Determination of representative crack density of cementitious materials

H. H. Pan, Y. W. Chen & D. H. Lin

Dept. of Civil Engineering, Kaohsiung University of Applied Sciences, Kaohsiung, Taiwan

ABSTRACT: Based on the micromechanics approach and SEM measurements, a representative crack density
parameter is chosen to evaluate elastic moduli and the stress intensity factor of high-performance concrete.
Following the predicted material properties containing no cracks, one can estimate the effective material
properties if their crack densities are given. SEM specimens were taken in the middle part of high-
performance concrete, and each sample was observed at the fixed and random position to measure the micro-
cracks with five magnifications from 500× to 5000×. The micromechanics-based calculations were compared
with the experimental data to verify the reliability of representative crack density parameter. It is found that
the optimum SEM magnifications for the representative crack density parameter of cementitious materials are
of 3000× ~ 5000×, especially the magnification near 3000× is suitable for HPC containing c1 = 0.67 .

1 INTRODUCTION where superplasticizer conforming to ASTM C494

Type-G with a specific gravity of 1.1. The total vol-
The crack density calculation of cementitous materi- ume fraction of the aggregates is c1 = 0.67 with
als depends on the number, the length, the width, the river sand having a specific gravity of 2.60 and the
arrangement and the distribution of the cracks inside absorption of 2.5 %, and coarse aggregate is a kind
the material, and the observation methods. For a of crushed sandstone with a specific gravity of 2.57
scanning electron microscope (SEM), for example, and the absorption of 1.45 %. A mixture proportion
different observation magnifications of the specimen of high-performance concrete (HPC) with w/b=0.36
will result in different measurements of the cracks. is shown in Table 1, and the slump is 200 ± 30 mm.
Although many theoretical and experimental meth-
ods have been proposed to estimate and measure the Table 1. Mixture proportion of high-performance concrete*.
crack density for brittle solids (Attiogbe & Darwin Water Cement Fly ash Sand Gravel SP
1986, Attiogbe & Darwin 1987, Erick 1988, Oilliv- 160 378 67 730 1020 2.23
ier 1985), it is still difficult to know the true crack *Unit: (kg/m3)
density of cement-matrix composites up to now.
In this paper, one tries to find a representative Material age is of 28 days, and specimen sizes of
crack density parameter (RCDP) of cementitous ma- concrete made by steel molds are of 100φ × 200 mm
terials depending on the SEM magnifications, and and 100 × 100 × 350 mm, respectively, to measure
this RCDP will be examined by the micromechanics the elastic moduli and the fracture toughness. At
theory and the experiments to confirm the reliability least six specimens were used to examine the mate-
in use. Two SEM observations were chosen to view rial properties. Specimens were under a uniaxial
the microcracks: fixed position and random position, compression by MTS machine with a constant strain
where the fixed view means that the inspection posi- rate ε& = 1 × 10−5 /sec to measure the longitudinal and
tion of the cracks always locates in the middle of the lateral strains and plot the stress-strain curves. Frac-
sample, and the random view is the cracks met by ture toughness was calculated from the three-point
chance without any favorite positions. bending test.
The crack density of high-performance concrete
was determined based on the SEM measurements.
2 EXPERIMENTAL PROGRAM The size of SEM specimens is about 3 × 3 × 1.5 mm.
The length and the number of microcracks were
The binder consists of cement, fly ash and superplas- measured from SEM specimens when the material
ticizer (SP), and water-to-binder ratio (w/b) is 0.36, was under no load, 0.3 f c' and 0.5 f c' respectively
in the uniaxial compression test, and was under the The definition of crack density parameter, ac-
fracture strength for the calculations of the fracture cording to Budiansky & O’Connell (1976), is
toughness, where f c' is the peak strength of con-
crete. The observation positions of SEM specimens 2N A2
η= (1)
were at the fixed position and the random one, re- π P
spectively. Each observation point was viewed by
five magnifications: 500×, 1000×, 3000×, 4000× and where η = crack density, N = total number of cracks
5000×. per unit volume, A = area of the crack, P = perimeter
To evaluate the crack properties, one used Photo- of the crack, and the angle brackets < ⋅ > = volume
shop7.0 software to deal with the SEM picture con- averaging of the quantity.
verted into the monotonic white and black color, and To determine the crack density, assuming that the
SigmaScan Pro5 software to measure the number SEM specimens can suitably represent the realistic
and the length of cracks. Besides, the window size of cracks inside the material, and all cracks are convex
the observation in SEM was also measured. Figure 1 and have the same size. Then, a theoretical calcula-
is the SEM picture at a magnification of 4000 × tion of the crack density measured from two-
while HPC was under the load 0.3 f c' , and Figure 2 dimensional cracks is used as (Budiansky &
is an image transformation of Figure 1 by Photo- O’Connell 1976)
shop7.0 software. 8
η=
2
M⋅ l (2)
π 3

where l = average trajectory of the cracks and M =

total crack number per unit area in SEM window.
Let n, h and w be referred to the crack number, and
the height and the width of window, respectively.
The total crack number per unit area then is calcu-
lated by
n
M= (3)
h× w
Here, the computation of the crack number n in
' SEM winder is following the rule that the crack at an
Figure 1. SEM-picture of cracks with 4000× at 0.3 f c . obvious turning point is treated as the beginning of a
new crack. For example, number 1 crack marked in
Figure 2 consists of three cracks in the calculation,
where two straight cracks and a bended crack were
counted approximately.

3.2 Effective elastic moduli and fracture toughness

High-performance concrete is assumed to be a two-
phase composite containing concrete without cracks
as the matrix and the cracks as the inclusion. At-
tiogbe (1987) proposed an analytical procedure used
to convert two-dimensional crack data into three-
Figure 2. Image transformation of Figure 1 by Photoshop7.0 dimensional crack distributions in cement paste and
software. mortar, and found that the degree of anisotropy K is
about -0.15 when the compressive strain is less
0.002. Thereby, the cracked concrete (composite) as
3 THEORETICAL ANALYSIS a whole is isotropic if the strain is small.
Pan & Weng (1995) used the inclusion theory
3.1 Crack density
(Mori & Tanaka 1973, Weng 1984) to examine the
Because SEM can only scan a small area of the effective elastic bulk modulus κ and effective
specimen at a time to view the microcracks, it is dif- elastic shear modulus μ of the composite, and con-
ficult to find out the true size and the shape of mi- cluded that the elastic moduli of isotropic cracked-
crocracks and their distributions in the material. That materials are of less crack-shape sensitivity. Besides,
is why one needs to establish some rules to
the effective elastic moduli of the material with cir-
straighten out the meaning of crack density in use,
and those rules have to be confirmed correctly. cular cracks were also found by
κ 1 45(2 −ν 0 )[45(2 −ν 0 ) +16(1 −ν 02 )(10 − 3ν 0 )] (10)
= (4) g=
κ0 16 1 −ν 0
2 45(2 −ν 0 )2[45+ 32(5 +ν 0 )η] +1024(1 −ν 02 )(5 −ν 0 )(5 − 2ν 0 )η 2
1+ η
9 1 − 2ν 0 where k1 = main crack contour factor, and the value
k1 = 0.072 for the steady-state propagating crack
μ 1 and k1 = 1 / 24 for the stationary crack respectively.
= (5)
μ 0 1 + 32 (1 − ν 0 )(5 −ν 0 )
Now, in our case, HPC containing microcracks is
η isotropic. One can use a uniaxial compression and
45 2 −ν 0 SEM to find the elastic Young modulus, Poisson’s
where κ 0 = elastic bulk modulus of the matrix (no ratio and the crack density η of HPC. Of course,
cracks), μ 0 = elastic shear modulus of the matrix, the other two elastic moduli κ and μ are also de-
and ν 0 = Poisson’s ratio of the matrix. Due to the termined from the isotropic relation of HPC.
less crack-shape sensitivity, one can use Equation 4 Once the crack density η and Poisson’s ratio ν
and Equation 5 to determine the effective elastic of the cracked body are known, the Poisson ratio of
moduli of isotropic HPC containing arbitrary shapes HPC without cracks (ν 0 ), calculated from Equation
of cracks. The elastic relation still holds for 6, is determined. This Poisson’s ratio ν 0 allows us
E = 9κμ /(3κ + μ ) . to obtain the elastic bulk modulus κ 0 and shear
From Equation 4, Equation 5 and the Hooke’s modulus μ 0 respectively by substituting ν 0 and
law, one can easily find out the relation of Poisson’s η into Equation 4 and 5, so as to find Young’s
ratio between the composite (cracked material) and modulus E0 simultaneously. Thereby, the same
the matrix as material subjected to new compressive loads will
produce new crack densities, the predicted effective
45(2 −ν 0 )ν 0 + 16ν 0 (1 −ν 02 )η bulk and shear moduli, κ and μ , are found by
ν=
45(2 −ν 0 ) + 16(1 −ν 02 )(10 − 3ν 0 )η
(6) Equation 4 and 5.
Similarly, from Equation 8, one can calculate the
where ν = Poisson’s ratio of the cracked material. stress intensity factor of the matrix K 0 if the crack
Meanwhile, the fracture toughness of the brittle density is given. As the material properties of HPC
material is usually expressed by the critical stress in- without microcracks are found theoretically, the
tensity factor K c . Let the stress intensity factor of stress intensity factor increment ΔK due to the mi-
the matrix, the crack-tip stress intensity factor of the crocracking is finally determined by means of Equa-
composite and the stress intensity factor change tion 7.
(toughness chance) be denoted as K0, K tip and
ΔK , respectively, where the crack-tip stress inten-
sity factor K tip = K 0 − ΔK and the critical stress in- 4 RESULTS AND DISCUSSION
tensity factor of the composite K c = K 0 + ΔK .
Based on a micromechanics approach, the ana- 4.1 Representative crack density parameter
lytic solution of toughness change for a two-phase As one knows, different observation magnifications
isotropic composite under Mode I loading has been in SEM measurements will lead to different values
derived (Pan 1999) and the form is of crack density in estimations. In this paper, one
tries to suggest a representative crack density pa-
K tip
= f g (7) rameter that can suitably employ to estimate the me-
K0 chanical properties of cracked cementitious materi-
als.
where f and g are material parameters. It is noted High-performance concrete with c1 = 0.67 was
that, in Equation 7, the ratio K tip / K 0 less than one tested by the uniaxial compression and had the peak
implies material toughening. From Equation 7 and stress f c' = 48.56 MPa. SEM with five magnifica-
the relations of K tip , K 0 , ΔK and K c , the stress tions from 500× to 5000× and the field at fixed and
intensity factor of the material with no cracks has random position were taken to view the cracks after
the form as the designed loading reached, and the results are
Kc shown in Figure 3. The estimated crack densities of
K0 = (8) high-performance concrete at the random position of
2− f g the view are always greater than those at the fixed
If the cracked material contains circular cracks, the position regardless the magnifications, and both val-
material parameters f and g are ues approach to asymptotes as the magnifications in-
crease.
27 + 96k1 (1 + ν 0 ) 2η Figure 4 shows the crack densities of concrete in
f = (9) the random view applied to no load, 0.3 f c' , 0.5 f c'
27 + 4(1 + ν 0 ) 2η
and f c' , respectively. Those crack densities also
tend to some asymptotic constants when the magni-
 '
fications are greater than 3000×. From Figure 3 and Table 2. Elastic moduli under 0.52 f c .
Figure 4, it seems that the estimated crack density E（GPa） ν μ（GPa） κ（GPa）
may be insensitive to the magnifications and the ob- 21.39 0.244 8.60 13.92
servation positions if the magnifications are of 3000
× or larger, and this range of crack density might be From Table 2 and Equations 4-10, the calculated
chosen as the representative crack density parameter results for the shear modulus, bulk modulus and
which we can use to evaluate the material properties. fracture toughness of the matrix (no cracks exist) are
However, it is still needed to inspect carefully. shown in Table 3, and their statistical variances are
also shown in Table 4. In Table 3, the differences of
2.5 average material properties at the magnification
fixed position
range of 500 × ~ 5000 × are pretty large in both
2.0 random position fixed view and random view. In Table 4, the vari-
ances of elastic moduli and fracture toughness at the
crack density

1.5 magnification range of 3000× ~ 5000× are far less

than those of 500× ~ 5000×. Therefore, the crack
1.0 density at the magnification range of 3000× ~ 5000×
is suitable for selecting as the representative crack
0.5 density parameter in concrete.

0.0 Table 3. Average properties of the matrix.

0 1000 2000 3000 4000 5000 6000 Observa- Magnification
Properties
magnification tion 500x~5000x 3000x~5000x
Figure 3. Crack density of different views after failure. Fixed 14.61 10.07
μ0 (GPa)
2.5
Random 18.79 10.33
no load Fixed 40.97 20.56
2.0
0.3fc' κ 0 (GPa)
0.5fc' Random 61.69 22.06
fc'
crack density

1.5 K0 Fixed 0.615 0.774

( MPa m ) Random 0.581 0.695

1.0

Table 4. Variance of average properties of the matrix.

0.5 Observa- Magnification
Properties
tion 500x~5000x 3000x~5000x
0.0
Fixed 661.02 3.67
0 1000 2000 3000 4000 5000 6000
μ0 (GPa)
magnification Random 3575.78 15.01
Figure 4. Crack density of random position with different
Fixed 0.048 0.017
loads. κ 0 (GPa)
Random 0.028 0.014
Although different SEM magnifications will lead K0 Fixed 661.02 3.67
to different crack densities in calculations, the prop-
( MPa m ) Random 3575.78 15.01
erties of cracked material are unique in experiments.
To examine the effect of the magnification, average
crack densities calculated from two groups of 500× 4.2 Theoretical verification
~ 5000× and 3000× ~ 5000× were chosen to esti-
mate the elastic moduli of the material without Now the representative crack density parameter η
cracks. is selected at the magnification range of 3000× ~
The elastic moduli of high-performance concrete 5000× in use. This representative crack density pa-
with c1 = 0.67 subjected to a uniaxial compression rameter (RCDP) is considered as an important factor
0.52 f c' are shown in Table 2. Those experimental to evaluate the mechanical properties of cracked ma-
data allow us to theoretically find the elastic moduli terial. Here, the elastic moduli and fracture tough-
of material containing no cracks (matrix). For exam- ness of the matrix in Table 3 at the range of 3000× ~
ple, the experimental crack densities at the magnifi- 5000× are taken to estimate the effective bulk and
cation range of 3000× ~ 5000× in fixed view and shear moduli, and fracture toughness of high-
random view are η = 0.123 and η = 0.146 , re- performance concrete containing the aggregate
spectively, and then from Equation 6 the Poisson ra- c1 = 0.67 .
tio of concrete without cracks ν 0 is found to be Based on the material properties of the matrix in
0.289 and 0.297 in turns. Table 3 and the representative crack density parame-
ters measured from HPC subjected to no load, 0.3 f c' Table 7. Comparisons of effective Young modulus (GPa).
and 0.5 f c' respectively, the experimental values and Observation
Magni- No
0.3 f c
'
0.5 f c
'
theoretical calculations for the effective bulk and fication load
shear moduli are shown in Table 5. The predicted ef- Experiment 21.55 19.25 18.33
fective bulk and shear modulus at fixed view are 500x 5.87 6.18 8.243
close to those at random view regardless of the ap- 1000x 10.49 11.32 10.63
plied stress. It means that one can use the SEM ob-
Fixed 3000x 22.81 21.09 20.13
servation either at fixed view or at random view to
measure the crack properties if the magnification is 4000x 23.62 23.46 21.97
of 3000× ~ 5000×. 5000x 23.89 23.54 22.70
500x 5.13 12.37 4.45
Table 5. Comparisons of effective elastic moduli. 1000x 9.44 17.21 12.73
' '
Properties Observation No load 0.3 f c 0.5 f c
Random 3000x 21.22 20.07 19.41
4000x 22.75 22.36 22.01
Experiment 8.66 7.74 7.37 5000x 24.96 24.18 23.99
μ (GPa) Fixed 9.27 9.01 8.65
25
Random 9.09 8.82 8.68

Experiment 14.03 12.53 11.93 20

κ
stress(MPa)
(GPa) Fixed 16.59 15.48 14.13 15

Random 15.81 14.76 14.22 10

experimental
linear elastic
500X
Fixed 0.062 0.085 0.118 1000X
η 5 c1=0.67 3000X
fixed position 4000X
Random 0.099 0.124 0.138 5000X
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

Compared with the experimental data in Table 5, strain(*10-3)

the predicted effective elastic bulk and shear moduli Figure 5. Stress-strain curves at fixed view.
have the errors from 5% to 18% approximately. Ta-
ble 6 shows the fracture toughness for experimental 25

data and the predictions. The predicted fracture

20
toughness is in an acceptable range as compared
with the experimental data in Table 6.
stress(MPa)

15

Table 6. Comparisons of fracture toughness. experimental

10 linear elastic
500X
Properties Experiment Fixed Random 1000X
5 c1=0.67 3000X
4000X
K c ( MPa m ) 0.512 0.550 0.542 random position 5000X
0
η --- 0.238 0.149 0 1 2 3 4 5 6
strain(*10-3)
Figure 6. Stress-strain curves at random view.
Finally, the effective Young modulus of high-
performance concrete subjected to different loads is 25
calculated with five magnifications, and the results c1=0.67
are shown in Table 7 and Figures 5-6. By comparing 20
fixed position

with the experimental data, the predicted results cal-

culated from the crack density at 500× and 1000×
stress(MPa)

15
are not acceptable shown in Figures 5-6. Let the
stress-strain relations in Figures 5-6 be enlarged near 10
the experimental curves and re-plotted in Figures 7- experimental
linear elastic
8, obviously, the predicted effective Young modulus 5 3000X
4000X
calculated from the magnification of 3000× is close 5000X
to the experimental data. Hence, the better choice for 0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
representative crack density parameter is the cracks
strain(*10-3)
measured at the magnification near 3000×.
Figure 7. Enlarged stress-strain curves at fixed view.
 ACKNOWLEDGMENT
25
c1=0.67 The funding of this research was partially supported
random position
20 by the Taiwan National Science Council under
Grant NSC 95-2221-E-151-046.
stress(MPa)

15

10
experimental
REFERENCES
linear elastic
3000X
5
4000X
Attiogbe, E. K. & Darwin, D. 1986. Correction of window size
5000X distortion of crack distributions on plane sections. J. of Mi-
0 croscopy 114(1): 71-82.
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 Attiogbe, E. K. & Darwin, D. 1987. Submicrocracking in ce-
-3
strain(*10 ) ment paste and mortar. ACI Journal 84-M43: 491-500.
Figure 8. Enlarged stress-strain curves at random view. Budiansky, B. & O’Connell, R. J. 1976. Elastic moduli of a
crack solid. Int. J. Solids Structures 12: 81-97.
Erick, R. 1988. Automatic quantification of microcracks net-
work by stereological method of total projections in mor-
5 CONCLUDING REMARKS tars and concretes. C.C.R. 18: 35-43.
Mori, T. & Tanaka, K. 1973. Average stress in the matrix and
Based on SEM measurements and the verification of average elastic energy of materials with misfitting inclu-
the micromechanics approach, the estimated crack sions. Acta Metall. 21: 571-574.
Oillivier, J. P. 1985. A non destructive procedure to observe
density at the magnification range of 3000× ~ 5000× the microcracks of concrete by scanning electron micros-
can be considered as the representative crack density copy. C.C.R. 15(3): 1055-1060.
parameter in concrete or cementitious materials. Pan H. H. & Weng, G. J. 1995. Elastic moduli of heterogene-
This representative crack density parameter allows ous solids with ellipsoidal inclusions and elliptic cracks.
us to determine the mechanical properties of cracked Acta Mechanica 110: 73-94.
Pan H. H. 1999. An overall approach for microcrack and in-
cementitious material if the microcracks are ran- homogeneity toughening in brittle solids. Chinese J. Me-
domly oriented. For high-performance concrete with chanics 15(2): 57-68.
the volume fraction of the aggregates c1 = 0.67 , the Weng, G. J. 1984. Some elastic properties of reinforced solids,
representative crack density is better observed with special reference to isotropic ones containing spherical
around the magnification of 3000×. inclusion. Int. J. Engng. Sci. 22: 845-856.