Soil Dynamics and Earthquake Engineering 110 (2018) 185–194

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

journal homepage: www.elsevier.com/locate/soildyn

Stiffness of a recycled composite aggregate T

a b,⁎ c
H. He , K. Senetakis , M.R. Coop
a
University of New South Wales (UNSW), Sydney, Australia
b
City University of Hong Kong, Hong Kong
c
University College London, London, UK

A B S T R A C T

Recycled concrete aggregate (RCA) is a promising substitute for natural aggregates and the reuse of this
demolished material would benefit the construction projects both economically and environmentally. A diffi-
culty associated with the prediction of the behavior of RCA is because of its multi-composition which is linked to
the grain size in consideration. In this study, a comprehensive laboratory testing program is conducted on
different fractions of RCA for which the ratio of silicon over calcium increases for the coarser fractions due to the
limitation of the cement mortar inclusion as the mean size increases. The study incorporates bender/extender
element tests capturing small-strain constrained moduli, Young’s moduli and Poisson’s ratio of RCA fractions
with different mean grain size and grain size distribution. For the uniform RCA fractions, the results showed that
during isotropic compression, the specimens had a clear dependency of their dynamic properties on mean grain
size. For the better graded specimens, it was shown that their behavior was dominated to some extend by the
finer fraction they were composed rather than being affected equally by the different fractions they consisted of.
The results also highlighted the sensitivity of the behavior of the samples to the over-consolidation stress history.
For RCA, a multi-composition material, the grain size has a dominant role on the modulus – pressure relationship
as well as the sensitivity of material behavior to stress history and these observations must be considered in the
analysis of geo-structures where RCA is used as construction material.

1. Introduction element test (BE) [38], which is a high-frequency dynamic testing

technique that uses a pair of piezo-element inserts, is widely adopted to
The amount of concrete aggregate recycled from demolished capture the shear wave velocity (Vs) for the purpose of deriving the
buildings or structures has risen rapidly in recent years (e.g. [29]). small-strain soil shear modulus (Gmax) (e.g. [41,18,22,45,23,1], Airey
Facing the increasing need of aggregates in engineering construction and Mohsin, [10]). The same pair of bender elements can also be
and the rising cost of natural aggregates, recycled concrete aggregate configured to send primary waves (P-waves) and measure their pro-
(RCA) has become a promising and economic substitution. Benefitting pagation velocity (Vp), denoted as extender element mode (EE) (used by
from extensive research on RCA, the material is starting to be used as a [24,23,19,20,13], among others). The small-strain constrained modulus
non-structural construction element in practical projects, including (Mmax), Young’s modulus (Emax) and Poisson’s ratio (ν) can be thereafter
geotechnical engineering and pavement geotechnics applications (e.g. derived from the knowledge of Vs, Vp and the specimen density, thus
[31,32,11,29,30,40,13,14]). This replacement of the natural aggregates the complete characterization of the small-strain behavior of geo-ma-
releases, partly, the pressure from both the demand of quarry sites and terials can be implemented based on measurements from bender/ex-
the cost of land-fill disposal. tender element tests.
In modeling the behavior of RCA when used as a geotechnical ma- In the light of examining in this study sand-sized recycled concrete
terial, stiffness at small strains is a key property. Material small-strain aggregate, particular focus has been paid on the associated literature on
stiffness is a critical parameter for the prediction of the deformations of granular materials. Based on published works on the dynamic proper-
geo-structures and soil-structure interaction problems. The stiffness of ties of sand-sized materials, it can be concluded that the major prop-
geo-materials, including the shear modulus (G), constrained modulus erties that affect their dynamic behavior at small strains at a given
(M) and Young’s modulus (E), reaches its maximum value in the elastic isotropic confining pressure are the void ratio, the coefficient of uni-
range of behavior (generally less than 10−3% strain). The bender formity, the particle shape and overall morphology as well as the

⁎
Corresponding author.
E-mail address: kostas.senetakis.cityu@gmail.com (K. Senetakis).

https://doi.org/10.1016/j.soildyn.2018.02.001
Received 20 November 2017; Received in revised form 29 December 2017; Accepted 1 February 2018
0267-7261/ © 2018 Elsevier Ltd. All rights reserved.
H. He et al. Soil Dynamics and Earthquake Engineering 110 (2018) 185–194

mineral composition (e.g. [12,7,4,17,33,25,6,42,43,35,27,28], among and unloading small-strain constants and the comparison with a
others). Although work on other materials, for example natural sands of Young’s modulus literature model are given;
single mineralogy or reference granular materials such as glass beads, (ii) Compare the dynamic properties of the poorly graded RCA speci-
has not reported any significant effect on Gmax when changing the mean mens with the uniform fractions and explain the different observed
grain size of the material (e.g. [25,42,35,44]), a recent study on re- trends from a fundamental point of view;
cycled concrete aggregate by He and Senetakis [13] reported a clear (iii) Provide empirical expressions, based on the experiments, to cor-
dependency of the sensitivity of Gmax to pressure on the mean grain size relate small-strain properties to important factors (e.g. pressure
(D50), which sensitivity must be captured for accurate predictions of and grain size) which comprise simple but useful tools in modeling
geo-structures deformations [8]. He and Senetakis [14] conducted a the behavior of geo-materials. This could provide practical but also
characterization of the material through consolidated-drained triaxial some theoretical aspects of the dynamic behavior of recycled
shearing tests, breakage analysis after one-dimensional compression concrete aggregate with many potential applications in geo-
tests and the shear and primary wave velocities as well as the Poisson’s technical engineering.
ratio (ν) measurements of two fractions of RCA, 0.60–1.18 mm and
1.18–2.36 mm. That study highlighted the significant stress history in- 2. Materials and methods
fluence on the dynamic properties of the aforementioned fractions. He
and Senetakis [14] also reported on the small to medium strain stiffness 2.1. Material properties
degradation and damping characteristics throughout the conduction of
resonant column tests. Although He and Senetakis [14] reported on the The recycled concrete aggregate tested in the current study was
shear and Young’s moduli of RCA, that study focused only on two demolished, crushed and supplied by an Australian supplier (New South
uniform fractions, while in another study, He and Senetakis [13] in- Wales). After the crushing process, the recycled concrete consisted of a
vestigated the effect of mean grain size on Gmax, covering a wide range well-graded aggregate which includes gravel, sand and silt sized grains.
of sizes but of uniform fractions with no information for constrained Therefore, sieving was conducted before testing and sand sized ag-
modulus, Young’s modulus and Poisson’s ratio which would be neces- gregates are focused on in this study. Five uniform fractions (with
sary to capture completely the small-strain properties of RCA. coefficient of uniformity, Cu ≈ 1.4) together with three better graded
For the complete modeling of the small-strain behavior of geo-ma- materials, denoted as poorly-graded fractions (Cu ≈ 2.8), which are
terials, it is necessary to quantify, apart from the shear modulus, the mixtures of three successive single fractions, were studied and the
constrained modulus or Young’s modulus as well as the Poisson’s ratio. grading curves for all the materials are given in Fig. 1. The particle sizes
However, in the literature, there has been a relatively limited amount of
work examining both shear and primary wave velocities (or shear and
Young’s modulus) as well as material Poisson’s ratio of sands (e.g.
[34,20,43,5,28]). Even though in many research works and practical
applications, Poisson’s ratio is considered as a constant, recent labora-
tory works have highlighted that Poisson’s ratio (ν) is notably affected
by the effective confining pressure in the range of small strains, while,
Kumar and Madhusudhan [20] found an additional effect of material
porosity. An effort to explain the micromechanics behind the pressure
dependency of Poisson’s ratio was presented in their study by Gu et al.
[10] through numerical simulations. Given that the Poisson’s ratio is
affected by the confining pressure, the necessity of the knowledge of
both shear and Young’s moduli of the material is stressed.
In this study, the constrained modulus, Young’s modulus and
Poisson’s ratio of a range of uniform fractions (from 0.15–0.30 mm to
2.36–4.75 mm) and poorly graded recycled concrete aggregate (RCA)
specimens are investigated with a set of bender/extender element tests.
The small-strain constants of different uniform fractions were measured
and compared through an array of dynamic tests to study whether there
is an effect of mean grain size on the dynamic behavior of the RCA,
particularly in correlating the different small-strain properties (i.e.
constrained modulus and Poisson’s ratio) with the size and size dis-
tribution of RCA grains. This is particularly important for this type of
geo-material since SEM-EDS analysis from He and Senetakis [13] has
shown that RCA is a multi-compositional material and this composition
varies with the fraction (i.e. grain size). The Young’s modulus constants
during the isotropic swelling stages are also examined for a limited
number of specimens. In practice, it is possible that RCA will be used in
a relatively better graded form rather than a single size fraction. Thus,
apart from the study of uniform fractions of RCA, which provides some
fundamental insights into the behavior of those complex materials,
additional experiments on better graded specimens (denoted in the
study as poorly graded samples) were conducted to study the effect of
grading. Thus, the major contributions of this work are:

(i) The complete characterization of the small-strain properties of

RCA fractions providing some insights into the role of their multi-
compositional nature (exploring constraint modulus, Young’s Fig. 1. Grading curves of recycled concrete aggregate (RCA) specimens tested:
modulus and Poisson’s ratio). A detailed analysis of the loading (a) uniform fractions (Cu ≈ 1.4) (b) poorly graded fractions (Cu ≈ 2.8).

186
H. He et al. Soil Dynamics and Earthquake Engineering 110 (2018) 185–194

Table 1
Dynamic testing program on single fraction (Cu = 1.4) dry RCA specimens.
Νο. Specimen code Fraction (mm) Mean grain size Specific gravity Preparation method Loading Path Max p′ (kPa) Initial void Initial unit weight γdo
D50 (mm) (Gs) ratio eo (kN/m3)

1 RCA02–1 0.15–0.30 0.21 2.62 Air-pluviation Loading 400 1.384 10.78

2 RCA02–2 0.15–0.30 0.21 2.62 Compaction Loading 600 1.274 11.30
3 RCA02–3 0.15–0.30 0.21 2.62 Air-pluviation Loading-Unloading 800 1.368 10.85
4 RCA02–4 0.15–0.30 0.21 2.62 Compaction Loading-Unloading 800 1.190 11.72
5 RCA02–5 0.15–0.30 0.21 2.62 Compaction Loading-Unloading 800 1.347 10.94
6 RCA03–1 0.30–0.60 0.42 2.48 Air-pluviation Loading 600 0.934 12.58
7 RCA03–2 0.30–0.60 0.42 2.48 Compaction Loading 600 1.056 11.83
8 RCA03–3 0.30–0.60 0.42 2.48 Compaction Loading-Unloading 800 1.054 11.83
9 RCA04–1 0.60–1.18 0.84 2.47 Air-pluviation Loading 400 1.264 10.70
10 RCA04–2 0.60–1.18 0.84 2.47 Compaction Loading-Unloading 800 1.175 11.14
11 RCA04–3 0.60–1.18 0.84 2.47 Compaction Loading 400 1.136 11.34
13 RCA05–1 1.18–2.36 1.67 2.47 Compaction Loading-Unloading 600 1.233 10.85
14 RCA05–2 1.18–2.36 1.67 2.47 Compaction Loading 800 1.200 11.01
15 RCA05–3 1.18–2.36 1.67 2.47 Compaction Loading 600 1.248 10.78
16 RCA05–4 1.18–2.36 1.67 2.47 Compaction Loading 400 1.183 11.09
17 RCA06–1 2.36–4.75 2.59 2.59 Air-pluviation Loading-Unloading 400 1.255 11.27
18 RCA06–2 2.36–4.75 2.59 2.59 Compaction Loading-Unloading 800 1.214 11.48

RCA03 and RCA04, with each uniform fraction taking up 33.3% of the
total. The mixing process increased the Cu, while keeping the D50 of the
middle fraction unchanged.
The uniform fractions were classified through their mean grain size,
specific gravity of solids (Gs), particle shape and mineral composition.
The specific gravity values, which are given in Table 1, were tested
adopting the ASTM standard [3]. The particle shape parameters for
different fractions were estimated by visual observation through an
optical microscope and images taken from a Scanning Electron Micro-
scope (SEM). An image of the parent concrete aggregate and a re-
presentative SEM photo of the RCA05 are given in Figs. 2 and 3, re-
spectively. An empirical chart proposed by Krumbein and Sloss [19]
was used as reference to quantify the particle shape characteristics
through visual comparison with the particles and achieve both the
sphericity (S) and roundness (R) measurements. The method is a well-
established, even though empirical, quantification approach to describe
the shape of granular materials. The arithmetic average value of the S
Fig. 2. Image of the parent concrete aggregate material (maximum grain size and R gives the regularity (ρ) of the particle [6]. At least 30 particles
between 10 and 20 mm). were examined by two operators for each fraction and the mean values
and the standard deviations (SDV) of the regularities for the uniform
of the uniform samples are 0.15–0.30 mm (denoted as RCA02), fractions are presented in Fig. 4 (after [13]). Within the scatter of the
0.30–0.60 mm (denoted as RCA03), 0.60–1.18 mm (denoted as RCA04), data, it is observed that ρ decreased from 0.60 to 0.40 from the finest
1.18–2.36 mm (denoted as RCA05) and 2.36–4.75 mm (RCA06), re- (RCA02) to the coarsest (RCA06) fractions.
spectively. The samples with higher Cu were produced by evenly mixing Except for the presence of minor amount of silty material and bricks,
three successive fractions. For example, RCA234 consisted of RCA02, the recycled concrete aggregate is mostly composed of aggregate and
cement mortar. From previous studies, it has been reported that the
percentage of cement mortar increases as the aggregate mean grain size

Fig. 3. Scanning Electron Microscope (SEM) image of RCA05 (1.18–2.36 mm)

fraction. Fig. 4. Particle shape regularity mean values ± one standard deviation [13].

187
H. He et al. Soil Dynamics and Earthquake Engineering 110 (2018) 185–194

Fig. 5. EDS testing results: Calcium to Silicon ratio for pure cement, different
uniform fractions of RCA and one poorly graded RCA (updated after [13]).

decreases [39]. As the dynamic properties of the soil, particularly the

modulus – pressure relationship, are closely related to the mineralogy
[17,33,35,4], the percentage of the aggregate and cement mortar in
different fractions is a key parameter. To measure the percentage of
different components of the RCA, a set of Scanning Electronic Micro-
scope-Energy Dispersive Spectroscopy (EDS) tests was performed on a
pure cement specimen, four specimens from four different uniform
fractions (RCA02, RCA03, RCA05 and RCA06) and a poorly graded
Fig. 6. The schematic plot of the Stokoe-type resonant and the dimensions of
specimen (RCA234). The EDS analysis is an effective and accurate way the specimen.
to identify, map and quantify the elements in specimens. He and Se-
netakis [13] conducted EDS analysis of the uniform RCA fractions.
Through their EDS results, it is shown that among a variety of elements, 0.5 × Vp2 − Vs2
ν=
the most abundant in the RCA specimens are calcium (Ca), silicon (Si) Vp2 − Vs2 (2)
and oxygen (O) (the amount of carbon (C) could not be captured be-
From the knowledge of Mmax and ν, the small-strain Young’s mod-
cause the specimens were carbon coated prior to the tests to increase
ulus could be derived from the following equation:
their conductivity). Since the cement mortar is mostly composed of
calcium compounds, while the natural aggregates are predominantly Mmax × (1 + ν ) × (1 − 2 × ν )
Emax =
silicon compounds, the Ca/Si ratio is used as an indication of the (1 − ν ) (3)
composition. The measured Ca/Si ratios for some representative frac-
tions, one pure cement specimen, together with one better graded
specimen (RCA234) are presented in Fig. 5 (updated after [13]). It was 2.3. Testing program
found that the calcium to silicon ratio appeared to decrease as the mean
grain size of the uniform fractions increases and this ratio was the All the specimens were constructed in a metal split mold in 3 layers
highest for the pure cement specimen. As expected, the RCA234 spe- of similar heights. The dry air-pluviation technique was adopted to
cimen gave a value close to that of the RCA03 (of similar mean grain prepare loose specimens with high initial void ratios (eo). For medium
size). The decrease of the ratio Ca/Si indicated that the relative amount dense to dense specimens, a light-weight rod was used to compact the
of cement mortar content decreases gradually and the coarser fractions specimens in layers achieving higher relative density. Before removing
are composed mostly of aggregate. the mold, a small vacuum (≈ 5 kPa) was applied to hold the specimen.
Then, the dimensions of the specimen were measured and recorded
2.2. Dynamic experimental methods under the suction for the calculation of the initial void ratio (eo). The
testing program of the uniform fraction specimens is given in Table 1,
A fixed-free type of Stokoe resonant column system with bender/ while the testing series for the poorly graded specimens is listed in
extender element inserts implemented was used to conduct dynamic Table 2. All the specimens were loaded isotropically in progressive
tests under different isotropic confining pressures. As is illustrated in stages. BE and EE tests were conducted at 50 kPa, 100 kPa, 200 kPa,
Fig. 6, the resonant column system houses a specimen with dimensions 400 kPa, 600 kPa (if applicable) and 800 kPa (if applicable), after
of 10 cm in height and 5 cm in diameter. The resonant column was used which, dynamic tests were also conducted on some of the uniform
to apply a range of isotropic pressures to the specimens from 50 to fraction specimens during the isotropic swelling stage (details are given
800 kPa. On both the top cap and the pedestal, a piezo-element insert is in Table 1). The maximum pressure each specimen reached was
located, which is capable of generating both shear waves (S-waves) and 400 kPa, 600 kPa or 800 kPa.
primary waves (P-waves). When configured to bender element (BE)
mode, one of the piezo-inserts triggers shear waves (S-waves) and the 3. Tests results and analysis
other acts as receiver, while in extender element (EE) mode, the role of
sender and receiver are exchanged to generate and receive primary 3.1. Extender element signal analysis and Mmax-p′ relationship
waves (P-waves) [24]. The shear wave velocity (Vs) or primary wave
velocity (Vp) is then calculated from the recorded wave propagation Both the first arrival and peak to peak methods were performed to
time and the distance between the tips of the inserts. Thereafter, the estimate the primary wave (P-wave) propagation time and a re-
small-strain constrained modulus (Mmax) and Poisson’s ratio (ν) can be presentative signal analysis from one of the specimens is given in Fig. 7.
calculated using Eqs. (1) and (2), where ρs is the mass density of the soil The input signal had a voltage amplitude of 14 V and the received signal
specimen. waves were amplified 500 times to be distinguishable. Similar to the
work by Kumar and Madhusudhan [21], both the first arrival and peak
Mmax = ρs × Vp2 (1) to peak methods gave satisfactorily close values for all the tests

188
H. He et al. Soil Dynamics and Earthquake Engineering 110 (2018) 185–194

Table 2
Dynamic testing program on poorly graded (Cu = 2.8) dry RCA specimens.
Νο. Specimen code Fraction (mm) Preparation method Max p′ (kPa) eo γdo (kN/m3)

1 RCA234–1 0.15–1.18 Compaction 800 0.895 13.30

2 RCA234–2 0.15–1.18 Compaction 800 1.002 12.35
3 RCA234–3 0.15–1.18 Air-pluviation 800 1.086 11.86
4 RCA345–1 0.30–2.36 Compaction 800 0.888 12.84
5 RCA345–2 0.30–2.36 Compaction 800 0.991 12.17
6 RCA345–3 0.30–2.36 Air-pluviation 800 1.090 11.59
7 RCA456–1 0.60–4.75 Compaction 800 1.000 12.30
8 RCA456–2 0.60–4.75 Compaction 400 1.039 12.06
9 RCA456–3 0.60–4.75 Air-pluviation 400 1.118 11.61

Fig. 8. Comparison between wave velocities (Vs and Vp) results derived from
first arrival and peak to peak methods during bender/extender element tests.

effective radial stresses, respectively.

Fig. 7. Example of the signal analysis methods for extender element test (spe-
cimen RCA05-1) Note: the amplitude for the source wave was at 14 V while the Typical plots of normalized Mmax versus mean effective stress (p′)
received wave was magnified 500 times to be discernable in the same plot. from two representative specimens and their best fitting power-law
equations are given in Fig. 9. All the tests were analyzed in a similar
way to Fig. 9, and a summary of the results from the tests on the uni-
(maximum error ≤ 10%) but the first arrival method was adopted for
form RCA fractions, in terms of the constrained modulus constants AM
further analysis. For the shear wave velocities, the average of the first
and nM are plotted against the mean grain size D50 is given in Fig. 10. As
arrival and peak to peak values were adopted as explained by He and
is shown on Fig. 10(a), there is a very clear trend that AM increases as
Senetakis [14]. This interpretation method of the shear wave velocity
the mean grain size increases. Based on a power-law best fitting curve
was validated by the satisfactory comparison between the bender ele-
of the AM-D50 plot (Fig. 10(a)), the power is found to be 0.44. Despite
ment and the resonant column test results by He and Senetakis [14]. A
some scatter of the finest fraction RCA02 (0.15–0.30 mm), the trend is
comparison of Vs and Vp measurements between the two different
quite clear with a coefficient of determination (R2) of 0.88.
methods used is given in Fig. 8.
Based on Eq. (1), with the measured Vp and the density of the
specimen, the small-strain constrained modulus Mmax can be calculated.
Since the RCA specimens are isotropically loaded, when plotting the
normalized Mmax against normalized mean effective stress p′ the data
follow a power-law type of the general expression (given in Eq. (4)) that
is widely adopted in soil dynamics research:

p′ ⎞nM
Mmax / f (e ) = (AM ) × ⎛
⎝ 1kPa ⎠ (4)

f (e ) = e−1.3 (5)

where AM is material constant and nM is a power that depicts the sen-

sitivity of the stiffness to pressure. The f(e) is a void ratio function
(given in Eq. (5)) used to eliminate the density effect in the analysis
(after [16]) and the mean effective stress is divided by 1kPa to remove
the unit. Note that since the specimens are isotropically consolidated, Fig. 9. Typical plot of normalized constrained modulus (Mmax) against mean
p″ = σ'a = σ'r, where σ'a and σ'r correspond to the effective axial and effective stress (p′) from two representative specimens: RCA06-1 and RCA03-2.

189
H. He et al. Soil Dynamics and Earthquake Engineering 110 (2018) 185–194

Fig. 10. Illustration of the effect of mean grain size (D50) on the small-strain
constrained modulus constants: (a) Small-strain constant AM and (b) Power nM. Fig. 11. Comparison between predicted and measured small-strain Young’s
modulus values using the expression proposed by Payan et al. [28]: (a) mea-
sured Emax and predicted Emax against pressure from two representative tests;
(b) all the measured and predicted Young’s moduli.
Similarly, the power nM is plotted against D50 in Fig. 10(b) for all
the specimens. The power values decrease as D50 increases. For RCA02, 3.2. Emax-p′ relationships
the nM values are between 0.49 and 0.54, while those for RCA06 are in
the range of 0.34–0.35. This drop is quite drastic as the average power Like Mmax, the small-strain Young’s modulus (Emax) can also be
value dropped about 33% from the finest to the coarsest fraction. The fitted by a power law general formula, which is given in the following
power law best fitting of the nM-D50 curve gives a power value of −0.15 equation.
and the R2 value is equal to 0.93.
He and Senetakis [13] found a decrease of nG values as particle size p′ ⎞nE
Emax / f (e ) = (AE ) × ⎛
increases, though much smaller in magnitude, but they did not report ⎝ 1kPa ⎠ (6)
any systematic effect of particle size on the small-strain constant AG
where AE and nE are the Young’s modulus constants and the void ratio
values. Yang and Gu [44] studied the small-strain shear modulus of a
function f(e) is given in Eq. (5). From a previous study, Payan et al. [28]
reference material (i.e. glass beads) of three different grain sizes (D50).
proposed a model for Emax prediction based on a set of tests on quartz
Through their analysis, both experimental and micromechanical, it was
sands with a range of particle regularities (ρ) and coefficients of uni-
found that the variation of the shear modulus between specimens
formity (Cu). Based on their tests, the proposed general model is given
constructed of different mean particle sizes was negligible. Therefore,
as follows,
the trends of increasing AM and decreasing nM of the RCA specimens
with the increase of grain size should be attributed, predominantly, to (Cu0.11)(−0.44ρ + 0.66)
p′
the change of the composition of the RCA fractions. This is because the Emax = (245Cu−0.09 × ρ0.82 ) × e−1.32 × ⎛⎜ ⎞⎟
stiffness – pressure relationship of granular materials is affected ⎝ pa ⎠ (7)
markedly by the grain contact response which in turn is affected by the where Cu is the coefficient of uniformity of the specimen, ρ is the par-
material of the surfaces in contact [33,36,4]. For smaller sizes of RCA, ticle shape regularity, e is the void ratio of the specimen at the corre-
there is an increase of cement-cement and cement-aggregate contacts sponding pressure stage and pa is the atmospheric pressure (100kPa).
since the content of cement mortar increases. This gives rise to a more The formula was used to predict the Young’s modulus of the RCA
pronounced plastic-to-brittle in nature contact response which is be- specimens and the comparison between the predicted and measured
lieved to contribute to the trends shown in Fig. 10. This is because the results is given in Fig. 11. The measured and predicted Emax against the
nature of particle contact response has a dominant effect on the mod- mean effective stress p′ from two representative specimens are plotted
ulus – pressure relationship, expressed through the power nM in in Fig. 11(a). It is found that for the RCA particle shapes and grading
Fig. 10(b) [4]. characteristics, the Young’s modulus values are predicted to be much
smaller than the measured ones. In Fig. 11(b), the maximum error of
the stiffness prediction is about 50% compared to the measured values.

190
H. He et al. Soil Dynamics and Earthquake Engineering 110 (2018) 185–194

Fig. 12(a) and (b) are equal to 0.69 and 0.89, respectively. In terms of
small-strain Young’s modulus, the constants do not vary as much as the
AM and nM values. The average power value nE for each uniform frac-
tion only dropped from 0.56 to 0.48 from RCA02 (D50 = 0.21 mm) to
RCA06 (D50 = 2.59 mm).
The trends of the constants indicate that the coarser fractions ex-
hibited higher Young’s moduli in the lower pressure range, but they
were less sensitive to pressure increments compared to the finer frac-
tions. From the previous study by He and Senetakis [13], a similar drop
of the small-strain shear modulus (Gmax) power nG, when D50 increased,
was found. However, unlike the AE or AM, they reported no specific
trend for the AG values of different mean grain size uniform fractions.

3.3. ν-p′ relationships

The Poisson’s ratio of the specimens can be calculated from Eq. (2)
with the knowledge of both Vs and Vp. The ν-p′ relationship can be
fitted with a power law type equation as follows,

p′ ⎞n ν
ν = Aν × ⎛
⎝ 1kPa ⎠ (10)

where Aν and nν are the Poisson’s ratio constants of the material. In

agreement with previous works on granular soils [20,26,43,5] as well
as the previous study on two of the fractions of RCA [14], the power nν
is negative value for the specimens tested as the Poisson ratio decreased
with the increase of the mean effective stress. The Poisson’s ratio of the
RCA specimens ranged between 0.30 and 0.37 at 25 kPa confining
pressure and decreased to 0.25–0.30 at 800 kPa confining pressure.
There was no effect of density on the constants Aν and nν observed from
any of the fractions tested, which is in accordance to the observations
by Wichtmann and Triantafyllidis [43] and He and Senetakis [14].
To illustrate the effect of D50 on the Poisson ratio, the mean absolute
values of Aν and nν and their standard deviations are plotted against the
Fig. 12. Illustration of the effect of mean grain size (D50) on the small-strain
mean grain size for different uniform fraction specimens in Fig. 13. No
Young’s modulus constants: (a) Small-strain constants AE versus D50 and (b) monotonic trend was observed from the data points and the constants
Power nE against D50. for the coarser fractions were more scattered.

In a previous study of the RCA Gmax, He and Senetakis [13] have also 3.4. Isotropic swelling Emax-p′ relationships
reported underestimation of the small-strain shear modulus from the
general formula proposed by Payan et al. [27] based on quartz sands. One to three specimens from each fraction were tested during both
The failure of the prediction in both cases is believed to be due to two isotropic loading (compression) and unloading (swelling) stages
reasons. Firstly, similar to the volcanic sands studied by Senetakis et al. (Table 1). Similar to the previous observations of Vp measurements for
[35,37], the RCA is a material with intra-particle voids due to the RCA04 and RCA05 by He and Senetakis [14] and small-strain moduli of
presence of cement mortar, which leads to higher void ratio and lower carbonate sand with crushable grains during the unloading stages by
dry density specimens compared to typical quartz sand specimens that Jovicic and Coop [17] and He et al. [15], the small-strain Young’s
are composed of massive grains. Payan et al. [27,28] proposed models modulus was increased during the unloading stages compared to the
for Gmax and Emax based on experiments on quartz sands that had no corresponding stress level (p′) in the loading stages. The loading and
intra-particle voids, so only inter-particle voids were considered when unloading normalized Emax behavior of one representative specimen
calculating the void ratio values. Consequently, as the void ratio is (RCA03-3) is plotted as an example in Fig. 14. The increase of the
elevated by the intra-particle voids, the Emax prediction would be stiffness in the unloading stages would result in a higher unloading AE
lowered. Secondly, the particle surface of the RCA is always coated with value (equal to 49.1 MPa) and a lower unloading nE value (equal to
a cement layer, which can be observed from the close view of the SEM 0.33) compared to the loading process constants (AE = 11.1 MPa, nE
images. The cement layer is expected to be much softer and more = 0.55), which indicates that the sensitivity of the stiffness to pressure
plastic in nature compared to quartz sand surfaces. Therefore, the (slope of modulus-pressure relationship curve) is reduced after the
stiffness sensitivity to pressure is also expected to be higher [36,4]. material experiences a stress history. The comparison of the loading and
Thereafter, when plotting the constants AE and nE versus D50 unloading power value nE for all the tests is given in Fig. 15. Although
(Fig. 12), quite similar trends to the AM and nM against D50 plots were there is a small scatter in the limited number of unloading power va-
observed as AE increases while nE decreases as the mean grain size lues, it is shown that during unloading, all the fractions are exhibiting a
increases. When fitted to the power law, the following relationships for quite similar level of sensitivity to the stress change, or similar nE values
AE and nE with D50 are found, (around 0.34). The effect of the stress history is believed to be attrib-
uted to different factors including the rearrangement of the grains, the
AE = 18.8(D50)0.23 (8) plastic-to-brittle contact response and the small amount of particle
breakage during the isotropic compression stage [14,4].
nE = 0.51(D50)−0.07 (9)
For the purpose of comparing the magnitude of the effect of stress
The coefficient of determination R values for the power fitting in
2
history between different fractions, the ratio between the unloading

191
H. He et al. Soil Dynamics and Earthquake Engineering 110 (2018) 185–194

Fig. 15. The comparison between loading and unloading power constant nE.

Fig. 13. Illustration of the mean value ± one standard deviation for Poisson’s
ratio constants versus the mean grain size (D50) from uniform fraction speci-
mens: (a) Constants Aν versus D50 and (b) Power nν against D50 (nv are shown as
absolute values).

Fig. 14. Loading and unloading behavior of normalized small-strain Young’s

modulus Emax/f(e) against mean effective stress p′ (from specimen RCA03-3).
Fig. 16. Unloading over loading small-strain Young’s modulus constants versus
mean grain size (D50): (a) Unloading AE/loading AE against D50; (b) Unloading
and the loading constant is calculated for each test and plotted against
nE/loading nE against D50.
D50 in Fig. 16. From the limited number of tests, it is shown that both
ratios of (unloading AE/loading AE) and (unloading nE/loading nE)
values for the finer fractions deviate more from unity than those for Therefore, this study extended the analysis of the stiffness – pressure
coarser fractions. The finer fractions are therefore more affected than relationship to a set of RCA fractions with increased coefficient of
the coarser ones in terms of modulus change from the loading (com- uniformity and compared to the results from the tests on uniform
pression) to the unloading (swelling) stage. fractions (mainly discussed in Section 3.2) to examine the effect of Cu
on Emax. Three soils (RCA234, RCA345 and RCA456, respectively), each
consisting of three consecutive uniform fractions, were tested with
3.5. Effect of Cu on Emax constants bender/extender element tests under several progressively increased
isotropic stress stages (Table 2). Thereafter, similar to the uniform
The RCA used for engineering projects may have a better grading.

192
H. He et al. Soil Dynamics and Earthquake Engineering 110 (2018) 185–194

Fig. 18. Comparisons of the effect of mean grain size (D50) on the small-strain
Young’s modulus constants between poorly graded (Cu = 2.8) and single frac-
tions (Cu = 1.4): (a) small-strain constants nE versus D50 for each single test and
Fig. 17. Comparisons of the effect of mean grain size (D50) on the small-strain (b) average nE against D50.
Young’s modulus constants between poorly graded (Cu = 2.8) and single frac-
tions (Cu = 1.4): (a) small-strain constants AE versus D50 for each single test and
(b) average AE against D50. is lower than those for the coarser fractions included. When the soil
contains a certain amount of softer (or weaker) material, the contact
response is more likely to be dominated to some degree by the softest
fraction tests, the Young’s modulus was derived, normalized and fitted
component, which is also found by the previous studies on soil-rubber
with the power law equation given in Eq. (6).
mixtures composed of soft rubber grains and stiff sand grains (e.g.
Fig. 17(a) provides the AE constant for each individual test against
[9,2]). In sand-rubber mixtures, which comprise as extreme example
its D50, including both the uniform fractions (Table 1) and the poorly
from the point of view of different properties between the two domi-
graded specimens (Table 2). Through the comparison, despite the small
nant elements (i.e. rubber and sand grains), the surface contact beha-
scatter, at the same mean grain size level, it is observed clearly that the
vior is changed drastically compared to the pure sand, since the
better-graded materials have lower small-strain constant (AE) than the
granulated rubber can be considered as highly deformable with nearly
uniform ones. In order to quantify the effect of Cu on the constant AE,
no stiffness in the small-strain range. Consequently, the wave propa-
Fig. 17(b) shows the average AE versus D50 for materials with different
gation velocity is significantly reduced by the rubber thus reducing the
gradings but same mean grain size (D50). From the limited number of
stiffness of the mixture. For the RCA mixed fractions of the present
specimens tested at Cu ≈ 2.8, the decrements of the AE are quite clear
study, the inclusion of 33% of softer fine fraction materials (i.e. in-
and significant (23%, 23% and 11% for materials with D50 = 0.42, 0.84
creased cement mortar component) would lower the general stiffness of
and 1.67 mm, respectively) as the coefficient of uniformity (Cu) in-
the mixture drastically.
creased from 1.4 (uniform fractions specimens) to 2.8 (poorly graded
In terms of the power constant nE, no specific trend was observed
specimens). From the previous study on quartz sands, Payan et al. [28]
when changing the grading of the recycled concrete aggregate (as
noticed that the increase of Cu would result in a decrease of the AE
shown in Fig. 18). Through the current set of tests, the sensitivity of the
constant, however, this effect was found to be less pronounced for
different gradings but same mean grain size materials appeared to be
quartz sands than the results on the RCA fractions of this study. From
quite similar as shown in both Fig. 18(a) and (b). The power nE of the
their empirical equation (Eq. (7)), when Cu is increased from 1.4 to 2.8,
poorly graded specimens displayed a decreasing trend as the mean
the AE constant of the quartz sand would be predicted to drop about
grain size increases, which is identical to what was observed in the
6.0%. The notable drop of the AE value for the RCA fractions is believed
uniform fraction tests. With the individual data points scattered in the
to be mainly due to the component of the finest fraction in the mixture
same range between both materials in Fig. 18(a), the average nE values
of the RCA fractions. For the uniform RCA specimens, it was found that
of two grading materials are close as is plotted in Fig. 18(b). However,
the finer fractions had lower AE constant values, which means that the
Payan et al. [28] found a drop of the power while increasing the
stiffness of the material is lower in comparison to the coarser fractions.
coefficient of uniformity. According to their empirical equation in Eq.
From the data analysis (Fig. 17(a)), it can be observed that despite a
(7), an increase of Cu from 1.4 to 2.8 would result in a drop of nE of
small scatter, the mixture of a fine fraction with two successive coarser
about 7.9%, which was not observed for the RCA tests.
fractions gave AE values close to the uniform fine fraction level, which

193
H. He et al. Soil Dynamics and Earthquake Engineering 110 (2018) 185–194

4. Conclusions [11] Hansen TC. Recycling of demolished concrete and masonry. CRC Press; 2004.
[12] Hardin B, Richart F. Elastic wave velocities in granular soils. J Soil Mech Found Div
1963;89:33–65. [SM1][Proc. Paper 3407].
The small-strain constrained modulus (Mmax), Young’s modulus [13] He H, Senetakis K. The effect of grain size on g max of a demolished structural
(Emax) and Poisson’s ratio (ν) of a recycled concrete aggregate (RCA) concrete: a study through energy dispersive spectroscopy analysis and dynamic
were examined in this study. The different fractions of RCA were element testing. Soil Dyn Earthq Eng 2016;89:208–18.
[14] He H, Senetakis K. A study of wave velocities and poisson ratio of recycled concrete
characterized by their particle size, specific gravity, and composition aggregate. Soils Found 2016;56(4):593–607.
using the Energy Dispersive Spectroscopy (EDS) technique. Through a [15] He H, Senetakis K, Ranjith P. The behavior of a carbonate sand subjected to a wide
set of bender and extender element tests, the dynamic properties of a set strain range of medium-frequency flexural excitation. Geomech Geophys Geo-
Energy Geo-Resour 2017;3(1):51–60.
of uniform fraction specimens and an array of poorly graded specimens [16] Jamiolkowski M, Leroueil S, Lo Presti D. Design parameters from theory to practice.
were measured and analyzed. The small-strain Young’s modulus values in: Proceedings of the international conference on geotechnical engineering for
of the uniform fraction specimens are compared with predicted values coastal development. In Geo-Coast 1991, Coastal Development Institute of
Technology, Yokohama, Japan; 1991. p. 877–917.
from an existing empirical equation and the comparison was found to
[17] Jovicic V, Coop MR. Stiffness of coarse-grained soils at small strains. Géotechnique
be not satisfactory, with the prediction underestimating the stiffness of 1997;47(3):545–61.
the RCA. It appeared that both the small-strain constrained modulus [18] Jovicic V, Coop MR, Simic M. Objective criteria for determining gmax from bender
and small-strain Young’s modulus of the RCA are quite sensitive to the element tests. Géotechnique 1996;46(2):357–62.
[19] Krumbein WC, Sloss LL. Stratigraphy and sedimentation. 2nd ed W. H Freeman and
particle mean grain size. The stiffness constants AM and AE increased, Company; 1963.
while the power values nM and nE decreased, as the mean grain size [20] Kumar J, Madhusudhan BN. Effect of relative density and confining pressure on
(D50) increased. This phenomenon is primarily attributed to the dif- poisson ratio from bender and extender elements tests. Géotechnique
2010;60(7):561–7.
ferent ratios of cement and aggregate in the different fractions. As for [21] Kumar J, Madhusudhan BN. A note on the measurement of travel times using
the Poisson’s ratio (ν), there was not observed any monotonic trend of bender and extender elements. Soil Dyn Earthq Eng 2010;30(7):630–4.
the constants as the grain size changed. After the maximum pressure [22] Lee J-S, Santamarina JC. Bender elements: performance and signal interpretation. J
Geotech Geoenviron Eng 2005;131(9):1063–70.
had been reached, some of the specimens were tested during isotropic [23] Leong EC, Cahyadi J, Rahardjo H. Measuring shear and compression wave velocities
swelling stages. The sensitivity to confinement change of the RCA of soil using bender-extender elements. Can Geotech J 2009;46(7):792–812.
specimens became smaller during isotropic swelling compared to iso- [24] Lings M, Greening P. A novel bender/extender element for soil testing.
Géotechnique 2001;51(8):713–7.
tropic compression. For finer fractions, the Young’s modulus constants [25] Menq F-Y. Dynamic properties of sandy and gravelly soils. USA.: University of Texas
were altered more by the stress history than those of the coarser ones, at Austin; 2003.
which means the higher percentage of cement mortar in the finer [26] Nakagawa K, Soga K, Mitchell JK. Pulse transmission system for measuring wave
propagation in soils. J Geotech Eng 1996;122(4):302–8.
fractions also resulted in a higher sensitivity to stress history. When
[27] Payan M, Khoshghalb A, Senetakis K, Khalili N. Effect of particle shape and validity
testing RCA specimens with a higher coefficient of uniformity (Cu), a of Gmax models for sand: a critical review and a new expression. Comput Geotech
clear drop of the small-strain constants AE was observed compared to 2016;72:28–41.
uniform fraction specimens with the same mean grain size. The pre- [28] Payan M, Senetakis K, Khoshghalb A, Khalili N. Effect of gradation and particle
shape on small-strain young's modulus and poisson’s ratio of sands. Int J Geomech
sence of the finer fraction may have softened the particle contact re- 2017;17(5).
sponse, which led to this drop. On the other hand, the power nE values [29] Poon C-S, Chan D. The use of recycled aggregate in concrete in hong kong. Resour
of the same mean grain size remained the same when Cu increased. Conserv Recycl 2007;50(3):293–305.
[30] Rahardjo H, Satyanaga A, Leong E-C, Wang J-Y. Unsaturated properties of recycled
concrete aggregate and reclaimed asphalt pavement. Eng Geol 2013;161:44–54.
Acknowledgements [31] Rilem Recommendation. Specifications for concrete with recycled aggregates.
Mater Struct 1994;27(173):557.
[32] Sagoe-Crentsil KK, Brown T, Mak SL, Taylor A. Engineering properties and per-
The authors acknowledge the financial support by the grant from formance of concrete made with recycled construction aggregates. In: National
the Research Grant Council of the Hong Kong Special Administrative symposium on the use of recycled materials in engineering construction: 1996;
Region (HKSAR), China Project No. 9041880 (CityU 112813) and the Programme & Proceedings. Institution of Engineers, Australia; 1996, p. 132.
[33] Santamarina C, Klein K, Fam M. Soils and waves. New York: John Wiley and Sons;
start-up grant of the College of Science and Engineering, City University
2001.
of Hong Kong (Code: 7200533-ACE). [34] Saxena SK, Reddy KR. Dynamic moduli and damping ratios for monterey no. 0 sand
by resonant column tests. Soils Foundations 1989;29(2):37–51.
[35] Senetakis K, Anastasiadis A, Pitilakis K. The small-strain shear modulus and
References
damping ratio of quartz and volcanic sands. Geotech Test J 2012;35(6):964–80.
[36] Senetakis K, Madhusudhan BN. Dynamics of potential fill–backfill material at very
[1] Alvarado G, Coop MR. On the performance of bender elements in triaxial tests. small strains. Soils Found 2015;55(5):1196–210.
Géotechnique 2012;62(1):1–17. [37] Senetakis K, Madhusudhan BN, Anastasiadis A. Wave propagation attenuation and
[2] Anastasiadis A, Senetakis K, Pitilakis K. Small-strain shear modulus and damping threshold strains of fully saturated soils with intra-particle voids. J Mater Civil Eng
ratio of sand-rubber and gravel-rubber mixtures. Geotech Geol Eng ASCE 2016;28(2):04015108.
2012;30(2):363–82. [38] Shirley DJ, Anderson AL. In situ measurement of marine sediment acoustical
[3] ASTM D845. 854-00 standard test method for specific gravity of soil solids by water properties during coring in deep water. IEEE Trans Geosci Electron
pycnometer. Annual book of ASTM standards 4. American Society for Testing and 1975;13(4):163–9.
Materials; 2000. p. 02. [39] Tam VWY, Tam CM. Crushed aggregate production from centralized combined and
[4] Cascante G, Santamarina JC. Interparticle contact behavior and wave propagation. J individual waste sources in hong kong. Constr Build Mater 2007;21(4):879–86.
Geotech Eng 1996;122(10):831–9. [40] Tatsuoka F, Tomita Y-I, Iguchi Y, Hirakawa D. Strength and stiffness of compacted
[5] Chen G, Zhou Z, Pan H, Sun T, Li X. The influence of undrained cyclic loading crushed concrete aggregate. Soils Found 2013;53(6):835–52.
patterns and consolidation states on the deformation features of saturated fine sand [41] Viggiani G, Atkinson J. Stiffness of fine-grained soil at very small strains.
over a wide strain range. Eng Geol 2016;204:77–93. Géotechnique 1995;45(2):249–65.
[6] Cho G-C, Dodds J, Santamarina JC. Particle shape effects on packing density, [42] Wichtmann T, Triantafyllidis T. Influence of the grain-size distribution curve of
stiffness, and strength: natural and crushed sands. J Geotech Geoenviron Eng quartz sand on the small strain shear modulus gmax. J Geotech Geoenviron Eng
2006;132(5):591–602. 2009;135(10):1404–18.
[7] Chung RM, Yokel FY, Drnevich V. Evaluation of dynamic properties of sands by [43] Wichtmann T, Triantafyllidis T. On the influence of the grain size distribution curve
resonant column testing. Geotech Test J 1984;7(2):60–9. on p-wave velocity, constrained elastic modulus m max and poisson's ratio of quartz
[8] Clayton C. Stiffness at small strain: Research and practice. Géotechnique sands. Soil Dyn Earthq Eng 2010;30(8):757–66.
2011;61(1):5–37. [44] Yang J, Gu X. Shear stiffness of granular material at small strains: does it depend on
[9] Feng Z, Sutter K. Dynamic properties of granulated rubber/sand mixtures. Geotech grain size? Géotechnique 2013;63(2):165.
Test J 2000;23(3):338–44. [45] Youn J-U, Choo Y-W, Kim D-S. Measurement of small-strain shear modulus gmax of
[10] Gu X, Yang J, Huang M. Laboratory measurements of small strain properties of dry dry and saturated sands by bender element, resonant column, and torsional shear
sands by bender element. Soils Found 2013;53(5):735–45. tests. Can Geotech J 2008;45(10):1426–38.

194