressive stresses. The material contains evolution law has a small effect on the computed
randomly oriented elliptical voids with negligible area. macroscopic strength simple linear softening is

116
assumed in the simulations. The linear cohesive
law takes the form

⎛ w⎞
σ = ft ⎜1 − ⎟ (9)
⎝ wf ⎠

where w is a crack opening and wf is the maximum


crack opening at zero stress. According to the
formulation of the isotropic damage model, the
uniaxial tensile stress obeys the law

σ ( ω )E ε. (10)

with ω being the isotropic damage parameter,


ω ∈ 0;1 . Figure 6. Unit cells for C-S-HLD and C-S-HHD, 20 × 20 ×
Let us consider the fracture energy in mode I, 40 brick elements (size 100 × 100 × 200 nm).
Gf, and the effective thickness of a crack band h,
which corresponds to the finite element size in
the direction of the maximum principal strain
(Bažant & Planas 1998). This ensures objec-
tive results, independent on finite element size
(Jirásek & Bažant 2002). Notice that w hωε
and wf = 2Gf/ft. The evolution of isotropic dam-
age is obtained by combining Equations (9) and
(10).
−1
⎛ ε ⎞ ⎛ hE ε 02 ⎞
ω = 1 − 0 ⎟ ⎜1 − ⎟ (11)
⎝ ε ⎠ ⎝ 2G f ⎠

3.2 Model for fracture properties of C-S-H


globules Figure 7. Tensile strength scaling in C-S-H gel with
various packing density of globules.
Nanoindentation experiments carried out previ-
ously (Constantinides & Ulm 2004) identified
elastic properties of C-S-H and limit packing den-
sities as ηC-S-HLD = 0.63 and ηC-S-HHD = 0.76 (Con-
stantinides & Ulm 2007). Our model used 2D
and 3D images for different packing densities.
Figure 6 provides such examples for C-S-HLD and
C-S-HHD.
In order to match uniaxial tensile test, periodic
boundary conditions occur on horizontal surfaces
and uniform static boundary conditions with zero
normal stress belong to vertical surfaces. Vertical
eigenstrain loads 2D unit cell and uniaxial macro-
scopic tensile stress recovers.
Figure 7 displays the evolution of the C-S-H
tensile strength with regards to packing density,
as determined from 200 × 200 nm microstructures.
Fracture energy 2 J/m2 was assigned to the globule. Figure 8. Tensile strength scaling on C-S-HLD gel
Analytical approximation reads. depending on size of microstructure.

⎛ 1.293( glob
g
13.011
− 1) ⎞ where instrinsic cohesion of a globule was found
ft ,0 ft ,glob exp ⎜ ⎟ (12) as ft,glog = 2500 MPa. This yields tensile strength
⎝ ηglob ⎠ ft ,0C-S-HLD = 247.6 MPa and ft ,0C-S-HHD = 489.3 MPa,

117
being comparable with molecular dynamics data simulations need to prove this hypothesis and vali-
and microbending tests. date experimental data.
Further, we explored strength scaling with the
size of a unit cell. In perfectly plastic materials
and parallel configuration of load-bearing phases, 4 CONCLUSIONS
no size effect occurs as opposed to brittle materi-
als or to serial configuration. Weak size effect was The following conclusions can be drawn on experi-
found on C-S-HLD up to 1000 nm signalizing that mental and numerical parts:
C-S-H globules are ductile enough to redistribute
• Micro-bending tests on beams prepared with
stress during strain softening (Figure 8). Weak sta-
FIB give access to local fracture properties of
tistical size effect occurs after 1000 nm signalizing
cement paste constituents. Tensile strength
random nature of C-S-H packing with increasing
of inner product, outer product and CH were
brittleness.
found in the range of 260–700 MPa. Supremum
of fracture energies lie in the range 4–20 J/m2 for
3.3 Model for level of cement paste the respective constituents.
• Cohesion of C-S-H globule was identified as
Ordinary cement paste with low w/c attains flexu-
2500 MPa. When intermixing C-S-H globules
ral strength up to approximately 12 MPa (Taplin
with gel porosity, experimentally measured
1959). This can be significantly increased by hot-
strengths are approximately recovered for both
cured, compacted cement pastes with amorphous
C-S-HHD and C-S-HLD. Also, the results from
C-S-H where compressive strength 690 MPa
molecular dynamics are consistent.
was reached (i.e. the tensile strength of about
• Preliminary findings show that microcracks or
690/8 = 86 MPa can be assumed) (Gouda & Roy
internal defects as long as 100 μm exist and they
1976). It was shown previously that significant
control strength of cement pastes. The origin of
decrease of strength from C-S-H to paste level
those defects, their formation and description
is caused by presence of capillary porosity, spa-
need further research.
tial gradient of C-S-H, and stress concentra-
tion around elastic inclusions (unreacted cement
clinker, crystalline hydration products, unreacted
ACKNOWLEDGEMENT
supplementary cementitious materials) (Hlobil,
Šmilauer, & Chanvillard 2016). However, those
Financial support of the Czech Science Founda-
effects are still insufficient for strength scaling
tion’s project 17-18652S is gratefully acknowledged.
from C-S-H to cement paste. Thus, further reduc-
ing mechanisms happen. They can be in the form
of interfaces (Qian, Schlangen, Ye, & van Breugel
2012) or additional defects introduced at level 3. REFERENCES
Defects at level 3 can arise from several causes;
shrinkage during water consumption, rearange- Bažant, Z.P. & J. Planas (1998). Fracture and Size Effect in
ment of hydrates, internal stresses etc. It is obvi- Concrete and Other Quasibrittle Materials. CRC Press.
ous that size of those defects must lay above the Constantinides, G. & F.-J. Ulm (2004). The effect of
two types of C-S-H on the elasticity of cement-based
size of our experimental cantilever beams, i.e. materials: Results from nanoindentation and micro-
above ≈5 μm. LEFM can estimate the bottom mechanical modeling. Cem. Concr. Res. 34(1), 67–80.
size of internal defects if considered as a crack of Constantinides, G. & F.-J. Ulm (2007). The nanogranular
length a in perfectly brittle material. For illustra- nature of C-S-H. Journal of the Mechanics and Physics
tion, let us consider cement paste with E = 15 GPa, of Solids 55, 64–90.
Gf,CSH = 2 J/m2, ft = 10 MPa Davie, C. & E. Masoero (2015). Modelling damage from
the nano-scale up. In 10th International Conference on
Mechanics and Physics of Creep, Shrinkage, and Dura-
EG
G f CSH bility of Concrete and Concrete Structuresg.
ft = , (13)
πa Ghebrab, T. & P. Soroushian (2010, June). Mechanical
Properties of Hydrated Cement Paste: Development
EG
G f CSH 15 ⋅ 109 ⋅ 2 of Structureproperty Relationships. International Jour-
a= ≈ = 95 ⋅ 10 −6 m. (14)
π ft 2
π (10 ⋅ 106 )2 nal of Concrete Structures and Materials 4(1), 3743.
Gouda, G.R. & D.M. Roy (1976). Characterization of
hotpressed cement pastes. Journal of the American
Such estimation is consistent with (Ghebrab & Ceramic Society 59(9–10), 412–414.
Soroushian 2010) who used a ≥ 50 μm arguing that Griffith, A. (1924). Theory of rupture. In C. Biezeno
large portlandite crystals are responsible for stress and J. Burgers (Eds.), First International Congress for
concentration and crack propagation. However, Applied Mechanics, Delft, pp. 55–63.

118
Hlobil, M., V. Šmilauer, & G. Chanvillard (2016). Micro- Němeček, J., V. Šmilauer, L. Polívka, & A. Jäger (2016).
mechanical multiscale fracture model for compressive Tensile strength of hydrated cement paste phases
strength of blended cement pastes. Cement and Con- assessed by microbending tests and nanoindentation.
crete Research 83, 188–202. Cement and Concrete Composites 73, 164–173.
Jirásek, M. & Z.P. Bažant (2002). Inelastic analysis of Plassard, C., E. Lesniewska, I. Pochard, & A. Nonat
structures. John Wiley & Sons, Ltd. (2005). Nanoscale experimental investigation of parti-
McClintock, F. & J. Walsh (1962). Friction of griffith cle interactions at the origin of the cohesion of cement.
cracks in rock under pressure. In Proc. Fourth U.S. Langmuir 21(16), 7263–7270. PMID: 16042451.
National Congress of Applied Mechanics, New York Qian, Z., E. Schlangen, G. Ye, & K. van Breugel (2012).
City, pp. 1015–1021. Multiscale lattice fracture model for cement-based
Němeček, J., V. Hrbek, L. Polívka, & A. Jäger (2016). materials. In 4th International Conference on Com-
Combined investigation of low-scale fracture in putational Methods (ICCM 2012), Gold Coast,
hydrated cement assessed by nanoindentation and Australia.
fib. In V. Saouma, J. Bolander, and E. Landis (Eds.), Taplin, J.H. (1959). A method for following the hydration
9th International Conference on Fracture Mechanics of reaction in Portland cement paste. Australian J. Appl.
Concrete and Concrete Structures, FraMCoS-9. Sci. 10, 329–345.
Němeček, J., V. Králík, & J. Vondřejc (2013). Microme-
chanical analysis of heterogeneous structural materi-
als. Cement and Concrete Composites 36, 85–92.

119

You might also like