Soils and Foundations 2016;56(2):189–204

HOSTED BY The Japanese Geotechnical Society

Soils and Foundations

www.sciencedirect.com
journal homepage: www.elsevier.com/locate/sandf

Tensile strength of compacted rammed earth materials

Hiroyuki Arakia,n, Junichi Kosekib, Takeshi Satoc
a
Formerly, Department of Civil Engineering, the University of Tokyo, Japan
b
Department of Civil Engineering, the University of Tokyo, Japan
c
Integrated Geotechnology Institute Ltd., Japan
Received 30 June 2015; received in revised form 3 August 2015; accepted 12 November 2015
Available online 15 March 2016

Abstract

A rammed earth technique is a traditional architectural technique to build soil structures by compacting geo-materials in a form. In this study,
tensile strength properties of rammed earth materials and effects of test conditions are evaluated by conducting direct tension tests and splitting
tests on several specimens. It is inferred from the result of these test that the direct tension test should be used to evaluate tensile strength of a
layer interface, while the splitting test might evaluate the strength related to the tensile strength inside a compaction layer. When rammed earth
structures are constructed, the results from the experiments indicate that it is important to scarify interfaces during compaction in order to prevent
reduction and variation of the tensile strength of the layer interfaces. The tensile strength of the rammed earth specimens made of compacted soils
without lime increased with the decrease in their water contents. The tensile strengths of the specimens represent 5.0–12.5% of corresponding
values of unconfined compression strength at the same water content range. The tensile strength of rammed earth specimens made of lime-mixed
soil was in the range of 15–20% of the corresponding unconfined compression strength at 28 days curing. In the series of cyclic tensile loading
test, no significant reduction of tensile strengths was observed even after undergoing the cyclic loading history.
& 2016 The Japanese Geotechnical Society. Production and hosting by Elsevier B.V. All rights reserved.

Keywords: Tensile strength; Direct tension test; Splitting test; Rammed earth; Compacted soil; Lime-mixed material; Cyclic loading

International Geotechnical Classification Numbers: D00 [General]

1. Introduction strengthen by drying or solidifying by chemical reaction of

The rammed earth technique is an ancient construction
technique to build soil structures by compressing soil mixtures
consisting of clay and/or sand, and other materials such as lime
or cement, if needed. The soil mixture is tamped by a rod in a
form after adjusting the soil mixture’s water content with water
or liquid solution. The compacted soil mixture is cured to
n
Corresponding author. Present address: Department of Civil and Environ-
mental Engineering, Chuo University, 1-13-27 Kasuga, Bunkyo-ku, Tokyo
112-8551, Japan.
E-mail addresses: araki@civil.chuo-u.ac.jp (H. Araki),
koseki@civil.t.u-tokyo.ac.jp (J. Koseki), tsato@iis.u-tokyo.ac.jp (T. Sato).
Peer review under responsibility of The Japanese Geotechnical Society.
lime or cement. This technique, which was used in Yellow
River Valley, China around B.C. 2000 in Asia (Onitsuka et al.,
2007), is still used for walls of traditional houses in several
places of the world (Minke, 2006).
For example, Rendell and Jauberthie (2009) reported that
some rammed earth housings and agricultural housings that are
in excess of 100 years old in east Brittany, France, were
generally in a good state of repair. In addition, Al Hambra
palace in Granada, Spain, Fujan Tulou and the Great wall,
China were built with the rammed earth technique and were
designated by UNESCO as a world heritage (UNESCO, 2014).
In Japan, the rammed earth technique was used for a wall
surrounding a temple and a shrine, such as “Abura-dobei” in

http://dx.doi.org/10.1016/j.sandf.2016.02.003
0038-0806/& 2016 The Japanese Geotechnical Society. Production and hosting by Elsevier B.V. All rights reserved.
190 H. Araki et al. / Soils and Foundations 56 (2016) 189–204
Ryoan-ji temple in Kyoto, “Oo-neribei” in Nishinomiya shrine
in Hyogo and “Oo-gaki” in Horyu-ji temple in Nara; these
walls have been registered as the important cultural properties
by the Agency of Cultural Affairs, Japan.
On the other hand, the advantages of rammed earth
construction have attracted interest in construction projects of
modern housing in recent years, e.g. the high capability for
humidity control, potential for recycling, and the reduction in
construction energy, among others (Minke, 2006). The com-
pression strength of various rammed earth materials (e.g. Hall
and Djerbib, 2004; Jayasinghe and Kamaladasa, 2007; Rendell
and Jauberthie, 2009) and thermal conductivity of rammed
earth walls (e.g. Hall and Allinson, 2009) have been studied in
order to apply the rammed earth wall to a modern housing.
However, the seismic performance of the rammed earth wall
has not been investigated sufficiently, except for a limited
number of studies which are summarized below.
Bui et al. (2009) conducted in-situ dynamic measurements
and laboratory tests of rammed earth walls in order to evaluate
elastic modulus. They reported that the elastic modulus
depended on the specimen scale and measurement method
because of the effects of adhesion between compaction layers,
their water content, and so on. In addition, Bui et al. (2011)
conducted in-situ dynamic measurements of rammed earth
structures located in France in order to evaluate dynamic
parameters, e.g. natural frequencies, damping ratio and Pois-
son’s ratio which were used to determine the dynamic
characteristics. They pointed out that the current seismic
standard (Eurocode 8) could be applicable to the design of
rammed earth structures.
Takadachi and Koshihara (2010) conducted a series of
shaking table tests on a model of rammed earth wall, which
revealed that tensile failure at a bottom part of the rammed
earth wall model causes over-turning of the wall in earth-
quakes. Hence, the tensile strength of the material consisting of
the rammed earth wall is one of important factors in evaluating
its seismic behavior.
In order to evaluate the tensile strength of geo-materials, a
direct tension test, a bending test and a splitting test are
typically conducted. Namikawa and Koseki (2007) conducted
these three kinds of tests on cement-treated sand and showed
that the tensile strengths measured by the direct tension tests
are larger than those measured by the corresponding splitting
tests. They pointed out that partial shear failure that occurred
nearby the loading plate in the splitting tests was one of the
reasons behind the underestimation of tensile strength.
It should be noted that the above study by Namikawa and
Koseki (2007) was conducted focusing on the tensile proper-
ties of cement-treated sand which was artificially cemented
under the saturated condition. Rammed earth walls, however,
are exposed to the atmosphere, and thus are under the
unsaturated condition. The mechanical properties and water
content of atmospherically-exposed soils are affected by soil
suction, which is also related to the relative humidity of the
atmosphere. Jaquin et al. (2009) reported that the water content
of an unsaturated rammed earth material decreased with the
increase in soil suction, which induced as well an increase in
its compressive strength. Bui et al. (2014) also studied the
influence of water content and soil suction on the mechanical
properties of the rammed earth materials including sandy soil,
clayey soil and stabilized soil with natural hydraulic lime, and
showed that soil suction was an important factor for the
compressive strength, secant modulus and Poisson’s ratio.
However, not all the possible effects of water content and soil
suction on the tensile strength of rammed earth materials have
been well studied.
In Japan, traditional rammed earth walls are often made of
the mixture of soil and/or add-in materials such as slaked lime
and quick lime, which are selected depending on soil proper-
ties. Artificial cementation with a chemical reaction would
resist tensile stress mobilized in the rammed earth structure
made of the mixture consisting of soil and add-in material, in a
way that is similar to the cement-treated sands studied by
Namikawa and Koseki (2007) among others. In the rammed
earth structures without add-in material, on the other hand, soil
suction might be assumed to resist the tensile stress as well as
the compressive stress studied by Jaquin et al. (2009) and Bui
et al. (2014).
In the traditional Japanese rammed earth technique, magne-
sium chloride solution or sea water is sometimes used to adjust
the water content of soil mixture with/without add-in material.
These solutions are believed to improve the performance of the
rammed earth wall. For instance, Akatani et al. (2012) reported
that the compressive strength of rammed earth specimens
mixed with calcium hydroxide (slaked lime) and magnesium
chloride solution increased with the increase of the concentra-
tion of the magnesium chloride solution in a case where the
concentration was smaller than 15%.
On the other hand, negative effects of adding magnesium
chloride have also been reported. Tanimoto et al. (2004)
conducted unconfined compression tests of compacted mix-
tures consisting of soil, slaked lime and magnesium chloride at
several mixing ratios, and they reported that the strengths of
some compacted mixtures with magnesium chloride were
smaller than that of the compacted soil mixture without
magnesium chloride. They pointed out that the formation of
artificial cementation by slaked lime was impeded because
calcium hydroxide was non-easily ionizable in the solution
containing high levels of magnesium chloride. Xing et al.
(2009) conducted unconfined compression tests and miner-
alogical analysis on salt-rich soil cement. They showed that the
unconfined compression strength of the salt-rich soil cement
decreased with the increase of the amount of magnesium ion,
chloride ion and sulfate ion in the soil because these ions
impeded the formation of cementation compounds such as
calcium-silicate-hydrate and calcium-aluminate-hydrate.
It is to be noted that magnesium chloride solution is also
known as one of the materials used for magnesium oxychloride
cement. The magnesium oxychloride cement, consisting of
magnesium chloride solution and magnesium oxide, produces
some crystal phases by the chemical reaction (e.g. Dehua and
Chuanmei, 1999), and it has a suitable molar ratio of
magnesium oxide, magnesium chloride and water for posses-
sing a certain crystal phase to maximize the compressive
 H. Araki et al. / Soils and Foundations 56 (2016) 189–204 191

strength (Li and Chau, 2007). However, such chemical
reaction might not occur in the traditional rammed earth
technique, because the magnesium oxide is not usually used.
As described above, the role of magnesium chloride in soil
has not been fully-clarified. In particular, the effects of
magnesium chloride on the mechanical properties of rammed
earth without lime or cement have not even been revealed.
In this study, in order to evaluate the tensile properties of
rammed earth structures, direct tension tests and splitting tests
are conducted on three kinds of compacted rammed earth
materials, which are soils with/without magnesium chloride
and sand with lime. The possible effects of several test
conditions are revealed also based on the results of a series
of tests. In addition, a suitable test method to evaluate the
tensile strength of rammed earth wall is proposed based on
these test results.
2. Tested materials
2.1. Mixtures H1 and H2
Three kinds of soils, called herein mixture H1, mixture H2
and mixture M, were used in this study. The mixtures H1 and
H2 consist of materials that were actually used for a
reconstruction project of rammed earth walls surrounding a
traditional temple in Japan. They were composed of a clayey
sand (ρs ¼ 2.648 g/cm3, Fc ¼ 48.4%, D50 ¼ 0.085 mm) and a
silty sand (ρs ¼ 2.666 g/cm3, Fc ¼ 19.2%, D50 ¼ 0.45 mm). The
rammed earth wall reconstructed in this project was con-
structed originally more than 300 years ago, while some part of
the wall was reconstructed later because of deterioration. The
clayey sand was made from the material consisting of the old
walls by crushing and sieving. The silty sand (called as “silty
sand A” hereafter) was a commercial soil by the name of
Nashime sand, which has been often used for construction of
earthen walls in Japan. The mixing ratio in terms of the dry
weight of the clayey sand to the silty sand A was set as 65% to
35%, referring to the mixing ratio used in the aforesaid
reconstruction project.
Magnesium chloride is often added to the water that adjusts
the water content of the rammed earth material before
compaction in the traditional rammed earth technique. Typi-
cally, a magnesium chloride solution with a concentration of
about 3% was used in the aforesaid reconstruction project. In
order to reveal the possible effects of magnesium chloride, in
this study, it was added only to mixture H2, while mixture H1
was prepared without it. The concentration of magnesium
chloride solution was set as 3%, which was defined as the
weight of the magnesium chloride divided by the weight of the
solution including the water contained in the soil.
The initial amount of water to be added to these mixtures
during preparation was set to satisfy the optimum water
content condition, which was evaluated by conducting mod-
ified Proctor compaction tests with a compaction energy, Ec of
2700 kJ/m3. The optimum water content and maximum dry
density of mixture H1 are 10.1% and 1.997 g/cm3, respec-
tively. Those of mixture H2 are 10.3% and 1.996 g/cm3,
respectively. The mixing proportions of base materials to
prepare mixtures H1 and H2 and the soil properties of mixtures
H1 and H2 are summarized in Table 1.
2.2. Mixture M
Mixture M was composed of another silty sand (ρs ¼ 2.670
g/cm3, Fc ¼ 32.8%, D50 ¼ 0.12 mm), a poorly graded sand
(ρs ¼ 2.812 g/cm3, Fc ¼ 0.2%, D50 ¼ 0.50 mm) and quick lime
(calcium oxide, CaO). The silty sand (called as “silty sand B”)
was retrieved from the site of a future construction project of
new rammed earth walls in Japan. The poorly graded sand was
made from commercially available poorly graded gravel by
crushing. The lime content ratio l, defined as the ratio of lime
to solid (i.e. silty sand B and poorly graded sand) by dry

Table 1
Mixing proportions and soil properties of the soil mixtures.

Mixture H1 Mixture H2 Mixture M

Mixing proportions in Clayey sand 59.0% 58.8% –

weight Silty sand A 31.8% 31.6% –
Silty sand B – – 61.8%
Poorly graded sand – – 10.1%
Quick lime – – 8.2%
Water 9.2% 9.3% 20.0%
Magnesium chloride (Anhydrous) – 0.3% –
Total 100.0% 100.0% 100.1%
Soil particle density, ρs 2.654 g/cm3 2.654 g/cm3 2.690 g/cm3
Grain size distribution Gravel (4 2.0 mm) 5.9% 5.9% 1.6%
Sand (0.075 mm to 2.0 mm) 55.9% 55.9% 70.2%
Silt and clay ( o0.075 mm) 38.2% 38.2% 28.2%
D50 0.269 mm 0.269 mm 0.146 mm
Suspended soil Electric conductivity (JGS 0212- 0.14 mS/cm 1.20 mS/cm 6.42 mS/cm
2009)
pH (JGS 0211-2009) 6.5 5.8 12.4
Soil classification (JGS 0051-2009) Fine-grainedy sand with gravel Fine-grainedy sand with gravel Fine-grainedy sand
(SF-G) (SF-G) (SF)
192 H. Araki et al. / Soils and Foundations 56 (2016) 189–204

weight was set as 11.4%, based on the results of preliminary
tests (Lee et al., 2009).
The molding water content, defined as the ratio of the initial
amount of water to the dry weight of solid (soils and lime),
was set as 24.1%, which ensures the maximum dry density of
1.615 g/cm3 by compaction with Ec of 2700 kJ/m3. The water
content after compaction was reduced to 20.5% approximately,
because of absorption by the hydration reaction and evapora-
tion during compaction. The mixing proportions of base
materials to prepare mixture M and the soil properties of
mixtures M are also shown in Table 1.
3. Test methods and conditions
3.1. Test Series
In this study, the tensile properties of the rammed earth
materials were evaluated by conducting both a splitting test
and a direct tension test. Fig. 1 shows schematically the test
apparatus for these tests. From the direct tension test results,
one can evaluate the tensile strength of the specimen directly,
while a specimen with a special shape, i.e. dog bone shape as
shown in Fig. 1a, needs to be employed to ensure the
distribution of tensile stress inside the specimen is as uniform
as possible (Yoshizawa et al., 2003; Koseki et al., 2008). On
the other hand, from the splitting test results, one can evaluate
the tensile strength indirectly, as will be discussed in detail
later. It should be noted, however, that the splitting test may
underestimate the tensile strength due to shear stress induced
partially inside the specimen near the loading platens
(Namikawa and Koseki, 2007).
In total twelve series of specimens as listed in Tables 2 and 3
were prepared in order to evaluate the tensile properties and to
reveal possible effects of several test conditions. All specimens
for direct tension test and splitting test are listed in Table 4 and 5,
respectively. The values of dry density and water content of the
specimens at the time of tests are also shown in Tables 4 and 5.
The degree of compaction, Dc are defined as the dry density

Fig. 1. Specimens and apparatus for (a) direct tension tests and (b) splitting test.

Table 2
Specimen series for direct tension tests.

DH1,g DH1 DH1u DH1,c DH1u,c DH2 DH2u DM

Mixture H1 H1 H1 H1 H1 H2 H2 M

Treatment of layer interface Scarified Scarified None Scarified None Scarified None Scarified

Mould type A A, B A A A A, B A B

Curing method I I I I I I I II

Curing time (days) 28 21-28 21-28 21, 23 24, 28 28 28 28

Load type Monotonic Monotonic Monotonic Cyclic Cyclic Monotonic Monotonic Monotonic

Bonding method Gypsum Epoxy resin Epoxy resin Epoxy resin Epoxy resin Epoxy resin Epoxy resin Epoxy resin
 H. Araki et al. / Soils and Foundations 56 (2016) 189–204 193

divided by the maximum dry density of each mixture by mixture was compacted by wooden tampers with a tip of 4 to
compaction with Ec of 2700 kJ/m3. The water content, defined 8 cm square. The thickness of each compaction layer was
as a weight of water divided by a weight of dry soil in this paper, around 6 cm, which was half the approximately 12 cm
were measured immediately after tests. spreading depth before compaction. The dry density of the
The direct tension test and splitting test were referred to reconstructed rammed earth wall after compaction was 1.95 to
Namikawa and Koseki (2007) and Koseki et al. (2008). JGS 2.07 g/cm3.
2551-2009 was also referred for the splitting test. It is to be In order to establish the specimen preparation condition, in
noted that JGS 2551-2009 is the test method for rocks, not this study, preliminary compaction tests were conducted for
soils. The details of test procedures and conditions will be mixture H1 by applying Ec of 600 kJ/m3 and 2700 kJ/m3. The
described below. values of the maximum dry density of mixture H1 by compac-
tion with Ec of 600 kJ/m3 and 2700 kJ/m3 were 1.861 g/cm3 and
3.2. Direct tension test 1.997 g/cm3, respectively. The maximum dry density with Ec of
2700 kJ/m3 was almost equal to that of the reconstructed
3.2.1. Specimens rammed earth wall described above. Moreover, the thickness
According to the report on the aforesaid reconstruction of a compaction layer with Ec of 2700 kJ/m3 was around half of
project (Nara Prefectural Board of Education, 2009), the soil that of spreading depth in the preliminary test. Therefore, the
compaction energy to prepare specimens was set as 2700 kJ/m3
Table 3 in this study. Although the thickness of each compaction layer
Specimen series for splitting tests. was around 6 cm in the prototype wall, the thickness of each
compaction layer in the specimen was set at 10 mm in order to
SH1 SH2 SMa SMb
ensure homogeneous compaction and to minimize dispersion.
Mixture H1 H2 M M The specimen series for direct tension tests are shown in
Treatment of layer interface Scarified Scarified Scarified Scarified Table 2. Detailed data for all specimens for direct tension test
Mould type C C C C is listed in Table 4.
Curing method I I III IV
The shape and dimensions of the specimen were cylindrical
Curing time (days) 7-245 7-65 28 28
with a diameter of 50 mm and a height of 140 mm, while the

Table 4
List of specimens for direct tension test.

Series No. Mixture Mould Curing method Curing days Dry density [g/cm3] Dc [%] Water content [%] Tfd [kPa]

DH1,g 02 H1 A I 28 1.938 97.0 1.71 75.4

DH1,g 03 H1 A I 28 1.982 99.2 1.75 113.8
DH1,g 04 H1 A I 28 1.936 96.9 1.79 96.3
DH1 11 H1 A I 28 1.960 98.1 1.58 314.0
DH1 12 H1 A I 28 1.944 97.3 1.55 295.3
DH1 14 H1 A I 27 1.974 98.8 1.54 338.2
DH1 15 H1 A I 21 1.982 99.2 1.11 402.6
DH1 16 H1 A I 23 1.995 99.9 1.23 380.8
DH1 31 H1 B I 28 1.980 99.1 1.86 275.9
DH1 32 H1 B I 28 1.987 99.5 1.85 258.5
DH1u 21 H1 A I 28 1.939 97.1 1.15 280.4
DH1u 22 H1 A I 28 1.941 97.2 1.18 313.8
DH1u 23 H1 A I 28 1.973 98.8 1.13 264.3
DH1u 25 H1 A I 21 1.983 99.3 1.14 371.0
DH1u 26 H1 A I 23 1.994 99.8 1.24 401.0
DH1,c 17 H1 A I 21 1.974 98.8 1.22 420.2
DH1,c 18 H1 A I 23 1.978 99.0 1.36 422.3
DH1u,c 24 H1 A I 28 1.966 98.4 1.14 279.4
DH1u,c 27 H1 A I 24 1.992 99.7 1.39 322.5
DH2 51 H2 A I 28 1.960 98.2 1.49 287.8
DH2 52 H2 A I 28 1.951 97.7 1.49 257.1
DH2 53 H2 A I 28 1.989 99.6 1.69 327.2
DH2 71 H2 B I 28 1.980 99.2 2.34 201.2
DH2 72 H2 B I 28 1.993 99.8 2.31 187.4
DH2u 61 H2 A I 28 1.958 98.1 1.49 265.0
DH2u 62 H2 A I 28 1.994 99.9 1.62 324.0
DH2u 63 H2 A I 28 2.000 100.2 1.63 336.6
DM 91 M B II 28 1.580 97.8 6.91 522.0
DM 92 M B II 28 1.575 97.5 8.67 373.2
DM 93 M B II 28 1.577 97.6 9.58 408.7
194 H. Araki et al. / Soils and Foundations 56 (2016) 189–204

Table 5
List of specimens for splitting test.

Series No. Mixture Curing method Curing days Dry density [g/cm3] Dc [%] Water content [%] Tfs [kPa]

SH1 1-1 H1 I 14 1.913 95.8 3.72 179.3

SH1 1-2 H1 I 14 1.954 97.8 4.62 166.3
SH1 1-3 H1 I 14 1.910 95.6 3.81 184.0
SH1 1-4 H1 I 14 1.952 97.7 4.74 164.1
SH1 1-5 H1 I 28 1.916 95.9 2.82 219.6
SH1 1-6 H1 I 28 1.936 96.9 3.66 195.1
SH1 1-7 H1 I 28 1.902 95.2 2.82 186.2
SH1 1-8 H1 I 28 1.930 96.6 3.63 202.8
SH1 1-9 H1 I 56 1.911 95.7 1.81 248.5
SH1 1-10 H1 I 56 1.929 96.6 2.40 206.7
SH1 1-11 H1 I 56 1.949 97.6 2.77 228.0
SH1 1-12 H1 I 56 1.902 95.2 2.29 262.9
SH1 1-13 H1 I 56 1.934 96.8 2.77 185.0
SH1 1-14 H1 I 113 1.905 95.4 1.95 243.9
SH1 1-15 H1 I 113 1.939 97.1 2.14 251.0
SH1 1-16 H1 I 113 1.911 95.7 2.20 266.5
SH1 1-17 H1 I 113 1.946 97.4 2.24 267.0
SH1 1-18 H1 I 8 1.953 97.8 1.58 266.2
SH1 1-19 H1 I 8 1.950 97.6 1.58 217.7
SH1 1-20 H1 I 7 1.965 98.4 1.65 325.7
SH1 1-21 H1 I 7 1.981 99.2 1.88 360.9
SH1 1-22 H1 I 245 1.931 96.7 1.40 355.5
SH1 1-23 H1 I 245 1.996 99.9 1.43 425.4
SH2 2-1 H2 I 28 1.999 100.2 1.40 506.9
SH2 2-2 H2 I 28 2.005 100.5 1.40 495.8
SH2 2-3 H2 I 28 1.962 98.3 1.57 340.4
SH2 2-4 H2 I 28 1.975 98.9 1.38 384.8
SH2 2-5 H2 I 7 2.003 100.4 2.78 321.8
SH2 2-6 H2 I 7 1.960 98.2 3.00 236.2
SH2 2-7 H2 I 7 1.993 99.8 3.01 226.7
SH2 2-8 H2 I 14 2.010 100.7 2.62 325.7
SH2 2-9 H2 I 14 2.045 102.5 2.64 334.8
SH2 2-10 H2 I 14 2.024 101.4 2.63 324.1
SH2 2-11 H2 I 65 2.008 100.6 1.83 362.9
SMa 3-1 M III 28 1.577 97.6 6.74 377.8
SMa 3-2 M III 28 1.641 101.6 9.42 483.3
SMa 3-4 M III 28 1.565 96.9 5.90 434.4
SMa 3-5 M III 28 1.571 97.3 5.99 446.5
SMa 3-6 M III 28 1.555 96.3 6.44 221.9
SMa 3-7 M III 28 1.620 100.3 9.42 443.2
SMb 3-8 M IV 28 1.572 97.3 19.59 356.2
SMb 3-9 M IV 28 1.579 97.8 19.87 405.6
SMb 3-10 M IV 28 1.571 97.3 20.23 388.2

diameter at the middle height was trimmed down to 45 mm
(Fig. 1a). This shape was adopted based on the results from
finite element analyses in order to avoid failure at the fixed
portions in which the tensile stress may be concentrated
(Yoshizawa et al., 2003, Koseki et al., 2008).
For specimen preparation, two types of split moulds, called
mould A and B, were used. Mould A is cylindrical with a
diameter of 50 mm and a height of 140 mm. The specimens were
compacted into 14 layers in mould A by applying the compaction
energy of 2700 kJ/m3, while the weight and the volume of the
mixture are controlled for each layer. The required number of
blows was determined based on a weight and a dropping height
of a rammer, which were 508 g and 25 cm, respectively. For
instance, the number of blows for each layer with a thickness of
10 mm and a diameter of 50 mm was set at 43. The radius of a
compaction zone of the rammer was 13 mm. The specimens were
removed from the mould and were trimmed down the middle part
by a knife immediately after compaction.
Mould B is shown in Fig. 2. It is shaped like a dog-bone with a
diameter of 50 mm and a height of 160 mm, where the diameter at
the middle height was reduced to 45 mm. A schematic on the
making procedure of a specimen using mould B is shown in Fig.
2a. In order to ensure well-compaction at the inflected portion of
the mould, after setting a spacer into the mould to replace the upper
part of the specimen (Fig. 2a, i), its lower part was compacted (Fig.
2a, ii). Then, the mould including the lower part of the specimen
was turned upside down (Fig. 2a, iii), and its upper part was
compacted (Fig. 2a, iv). A total of 16 compaction layers were
made in the specimen inside mould B. The compaction energy was
2700 kJ/m3, which was applied as described above. The specimen
 H. Araki et al. / Soils and Foundations 56 (2016) 189–204
Fig. 2. (a) Schematics on making procedure of a specimen by using mould B and (b) photographs for the processes of (i) and (iv).
height was trimmed to 140 mm after demoulding. The specimens
made in the mould B were No.31 and No.32 in series DH1, No.71
and No.72 in series DH2 and all in series DM. Except for these
specimens, designed for thedirect tension test, all specimens were
made in mould A (Table 4).
Before making the next layer, the surface of each compacted
layer was scarified thoroughly to a depth of about 3 mm by a
minus driver in order to strengthen the interface of these layers,
except for the specimens of series DH1u and DH2u which
were prepared without scarifying. The specimen series DH1u
and DH2u simulates the conditions which were adopted in the
Japanese traditional method of making the rammed earth walls.
The specimens made of mixtures H1 and H2 were cured at
ambient temperature and relative humidity conditions in a
room without sealing until the test day. The room temperature
varied seasonally in the range of about 10–30 1C. The relative
humidity in the room was in the range of 60 to 70%RH at the
curing period for specimens No.31 and 32 in series DH1 and
No.71 and 72 in series DH2. Except for the curing period for
the above specimens, the relative humidity in the room was in
the range of 20 to 40%RH at the curing period for specimens
made of mixtures H1 and H2. This curing method is called
curing method I. The curing time was between 21 and 28 days.
The specimens made of mixture M were cured at ambient
temperature and relative humidity conditions in the room in the
mould for initial 14 days. The relative humidity was in the range of
60 to 70%RH. Then, after demoulding, they were wrapped by a
sheet of plastic film and were cured under the room temperature
and the relative humidity for 14 days. Thus, the total curing time
was 28 days. This curing method is called curing method II.
3.2.2. Test method
The apparatus shown in Fig. 1a was used. In the direct
tension test, the specimens were directly supported by holders,
which were attached to the top cap and the pedestal. In order to
reduce the bending moment applied unnecessarily to the
specimen, a pair of universal joints was inserted on both ends.
In this study, gypsum and epoxy resin (SCANDIPLEX,
produced by SCAN-DIA Hans P. Tempelmann Gmbh & Co.
195
KG) were used as an adhesive material to bond the specimen
and the holders (Table 2).
The axial displacement was measured by a pair of external
displacement transduces set at an opposed position around the
loading shaft, as schematically shown in Fig. 1a, and a pair of
LDTs (Local Deformation Transducers, Goto et al., 1991)
attached directly on the side of the specimens at their middle
height with reduced diameter. A tensile load was applied to the
specimen at an axial strain rate of 0.005%/min, so that the peak
tensile strength was measured accurately, referring to Koseki et
al. (2008).
The tensile stress std that was mobilized in the specimen
was evaluated by:
Pd
std ¼ ð1Þ
A0
where Pd is the tensile force applied to the specimen; and A0 is
the cross-sectional area of the specimen at the middle height.
The peak value of the tensile stress is employed as the tensile
strength, Tfd, by the direct tension test.
The specimens in series DH1,c and DH1u,c, as listed in
Tables 2 and 4, were subjected to cyclic loading, while referring
to plane strain cyclic compression tests conducted by Koseki et
al. (2005). Cyclic loading was conducted in order to evaluate the
behavior of the rammed earth structure during earthquake and
the strength of rammed earth structure subjected to seismic
force. A typical stress strain behavior of the cyclic loading test is
shown in Fig. 3. In Fig. 3, the initial and final portions (a-b and
c-d) are monotonic tensile loading, and the middle portion (b-c)
is cyclic loading. The value of the initial tensile stress, q0 of
specimen series DH1,c and DH1u,c was set at approximately
50% of Tfd, which was estimated based on the results of the
specimen series DH1 and DH1u, respectively. The single-
amplitude value of the cyclic stress amplitude, qd was set at
approximately 80 to 85% of q0. The number of cyclic loading
was 20. The apparatus shown in Fig. 1a was used also in cyclic
loading. The strain rate was 0.005%/min at the monotonic
loading portions of a–b and c–d, and it was increased to 0.02%/
min at the cyclic loading portion of b–c.
196 H. Araki et al. / Soils and Foundations 56 (2016) 189–204
Fig. 3. Typical stress strain relationship in series DH1,c.
3.3. Splitting test
The specimen series for splitting test are shown in Table 3.
The detail data of all specimens for splitting test are listed
in Table 5.
A split mould with an inner diameter of 50 mm (denoted as
mould C) was used for making a cylindrical specimen with a
diameter of 50 mm and a height of 50 mm. Specimens were
compacted in the mould into five layers by employing compaction
energy of 2700 kJ/m3. All interfaces of the compaction layers were
scarified by a minus driver. The compacting and scarifying manner
was similar to that employed in preparing the specimens for direct
tension tests (i.e., series DH1, DH2 and DM as listed in Table 2).
The curing method I as described before was employed on
the specimens made of mixtures H1 and H2. The curing time
was between 7 and 245 days. The relative humidity and the
temperature in the curing room varied seasonally in the range
of 20 and 70%RH and 10 to 30 1C, respectively.
The specimens made of mixture M were cured by two curing
methods, called curing methods III and IV. In curing method III, the
specimens in the moulds were cured for 28 days under the same
ambient temperature and relative humidity conditions that were
employed in curing method I. In curing method IV, the specimens
in the moulds were cured inside desiccators with a saturated
solution of potassium sulfate (K2SO4) to keep the relative humidity
at 98%RH under a constant temperature of 16 1C for 28 days.
In conducting the splitting test, the apparatus shown in Fig.
1b was used. Specimens were placed horizontally on a pedestal
and loaded in the vertical direction at a rate of 0.025 mm/min
referring to Koseki et al (2008). The tensile stress sts was
evaluated based on the linear elasticity theory by:
2Ps
sts ¼
π Ut UD
where Ps is the compression force applied to the specimen;
t is the width of the contact area between the specimen and
the loading plates (¼ 50 mm); and D is the diameter of
the specimen ( ¼ 50 mm). The peak value of the tensile
stress is employed as the tensile strength, Tfs, by the
splitting test.
Fig. 4. Typical stress strain relationships in series SH1.
4. Test results
4.1. Splitting tests
The values of Tfs and water content of specimens immedi-
ately after tests are shown in Table 5. Fig. 4 shows the stress
strain relationships of the specimens with different curing time
in series SH1 (Tables 3 and 5), which were made by using
mixture H1 without magnesium chloride. The tensile stress is
calculated with Eq. (2). The strain shown in Fig. 4 is not
tensile strain but average compressive strain, which is calcu-
lated by dividing the vertical displacement of the upper loading
plate, which was measured with the external displacement
transducer (Fig. 1b), with the diameter of specimen (50 mm). It
can be seen that the tensile strength increased with the increase
in the curing time, while the effect of the curing time on the
average strain mobilized at the peak stress state was not
significant (Fig. 4). Photograph 1a shows typical condition of a
specimen after the peak stress state.
Fig. 5 shows the relationship between Tfs and the water
content of the specimens in series SH1. For comparison, test
results on the specimens in series SH2, which were made by
using mixture H2 with magnesium chloride are also shown in
the figure. The water contents were measured immediately
after the tests. The values of water contents in series SH1 and
SH2 were in the ranges of 1.5–4.5% and 1.5–2.5%, respec-
tively. The values of Tfs gradually increased with the decrease
in the water content. The relationship between Tfs and the
water content of the specimens in series SH2 was similar to
that of the specimens in series SH1.
It should be noted that the above change of the water content
depends on the curing time and the relative humidity. The
relative humidity in the curing room was not controlled, which
ð2Þ varied seasonally in the range of 20 and 70%RH. The test
results shown in Fig. 5 were obtained on specimens cured for
different time and under different relative humidity, while
rather a unique relationship between Tfs and w could be found.
This behavior will be further analyzed in Section 5.
Fig. 6 shows the relationships between Tfs and w of the
specimens in series SMa and SMb at 28 day-curing with
different curing methods. The specimens in series SMb were
 H. Araki et al. / Soils and Foundations 56 (2016) 189–204 197

Photograph 1. Typical condition of specimens after (a) splitting test, (b) direct tension test by using epoxy resin and (c) direct tension test by using gypsum.

Fig. 5. Relationships between Tfs and water content of specimens with Fig. 6. Relationships between Tfs and water content of specimens in series
different curing time in series SH1 and SH2. SMa and SMb.

cured in a desiccator where the relative humidity was main-

tained at 98%RH, while those in series SMa were cured
without such control of the relative humidity, as mentioned
before. As a result, the values of the water content of the
specimens in series SMa and SMb were around 8% and 20%,
respectively. However, the values of Tfs in series SMa and
SMb were not largely different from each other; they were in
the ranges of 350 to 420 kPa mainly and 370 to 470 kPa,
respectively. This behavior will be also further analyzed
in Section 5.

4.2. Direct tension test

The values of Tfd and water content of specimens immedi-
ately after tests are shown in Table 4. The stress strain
relationships obtained during the monotonic loading in the
direct tension tests are shown in Figs. 7–10. Figs. 7–10 show
the stress strain relationships of the specimens in series DH1,g,
DH1 and DH1u, DH2 and DH2u and DM, respectively. The
numbers shown in the legend of these figures indicate the test
names of specimens. The tensile stress is calculated with
Fig. 7. Stress strain relationships of specimens in series DH1,g.
Eq. (1) based on tensile load and a cross-sectional area of the
specimen at the middle height. The tensile strain is evaluated
as an average of two values measured by a pair of LDTs.
Photograph 1b shows typical condition of a specimen after the
peak stress state.
As for the test results in series DH1 and DH1u (Fig. 8), the
failure strains mobilized at the peak stress state, εf ranged from
198 H. Araki et al. / Soils and Foundations 56 (2016) 189–204

Fig. 8. Stress strain relationships of specimens in (a) series DH1 and (b) DH1u.

Table 6
Specimens in series DH1,c and DH1u,c for cyclic direct tension test.

Specimen No. Curing Water Estimated q0 (kPa) qd (kPa) Number

days content Tfd (kPa) of cycles
(%)

DH1,c No.17 21 1.22 400 200 170 20

DH1,c No.18 23 1.36 400 200 170 20
DH1u,c No.24 28 1.14 280 140 110 20
DH1u,c No.27 24 1.39 280 140 110 20

shapes of these stress strain relationships in series DM are

Fig. 9. Stress strain relationships of specimens in series DH2 and DH2u.
Fig. 10. Stress strain relationships of specimens in series DM.
0.015% to 0.025%, and the values of Tfd ranged from 250 to
400 kPa. Similarly, the values of Tfd and εf in series DH2 and
DH2u (Fig. 9) ranged from 200 to 330 kPa and from 0.010%
to 0.030%, respectively. The shapes of these stress strain
relationships were also similar to each other.
The values of Tfd and εf in series DM (Fig. 10) ranged from
360 to 530 kPa and from 0.005% to 0.025%, respectively. The
similar to those in series DH1 and DH2.
Fig. 7 shows the stress strain relationships of specimens in
series DH1,g, in which the gypsum slurry was employed for
bonding. The values of Tfd and εf are significantly smaller than
those in series DH1 (Fig. 8a), where the specimens were prepared
using epoxy resin for bonding. It must be noted that these
specimens in series DH1,g and DH1 were prepared with the
mould A, except for No.31 and No.32 in series DH1 (Table 4).
In series DH1,g, all specimens exhibited tensile failure at the
regions close to the gypsum as shown in Photo. 1c.
In the tests with cyclic loading, large amplitude tensile
cyclic load was applied to the specimens in series DH1,c and
DH1u,c before reaching the peak stress state. In setting the
cyclic loading conditions as listed in Table 6, the values of Tfd
in these series DH1,c and DH1u,c were estimated to be
400 kPa and 280 kPa, respectively, which were based on the
corresponding monotonic loading test results on specimens
having similar water contents. The stress strain relationships in
these series are shown in Fig. 11. Possible effects of cyclic
loading history will be further analyzed in Section 5.
Fig. 12a summarizes the location of the failure plane observed
in the direct tension tests on specimens using epoxy resin for
bonding, where the parameter L is defined as a distance from the
bottom of the first compaction layer to the observed failure plane
as typically shown in Fig. 12b. The first compaction layer is
located at the bottom of the specimen that was compacted using
mould A. On the other hand, the specimen that was compacted
 H. Araki et al. / Soils and Foundations 56 (2016) 189–204
Fig. 11. Stress strain relationships of specimens in (a) series DH1,c and (b) DH1u,c.
Fig. 12. (a) Relationships between Tfd and L and (b) measurement positions of L.
using mould B (i.e. all specimens in series DM and some
specimens in series DH1 and DH2 as listed in Table 2) has the
first compaction layer at its middle height as shown in Fig. 12b
due to the special specimen preparation method as described
before. The broken lines plotted in Fig. 12a indicate the location
of layer interfaces that were estimated based on the test condition
where the specimens were compacted to a thickness of 10 mm.
Most of the test results were plotted near the broken lines,
suggesting that the tensile failure occurred near the layer
interfaces of the compacted specimens.
5. Discussions
5.1. Relationship between tensile strength, water content and
curing time
The water content in series SH1 and SH2 shown in Fig. 5 were
obtained for the specimens cured for different times and under
different relative humidities in the room. The relative humidity in
the curing room varied seasonally between 20 to 70%RH. In
series SH1, the water content of the specimens with the curing
199
time of 14 to 245 days decreased with increasing curing time. On
the other hand, the water content of specimens at 7 day-curing
were around 1.5%, which was in the same range of the water
content of specimens at 245 day-curing. The curing time to reach
a certain water content possibly depended on not only the curing
time but also the relative humidity.
For example, the values of water content of the specimens in
series SH1 at 7 and 8 day-curing were smaller than those of
specimens at 14 and 28 day-curing because the values of
ambient relative humidity at these tests for the specimens at
7 and 8 day-curing and 14 and 28 day-curing were 30%RH
and 40%RH, respectively. In addition, the values of Tfs of the
specimens in series SH1 at 7 and 8 day-curing and 245 day-
curing were in the same range because the values of ambient
relative humidity at these tests and water content of the
specimen were similar, at around 30%RH and 1.5%, respec-
tively. These comparisons suggest that the effect of curing time
was not a predominant factor in the change of the tensile
strength of the mixtures H1 an H2.
The values of Tfs in series SH1 and SH2 increased with the
decrease in the water content as mentioned previously (Fig. 5).
A similar trend has been found in the results of unconfined
200 H. Araki et al. / Soils and Foundations 56 (2016) 189–204
Fig. 13. Relationships between qu and water content for mixture H1 and H2.
Fig. 14. Relationships between Tfs and Tfd and water content of specimens
made of mixture H1.
compression tests for mixtures H1 and H2. The relationships
between the unconfined compression strength, qu and the water
content are shown in Fig. 13. The values of qu of mixtures H1
and H2 increased with the decrease in the water content, which
can be approximated by the formulas shown in Fig. 13. Based
on these results, the tensile and compressive strength of
mixture H1 and H2 would affect by the slightly change of
water content with several percent.
Figs. 14 and 15 show the relationships between the tensile
strengths, Tfs and Tfd, and the water content of mixtures H1 and
H2, respectively. The broken curves plotted in Figs. 14 and 15
were derived based on the formulas shown in Fig. 13. By
comparing these curves with the test results, it can be seen that
the tensile strengths of mixtures H1 and H2 were in the range of
7.5–12.5% and 5.0–10.0% of the corresponding values of qu,
respectively, except for the test results in series DH1,g.
Jaquin et al. (2009) and Bui et al. (2014) showed that the
values of qu of compacted rammed earth materials without
add-in material increased with the increase in soil suction; the
soil suction increased with the decrease in water content. Bui
et al. (2014) also showed that the value of qu of compacted
rammed earth materials with add-in material increased with the
decrease in water content in a case where an amount of add-in
Fig. 15. Relationships between Tfs and Tfd and water content of specimens
made of mixture H2.
material, natural hydraulic lime, was small. They pointed out
that the compacted soil became few sensitive to water if an
amount of the hydraulic binders formed by lime was sufficient
to coat all grains of the soil, because the hydraulic binders to
resist compressive stress are not sensitive to water. In other
words, the strength of the compacted soil increased with the
increase in soil suction induce by the decrease in water
content, except when the compacted soil contained rich
artificial cementation derived from lime or cement.
For the material H1 and H2 as mentioned above, artificial
cementation would not be formed in the specimen because add-
in material was not contained, and the curing time did not affect
predominantly for the tensile strength. The increase of tensile
strengths of the specimens made of the mixtures H1 and H2 can
be attributed to the increase in soil suction that the decrease in
water content would induce, as well as the compressive strength
studied by Jaquin et al. (2009) and Bui et al. (2014), while the
value of soil suction acting on a specimen was not measured in
this study. Further investigation is required to verify the possible
relationship between the strength properties and suction, while
evaluating the suction in the specimen.
Fig. 16 shows the relationship between the tensile strengths,
Tfs and Tfd, and the water content of specimens made of
mixture M that were cured for 28 days. As mentioned
previously, the values of Tfs in series SMa and SMb were
not much different from each other, although the value of
water content of specimens in series SMb was nearly double
that of the water content in series SMa. Thus, the tensile
strength of mixture M is not significantly affected by the water
content, due possibly to the more predominant effect of
cementation formed by the lime content, which would be
rather affected by the curing time. This trend is similar to the
result reported by Bui et al. (2014) which the value of qu of a
compacted soil hardly increased with the decrease in water
content in case of the compacted soil with rich hydraulic lime.
In relation to the above discussion, Lee et al. (2009) showed
that the qu values of mixture M increased with the increase in
the curing time. They also showed that the average value of qu
of 28 days-cured specimens, qu28 is 2465 kPa. Dashed lines in
Fig. 16 were drawn based on this result, and the values of Tfs
 H. Araki et al. / Soils and Foundations 56 (2016) 189–204
Fig. 16. Relationships between Tfs and Tfd and water content of specimens
made of mixture M.
and Tfd of mixture M were found to be in-between 15% and
20% of the corresponding value of qu28.
5.2. Effect of bonding method
As shown in Fig. 14, the values of Tfd of the specimens in
series DH1,g using gypsum slurry for bonding were signifi-
cantly smaller than those of the specimens in series DH1 using
epoxy resin for bonding.
The above underestimation of the tensile strength in series
DH1,g, may be related to the water contained in the gypsum
slurry, which may have moved into the neighboring sections of
the specimens, causing a local increase of the water content
that initiated earlier tensile failure there. This explanation is
supported by the fact that the tensile failure of all the
specimens in series DH1,g occurred at the bottom or top part
of the specimen which was close to the gypsum. In conducting
direct tension tests under such circumferences, the epoxy resin
which would not affect the water content of the specimen
should be used to bond the specimen with holders, especially
for the specimen made of the soil mixture without lime or
cement.
5.3. Effect of mechanical damage of specimens
The results of the direct tension tests on specimens made of
mixture M conducted by Lee et al. (2010) are also plotted in
Fig. 16. The mixing ratio, compaction energy and curing time
employed are the same as those employed for series DM in this
study. These specimens were compacted inside mould A, and
then were trimmed by a machine while pouring water after
curing for several days. These preparation processes had to be
used because the specimens were otherwise too weak to stand
by themselves just after compaction, and were too hard to be
trimmed down by a knife even after curing for several days.
However, these processes might have caused an increase in the
water content and mechanical damage to the specimens. In
addition, gypsum slurry was used for bonding.
On the other hand, the specimens in series DM in this study
were made inside mould B without trimming, were not
201
trimmed down in the later stages, and were bonded to holders
using epoxy resin. Thus, local change of water content and
mechanical damage of the specimens by trimming might not
have been induced in the specimens in series DM.
As discussed above, the change of water content did not
significantly affect the tensile strength of mixture M (Tfs in Fig.
16). Therefore, mechanical damage of the specimens by
trimming would have been the major reason why Tfd of
mixture M was underestimated in the tests by Lee et al.
(2010). In order to avoid such underestimation, specimens of a
rammed earth material like mixture M for direct tension test
should be made inside a special mould like mould B.
5.4. Effect of layer interface condition
The specimens in series DH1u and DH2u were prepared
referring to the Japanese traditional method to construct
rammed earth walls. At similar water contents of around
1.2%, though the values of Tfd in series DH1u scattered
widely, they were in general smaller than those in series DH1
(Fig. 14). A similar trend was observed also in the results of
series DH1,c and DH1u,c at similar water contents of around
1.4%. On the other hand, the values of Tfd in series DH2u were
approximately similar to those in series DH2 (Fig. 15).
As shown in Fig. 12, the tensile failure of most of the
specimens took place at the interfaces between the neighboring
compaction layers, which may suggest that these interfaces
were weaker even if the surface of each compacted layer was
scarified before making the next layer. In addition, the effect of
increase of strength by scarifying was not clearly found among
the specimens in series DH2 and DH2u, which were prepared
with and without scarifying, respectively (Fig. 15). However,
the comparison between the test results in series DH1 and
DH1u as shown in Fig. 14 suggests it is better to apply
scarifying in order to prevent reduction and variation of the
tensile strength at the layer interfaces when rammed earth
structures are constructed.
5.5. Effect of test method
Namikawa and Koseki (2007) conducted a series of experi-
ments and their numerical simulation on tensile strength of a
cement-treated sand. The numerical simulation showed that the
value of the tensile strength in the splitting test was smaller
than that in the direct tension tests because local shear failure
occurred by large deviator stresses concentrated below the
loading plate in splitting tests (i.e. Tfd 4Tfs). They showed also
that the value of the direct tension strength evaluated by the
numerical simulation was almost equal to the assigned value of
the actual tensile strength.
In this study, the test results on specimens made of mixtures
H1 and M showed that the value of Tfd was in general equal to
that of Tfs (Figs. 14 and 16). In addition, the value of Tfd was
smaller than that of Tfs in case of specimens made of mixture
H2 (Fig. 15). These trends are not consistent with the
explanation made by Namikawa and Koseki (2007).
202 H. Araki et al. / Soils and Foundations 56 (2016) 189–204
Fig. 17 shows schematically the direction of tensile stress
and that of the layer interfaces of the specimens for splitting
and tensile tests. The directions of tensile stress in the direct
tension test and the splitting test were perpendicular and
parallel to those of the layer interfaces, respectively.
The actual tensile strengths inside the compacted layers and
at the layer interfaces are denoted as Tf and Tf0 herein,
respectively. The fact that the tensile failure was observed at
the layer interfaces in the direct tension test implies that the
value of Tf0 was smaller than that of Tf in this study. The
insufficient compaction at the layer interfaces would be the
major reason why the value of Tf0 was smaller than that of Tf,
even if layer interfaces were scarified when the specimens
were prepared. The specimens in this study have the con-
texture of the alternate layers with the tensile strengths of Tf
and Tf0 . The contexture of alternate layers like the specimens
in this study might be also formed inside the rammed earth
walls made by following traditional rammed earth technique.
According to the numerical simulation by Namikawa and
Koseki (2007), the value of Tfd is almost equal to the value of
the actual tensile strength at the tensile failure section. In this
study, therefore, the value of Tfd in direct tension tests is
almost equal to the value of Tf0 because the tensile failure of
most of the specimens took place at the interfaces as shown in
Fig. 12. In splitting tests, on the other hand, the failure plane
was perpendicular to layer interfaces as shown in Photo. 1a.
The strength of the layer interface would hardly affect the
evaluation of Tfs because the tensile stress is mostly mobilized
inside the compacted layers. In this study, therefore, the value
of Tfs was related to not the value of Tf0 but the value of Tf. In
addition, it should be noted that the value of Tfs was possibly
smaller than the value of Tf as shown in the aforesaid
numerical simulation conducted by Namikawa and Koseki
(2007). In case of the specimens made of mixtures H1 and M,
the value of Tfd was incidentally in the same range as the
values of Tfs at the same water content (Figs. 14 and 16).
The difference between the values of Tf and Tf0 of the
specimens employed in this study was possibly larger than that
of the cement-treated sand specimens employed in the previous
relevant studies. In case of the specimens made of mixture H2,
the value of Tf0 was significantly smaller than that of Tf, which
Fig. 17. Schematic on relationships between layer interfaces and directions of
tensile stress mobilized in specimens.
was possible reason why the value of Tfd was smaller than that
of Tfs (Fig. 15). The possible effect of adding magnesium
chloride will be discussed later.
The tensile strength of the weakest part of the rammed earth
specimens, i.e. the strength at the layer interfaces, should be
measured in order to evaluate the seismic performance of a
whole rammed earth wall. For this purpose, the direct tension
test should be conducted. The splitting test might over-estimate
the tensile strength in case of specimens of a rammed earth
material like mixture H2 in this study.
5.6. Effect of large amplitude tensile cyclic loading
Fig. 11 shows the stress strain relationships of the specimens
in series DH1,c and DH1u,c. The specimens behaved rather
elastically during the large amplitude cyclic loading, exhibiting
limited amount of strain accumulation. During the subsequent
monotonic loading after the cyclic loading stage, however, the
strain started to increase again until the failure strain mobilized
at the peak strength in the test results in series DH1,c and
DH1u,c.
In Fig. 14, the peak tensile strength that was achieved during
the monotonic loading after cyclic loading was compared to
those of the specimens subjected to monotonic loading only.
The values of Tfd of specimens in series DH1,c with cyclic
loading were in the same range of those of the specimens in
series DH1 without cyclic loading at the same range of the
water content. A similar trend was also observed in the
comparison between the values of Tfd of the specimens in
series DH1u,c and DH1u. In addition, the value of εf in series
DH1,c and DH1u,c was in the same range of the value in series
DH1 and DH1u, respectively (Figs. 8 and 11).
In this test condition, therefore, the influence of the cyclic
loading history was found to be insignificant. The rammed
earth structure would be recovered elastically after unloading
unless the tensile stress mobilized at a layer interface exceeds
Tfd evaluated by a monotonic loading test.
In addition, the values of Tfd in series DH1u,c were also
smaller than those in series DH1,c as mentioned above. This
comparison also suggests that scarifying the layer interface
should not be skipped in order to prevent reduction of the Tfd
at the construction of rammed earth structures.
5.7. Comparison between specimens with/without magnesium
chloride
The values of maximum dry density of mixtures H1 and H2
were 1.997 g/cm3 and 1.996 g/cm3, respectively, as described
previously. The values of optimum water content of the
mixtures H1 and H2 were also in the same range. Therefore,
the compaction characteristics of the rammed earth material
were hardly affected by adding the magnesium chloride.
Fig. 5 shows that the value of Tfs of mixture H2 with
magnesium chloride was slightly larger than that of mixture
H1 without magnesium chloride at the same water content
level. A similar trend was observed in the relationships
between the value of qu and the water content, as shown in
 H. Araki et al. / Soils and Foundations 56 (2016) 189–204
Fig. 13. Addition of magnesium chloride resulted in such
limited increase in these strength properties.
On the other hand, the value of Tfd of series DH2 was
slightly smaller than that of series DH1 at the same water
content level, as shown in Figs. 14 and 15. The value of Tfd
corresponds to the strength of the layer interfaces discussed
above. The difference between the values of Tfd of specimens
in series DH1 and DH2 implied that adding magnesium
chloride affects the tensile strength of the interfaces in a
negative manner. In this study, therefore, the strength of the
interfaces decreased by adding magnesium chloride when the
water content was in the same range.
It should be noted that the above comparisons were made
between the test results of the specimens at the same water
content level, without quantitatively evaluating the possible
effects of soil suction. Further study is required to reveal the
reason why the strength at layer interfaces decreases by adding
magnesium chloride.
Although the effect of the magnesium chloride on the
strength properties of the rammed earth specimens was slightly
negative, it might have other roles in improving the physical
properties of a rammed earth wall. For instance, the water
retaining capability of soil increases by adding the magnesium
chloride (Tanimoto et al., 2004). The falling down of a wall
surface due to hard-drying might be relieved by adding the
magnesium chloride because the water in the wall is less-easily
evaporated. Moreover, the freezing point of magnesium
chloride solution is lower than that of pure water due to
freezing point depression. Thus, the deterioration of rammed
earth walls caused by freezing and thawing might be also
prevented by using the magnesium chloride solution.
6. Conclusions
In this study, direct tension tests and splitting tests on three
kinds of rammed earth materials were conducted in order to
evaluate tensile strength properties. The following conclusions
could be derived from the results of the tests.
1) The tensile strength of mixtures H1 and H2 increased with the
decrease in their water contents, possibly due to the increase of
suction. The tensile strengths of mixtures H1 and H2 were in
the ranges of 7.5–12.5% and 5.0–10.0% of the corresponding
unconfined compression strength, respectively.
2) The direct tension test using gypsum slurry for bonding
might have under-estimated the strength of unsaturated
specimens like mixture H1 because the gypsum slurry
increases locally the water content inside the specimen.
Epoxy resin should be used as the bonding material to
connect specimens with holders to prevent the change of the
water content of specimens.
3) In order to avoid underestimation of strength by mechanical
damage of specimen, specimens of a material like mixture
M should be made inside a special mould, like mould B for
the direct tension test.
4) The effect of increase in strength by scarifying at the layer
interfaces was not clarified. However, it is better to scarify
203
layer interfaces rather than not to scarify in order to prevent
reduction and variation of the tensile strength at the layer
interfaces.
5) A direct tension test should be used to evaluate the tensile
strength of layer interfaces that were the weakest part inside
the rammed earth specimens. A splitting test evaluates the
strength related to the tensile strength inside the compaction
layers because the tensile stress is mobilized inside the
compaction layers.
6) The influence of the application of cyclic loading history
was not significant. No reduction in the peak strength of the
specimens subjected to large amplitude cyclic loading was
observed.
A rammed earth structure exposed to atmosphere would
have tensile strength even if it contains no add-in materials
such as lime. In the evaluation of the seismic performance of
the rammed earth structure, the tensile strength at a layer
interface would be an important factor, which should be
evaluated by a direct tensile test. The specimen for the direct
tension test should be prepared by the manner referring to that
used in the construction of the rammed earth structure. It is
also important to prepare the specimen with a certain water
content, that which is assumed in the rammed earth structure
exposed to the atmosphere.
References
Akatani, K., Otsuka, S., Takemura, M., Tono, M., Inagawa, K., 2012. Effects
of mix proportion and construction factors on strength development of
rammed earth wall using traditional technique. Bull. Inst. Technol. (3),
42–49 (in Japanese).
Bui, Q.B., Hans, S., Morel, J.C., Do, A.P., 2011. First exploratory study on
dynamic characteristics of rammed earth buildings. Eng. Struct. 33 (12),
3690–3695.
Bui, Q.B., Morel, J.C., Hans, S., Meunier, M., 2009. Compression behaviour
of non-industrial materials in civil engineering by three scale experiments:
the case of rammed earth. Mater. Struct. 42 (8), 1101–1116.
Bui, Q.B., Morel, J.C., Hans, S., Walker, P., 2014. Effect of moisture content
on the mechanical characteristics of rammed earth. Constr. Build. Mater.
54, 163–169.
Dehua, D., Chuanmei, Z., 1999. The formation mechanism of the hydrate
phases in magnesium oxychloride cement. Cem. Concr. Res. 29 (9),
1365–1371.
Goto, S., Tatsuoka, F., Shibuya, S., Kim, Y.S., Sato, T., 1991. A simple gauge
for local small strain measurements in the laboratory. Soils Found. 31 (1),
169–180.
Hall, M., Allinson, D., 2009. Assessing the effects of soil grading on the
moisture content-dependent thermal conductivity of stabilized rammed
earth materials. Appl. Therm. Eng. 29, 740–747.
Hall, M., Djerbib, Y., 2004. Rammed earth sample production: context,
recommendations and consistency. Constr. Build. Mater. 18, 281–286.
JGS 0051-2009, Method of Classification of Geomaterials for Engineering
Purposes, Japanese Geotechnical Society.
JGS 0211-2009, Test Method for pH of Suspended Soils, Japanese Geotech-
nical Society.
JGS 0212-2009, Test Method for Electric Conductivity of Suspended Soils,
Japanese Geotechnical Society.
JGS 2551-2009, Method for Splitting Tensile Strength Test on Rocks,
Japanese Geotechnical Society.
Jaquin, P.A., Augarde, C.E., Gallipoli, D., Toll, D.G., 2009. The strength of
unstabilised rammed earth materials. Géotechnique 59 (5), 487–490.
204 H. Araki et al. / Soils and Foundations 56 (2016) 189–204
Jayasinghe, C., Kamaladasa, N., 2007. Compressive strength characteristics of
cement stabilized rammed earth walls. Constr. Build. Mater. 21,
1971–1976.
Koseki, J., Salas-Monge, R., Sato, T., 2005. Plane strain compression tests on
cement-treated sands, Geomechanics: Testing, Modeling, and Simulation,
Geotechnical Special Publication No.143, ASCE, pp. 429–443.
Koseki, J., Tsutsumi, Y., Namikawa, T., Mihira, S., Salas-Monge, R., Sano,
Y., Nakajima, S., 2008. Shear and tensile properties of cement-treated
sands and their application to liquefaction mitigation, Keynote Lecture, In:
Proceedings of 4th International Symposium on Deformation Character-
istics of Geomaterials, IS-Atlanta 2008, Vol. 1, pp. 27–50.
Lee, J., Koseki, J., Sato, T., Namikawa, T., Araki, H., 2010. Mechanical
properties of lime-treated soil and application to earthen walls, Bulletin of
Earthquake Resistant Structure Research Center, Institute of Industrial
Science. Univ. of Tokyo 43, 145–159.
Lee, J., Sato, T., Koseki, J., 2009. Effects on lime and water contents on
strength properties of a lime treated soil. In: Proceedings of International
Summer Symposium, International Activities Committee, JSCE, pp.133–
136.
Li, Z., Chau, C.K., 2007. Influence of molar ratios on properties of magnesium
oxychloride cement. Cem. Concr. Res. 37 (6), 866–870.
Minke, G., 2006. Building with Earth: Design and Technology of a Sustainable
Architecture. Birkhauser Publishers for Architecture, Basel.
Namikawa, T., Koseki, J., 2007. Evaluation of tensile strength of cement-
treated sand based on several types of laboratory tests. Soils Found. 47 (4),
657–674.
Nara Prefectural Board of Education, Office of Cultural Assets, 2009.
Summary of restoration Work at the HORYUJI SAIIN OGAKI TOMEN
(TODAIMON HOPPO), an Important Cultural Property, Nara Prefectural
Board of Education, Japan. (in Japanese).
Onitsuka, K., Chen, P., Tong, P., Negami, T., Hayakawa, K., 2007.
Geotechnical properties and construction technique of the ‘Han-chiku’
earth-filled remains located in the Yellow River Valley, China, Japanese
Geotechnical Journal, Japanese Geotechnical Society, Vol. 2, No. 4,
pp. 287–295. (in Japanese).
Rendell, F., Jauberthie, R., 2009. Performance of rammed earth structures in
east Brittany. In: Proceedings of the 11th International Conference on Non-
conventional Material and Technologies, Bath, UK, Paper No.144.
Takadachi, A., Koshihara, M., 2010. A study of seismic performance of
rammed earth wall, SEISAN KENKYU, Institute of Industrial Science,
Univ. of Tokyo, Vol. 62, No. 4 , pp. 439–443. (in Japanese).
Tanimoto, K., Takeshita, Y., Watanabe, M., Tamura, J., 2004. Consideration of
Geotechnical Characteristics of Tataki soil. In: Proceedings of 39th Annual
Conference of Japanese Geotechnical Society, pp. 803–804. (in Japanese).
UNESCO, 2014. World Heritage Centre, World Heritage List, 〈http://whc.
unesco.org/en/list/〉. (accessed on 15.1.14).
Xing, H., Yang, X., Xu, C., Ye, G., 2009. Strength characteristics and
mechanisms of salt-rich soil–cement. Eng. Geol. 103 (1), 33–38.
Yoshizawa, M., Shiomi, T., Namikawa, T., Koseki, J., Mihira, S., 2003.
Analytical study on shape of tension test specimen made of cement-treated
soil. In: Proceedings of 38th Annual Conference of Japanese Geotechnical
Society, pp.873–874. (in Japanese).